Patent Application
20210056950
Communication networks support a broad array of electronic communications among users. Text-based electronic communications may take a variety of different forms, including email, text/SMS messages, real-time/instant messages, multimedia messages, social networking messages, messages within multi-player video games, etc. Users may read and type responses to these forms of electronic communications via a personal electronic device, such as a mobile device or desktop computer.
This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter. Furthermore, the claimed subject matter is not limited to implementations that solve any or all disadvantages noted in any part of this disclosure.
In an example, a computing system obtains an electronic communication for a recipient and identifies presence of a predefined feature within the electronic communication. The computing system extracts a data subset from the electronic communication. The data subset is identified by a data extraction template selected for the predefined feature identified within the electronic communication. The computing system derives a narrative based on the data subset using an audio presentation template selected for the predefined feature identified within the electronic communication. The audio presentation template is configured to translate an aspect of the data subset into narrative form. The computing system outputs the narrative in an electronic format for audio presentation via an audio output interface in which the narrative describes the aspect of the data subset extracted from the electronic communication.
The use of text-based electronic communications such as email, text messaging, and instant messaging has grown to become a primary mode of communication in modern society. Mobile computing devices have enabled people to receive their electronic communications at virtually any time and location. As people go about their day, they may be frequently interrupted by the need or desire to review new electronic communications. Visual consumption of text and multimedia content through graphical user interfaces may distract people from simultaneously performing other tasks or may preclude people from performing tasks until after the electronic communications have been visually reviewed. For example, while operating a vehicle, it may be impossible or dangerous for a person to visually review new text-based communications.
According to an aspect of the present disclosure, the use of graphical user interfaces to present text and multimedia content of electronic communications may be augmented or replaced by audible presentation of the electronic communications in a manner that provides users with context for the presentation experience and control over the audible presentation. Such an audible presentation may provide a user experience that is commensurate with or improved over the visual consumption of the electronic communications, while enabling users to simultaneously perform tasks that are difficult or impossible to perform while using a graphical user interface. In essence, the disclosed audible presentation can translate text-based communications into an experience similar to listening to a podcast.
As shown in user speech 130, user 110 begins a dialog with device 120 by speaking the command “Read messages.” In response to the spoken command of the user, in device speech 140, device 120 outputs audio information that includes: “Hi Sam! I've got 6 conversations for you. This'll take about 5 minutes.” In this portion of device speech 140, device 120 outputs audio information in the form of natural language that greets user 110 by the user's name (i.e., “Sam”), identifies a quantity (i.e., “6”) of conversation threads that contain unreviewed electronic communications for the user as a recipient of those communications, and identifies a duration of time (i.e., “about 5 minutes”) for the user to review the conversation threads through audible output of the contents of the electronic communications. Thus, user 110 is informed by device 120 as to the anticipated duration of an audio presentation of the unreviewed electronic communications prior to progressing through the audio presentation, thereby enabling the user to make informed decisions as to whether particular electronic communications should be reviewed or skipped.
Continuing with the example dialog of
Next, device 120 outputs a summary of a first electronic communication of the first conversation thread to user 110, which identifies a relative time (i.e., “a few hours ago”) that the first electronic communication was received, identifies a sender of the first electronic communication (i.e., “Greg”), identifies a type of the first electronic communication (i.e., “email”), identifies a quantity of other recipients or the audience of the first electronic communication (i.e., “a large group”), identifies the presence of an attachment to the first electronic communication (i.e., “with an attachment”), and identifies at least a portion of text content of a message of the first electronic communication (e.g., “Goal! Can you believe it's already World Cup time? . . . ”).
In this example, upon hearing a portion of the text content of the first electronic communication, in user speech 130 user 110 speaks the command “Next conversation.” Responsive to this spoken command by user 110, device 120 advances the audio presentation of the unreviewed electronic communications to a second conversation thread, thereby skipping audio presentation of remaining unreviewed electronic communications of the first conversation thread. For example, device 120 responds to user 110 by outputting a summary of the second conversation thread, which identifies a quantity of unreviewed electronic communications of the second conversation thread (i.e., “3”), identifies a type of electronic communications of the second conversation thread (i.e., “email”), and identifies a subject of the second conversation thread (i.e., “HR incident”).
Device 120 may progress through conversation threads in the manner described above until all the unreviewed electronic communications have been reviewed by user 110 or the user preemptively elects to stop the dialog. By device 120 summarizing conversation threads and their contents, user 110 is provided with sufficient information to make informed decisions regarding whether a particular conversation thread or electronic communication should be reviewed by the user in the current session. In an example in which user 110 does not advance or interrupt audio presentation of unreviewed electronic communications, the audio presentation by device 120 will conclude in approximately the duration of time (e.g., “5 minutes”) identified by the device. However, by advancing the audio presentation, user 110 may review electronic communications within a shorter time period.
Logic subsystem 212 includes one or more physical devices (e.g., a processor) configured to execute instructions. Storage subsystem 214 includes one or more physical devices (e.g., memory) configured to hold data 220, including instructions 222 executable by logic subsystem 212 to implement the methods and operations described herein. Additional aspects of logic subsystem 212 and storage subsystem 214 are described below.
As shown in
Personal assistant machine 230 is one example of a conversational computing interface. However, a conversational computing interface may take other suitable forms. Accordingly, it will be understood that the various features and techniques described herein with reference to a personal assistant machine may be applicable to other forms of a conversational computing interface. A device, such as computing device 210 or computing system 200, that implements personal assistant machine 230 may be referred to as a personal assistant device. Similarly, a device, such as a computing device or computing system that implements another form of a conversational computing interface may be referred to as a conversational computing interface device, such as previously described with respect to device 120 of
Personal assistant machine 230 may engage in a dialog with a user by receiving and processing spoken commands of the user to perform tasks, including outputting information to the user. As an example, personal assistant machine 230 may output an audio presentation of a plurality of conversation threads and/or electronic communications for a recipient according to a presentation order. Personal assistant machine 230 may include natural language processing, thereby supporting a natural language interface by which a user may interact with computing device 210.
Speech output machine 232 receives data, such as machine-readable data and/or text-based data from personal assistant machine 230 to be output to the user, and converts such data into audio data containing speech having natural language components. In an example, speech output machine 232 may provide text-to-speech conversion. For example, personal assistant machine 230 may provide select portions of text content of an electronic communication to speech output machine 232 to convert the text content into an audible output of the text content for audible consumption by the user. In
Speech input machine 234 receives audio data representing human speech, and converts the audio data into machine-readable data and/or text data that is usable by personal assistant machine 230 or other suitable components of computing device 210. In an example, speech input machine 232 may provide speech-to-text conversion. In
The one or more communications applications 236 may support the sending and receiving of electronic communications 238, of which electronic communication 240 is an example. A communication application may support one or more types of electronic communications, including email, text/SMS messages, real-time/instant messages, multimedia messages, social networking messages, messages within multi-player video games, and/or any other type of electronic communication. Personal assistant machine 230 may interface with communications applications 236, enabling the personal assistant machine to receive, process, and send electronic communications of one or more different types on-behalf of a user.
I/O subsystem 216 may include one or more of an audio input interface 250, an audio output interface 252, a display interface 254, a communications interface 256, and/or other suitable interfaces.
Computing device 210 receives audio data representing audio captured via audio input interface 250. Audio input interface 250 may include one or more integrated audio microphones and/or may interface with one or more peripheral audio microphones. For example, computing device 210 may receive audio data representing user speech captured via audio input interface 250, such as user speech 130 of
Computing device 210 outputs audio representing audio data via audio output interface 252. Audio output interface 252 may include one or more integrated audio speakers and/or may interface with one or more peripheral audio speakers. For example, computing device 210 may output an audio representation of speech having natural language components via audio output interface 252, such as device speech 140 of
Computing device 210 may output graphical content representing graphical data via display interface 254. Display interface 254 may include one or more integrated display devices and/or may interface with one or more peripheral display devices. Display interface 254 may be omitted in at least some examples.
Computing device 210 may communicate with other devices such as server system 260 and/or other computing devices 270 via communications interface 256, enabling computing device 210 to send electronic communications to and/or receive electronic communications from the other devices. Communications interface 256 may include one or more integrated transceivers and associated communications hardware that support wireless and/or wired communications according to any suitable communications protocol. For example, communication interface 256 may be configured for communication via a wireless or wired telephone network and/or a wireless or wired personal-area network, local-area network, and/or wide-area network (e.g., the Internet, a cellular network, or a portion thereof) via a communication network 280. Communications interface 256 may be omitted in at least some examples.
I/O subsystem 216 may further include one or more additional input devices and/or output devices in integrated and/or peripheral form. Additional examples of input devices include user-input devices such as a keyboard, mouse, touch screen, touch pad, game controller, and/or inertial sensors, global positioning sensors, cameras, optical sensors. Additional examples of output devices include vibration motors and light-emitting indicators.
Computing system 200 may further include server system 260 of one or more server computing devices. Computing system 200 may further include a plurality of other computing devices 270 of which computing device 272 is an example. Server system 260 may host a communications service 262 that receives, processes, and sends electronic communications between or among senders and recipients addressed by the electronic communications. For example, users may operate computing devices 210 and 270 to send or receive electronic communications via communications service 262. Communications service 262 is depicted including a plurality of electronic communications 264 of which electronic communication 266 is an example. Electronic communication 266 may be received from computing device 272 via network 280 for processing and/or delivery to computing device 210 via network 280 in an example. One or more of communications applications 236 may be configured for coordinated operation with communications service 262 enabling electronic communications to be sent, received, and/or processed for senders and recipients as users of computing devices 210 and 270.
In an example, a user acting as a sender of electronic communication 300 may define, through user input, one or more of recipients 312, subject 316 including text content 318, message 320 including text content 322 and/or media content 324, attachments 326, calendar data 328, and/or other data 332 of electronic communication 300. Timestamp 314 may be assigned by the communications application or communications service as a timing of transmission or reception of electronic communication 300. Communication type 330 may be dependent upon the communications application or service used by the sender, or may be defined or otherwise selected by user input of the sender in the case of a communications application or service that supports multiple communication types.
At 410, a greeting may be presented as an audible output. In an example, the greeting may be presented in response to an instruction 412 received by the personal assistant device to initiate presentation of unreviewed electronic communications for a recipient. Instruction 412 may take the form of a spoken command by a user or other type of user input received by the personal assistant device. For example, in
At 414, a presentation roadmap may be presented as an audible output. The presentation roadmap may identify one or more of: a quantity of conversation threads that include one or more unreviewed electronic communications for the recipient, a quantity of the unreviewed electronic communications, a time estimate of the audio presentation for presenting the conversation threads that include unreviewed electronic communications, a length estimate of the unreviewed electronic communications, one or more highlighted items, and/or other suitable information. The one or more highlighted items may include a narrative that describes the contents of one or more of the unreviewed electronic communications (see e.g.,
At 416, a barge-in notice may be presented as an audible output. The barge-in notice may be used to notify the user that a spoken command may be provided by the user to perform an action with respect to the audio presentation or its contents. Referring to the example of
At 418, one or more changes to the day of the user may be presented as an audible output. Changes to the day may include updates to the user's calendar, and optionally may be derived from calendar data of one or more of the unreviewed electronic communications.
As described in further detail with reference to
A first conversation thread that includes one or more unreviewed electronic communications for the user may be presented at 470, including a conversation thread summary 420 of the first conversation thread, a communication summary 422 for each unreviewed electronic communication of the first conversation thread, and message content 424 for each unreviewed electronic communication of the first conversation thread.
At 420, the conversation thread summary for the first conversation thread may be presented as an audible output. The conversation thread summary may identify one or more of: a subject of the conversation thread as identified from an electronic communication of the conversation thread, a type of the electronic communications of the conversation thread, a quantity of unreviewed electronic communications of the conversation thread, recipients and/or an audience (e.g., quantity, identities of the recipients, and/or a quantity/identity of recipients that were added or removed in relation to prior reply-linked communications) of the conversation thread as identified from an electronic communication of the conversation thread, a time estimate of a portion of the audio presentation for presenting the unreviewed electronic communications of the conversation thread, a length estimate of the unreviewed electronic communications of the conversation thread, a narrative that describes the contents of one or more of the unreviewed electronic communications of the conversation thread (see e.g.,
Example outputs by the personal assistant device with respect to a quantity of unreviewed electronic communications of a conversation thread are described in further detail with reference to
At 422, a first communication summary for a first unreviewed electronic communication of the first conversation thread may be presented as an audible output. The communication summary may identify one or more of a subject of the electronic communication, a type of the electronic communication, a timing of the electronic communication based on the timestamp of the electronic communication, a sender of the electronic communication, recipients and/or an audience of the electronic communication, a time estimate of a portion of the audio presentation for presenting the electronic communication, a length estimate of the electronic communication, an indication of whether one or more attachments are included with the electronic communication, a narrative that describes the contents of the electronic communication (see e.g.,
At 424, message content of the first unreviewed electronic communication of the first conversation thread may be presented as an audible output. For example, an audible output of the text content of the message of the first unreviewed electronic communication may be presented in part or in its entirety at 424. In
Following presentation of the first unreviewed electronic communication, the audio presentation may proceed to a second unreviewed electronic communication of the first conversation thread. For example, at 426, a second communication summary for a second unreviewed electronic communication of the first conversation thread may be presented as an audible output. At 428, message content of the second unreviewed electronic communication of the first conversation thread may be presented as an audible output. The audio presentation may proceed sequentially through each unreviewed electronic communication of the first conversation thread. In at least some examples, the unreviewed electronic communications of a conversation thread may be presented according to a chronological-sequential order based on the respective timestamps of the unreviewed electronic communications, beginning with the oldest unreviewed electronic communication and continuing through to the most recent unreviewed electronic communication of the conversation thread.
At 430, a guided notice may be presented as an audible output. The guided notice may be used to inquire whether the user would like to perform an action with respect to the first conversation thread. As an example, the guided notice may provide a general notice to the user, such as “perform an action or proceed to the next conversation?” or may provide targeted notices, such as “would you like to reply to this conversation?”. At 432, a silent period may be provided to enable the user to provide an instruction or otherwise take action with respect to the conversation thread before proceeding to the next conversation thread of the audio presentation.
Following presentation of the first conversation thread at 470, the audio presentation may proceed to presenting a second conversation thread at 472 that includes one or more unreviewed electronic communications for the recipient. Presentation of the second conversation thread may similarly include presentation of a thread summary for the second conversation thread at 440, a communication summary for a first unreviewed electronic communication of the second conversation thread at 442, message content of the first unreviewed electronic communication of the second conversation thread at 444, a communication summary for a second unreviewed electronic communication of the second conversation thread at 446, message content of the second unreviewed electronic communication of the second conversation thread at 448, etc., until each unreviewed electronic communication of the second conversation thread has been presented as an audible output.
The audio presentation may proceed through each conversation thread that includes one or more unreviewed electronic communications for the recipient, as previously described with reference to the presentation of the first conversation thread at 470. Following presentation of the conversation threads that included one or more unreviewed electronic communications, additional information that the personal assistant device determines as being potentially relevant to the user may be presented as an audible output at 460. At 462, the user may be signed-off from the audio presentation session by the personal assistant device.
Continuing with the example timeline of
By organizing electronic communications into conversation threads, a user may perform an action with respect to the electronic communications of that conversation thread. For example, as described above, a user may skip audio presentation of a particular conversation thread, including the unreviewed electronic communications of the conversation thread, by providing a spoken command, such as “Next conversation” of
In at least some examples, audible indicators may be presented as an audible output by the personal assistant device to notify the user of transitions between portions of the audio presentation. For example, audible indicator 482 may be presented between presentation of the changes to the day at 418 and the thread summary at 420, audible indicators 484 and 490 may be presented between electronic communications, audible indicators 486 and 492 may be presented between the guided notice and the silent period, and audible indicators 488 and 494 may be presented between the silent period and a subsequent conversation thread and the additional information presented at 460 or sign-off at 462. An audible indicator may take the form of an audible tone or any suitable sound. Audible indicators having distinguishable sounds may be presented at different portions of the audio presentation. For example, audible indicator 484 identifying a transition between electronic communications may differ from audible indicator 488 identifying a transition between conversation threads. Such audible indicators may help a user easily understand whether the personal assistant device has started or completed a particular portion of the audio presentation, whether the personal assistant device has completed a particular action as instructed by the user, or whether the personal assistant device is currently listening for an instruction to be provided by the user.
The personal assistant device may support various presentation modes, including a continuous presentation mode and a guided presentation mode. In the continuous presentation mode, the personal assistant device may proceed through the audio presentation in the absence of instructions from the user. In the guided presentation mode, the personal assistant device may pause the audio presentation at transition points to await an instruction from the user to proceed. For example, in the guided presentation mode, the personal assistant device may pause the audio presentation and output the inquiry: “Would you like to hear this conversation thread” following presentation of the conversation summary.
Timeline 500 is divided into multiple conversation threads 510-520, each including one or more electronic communications of a recipient. In this example, conversation thread 510 includes electronic communications 530-540, conversation thread 512 includes electronic communications 550-558, conversation thread 514 includes electronic communications 560-564, conversation thread 516 includes electronic communication 570, conversation thread 518 includes electronic communication 580, conversation thread 520 includes electronic communications 590-594.
Multiple electronic communications of a conversation thread may be referred to as being reply-linked electronic communications in which one or more electronic communications are replies to an original electronic communication, thereby linking these electronic communications to each other by a common conversation thread. A first electronic communication that is a reply to an earlier second electronic communication which in-turn is a reply to an even earlier third electronic communication may be considered as being reply-linked to both the second and third electronic communications, thereby forming a common conversation thread. For example, electronic communication 534 is a reply to electronic communication 532, which in-turn is a reply to electronic communication 530. Therefore, each of electronic communications 530, 532, and 534 form part of conversation thread 510. For some types of electronic communications, such as collaborative messaging platforms or multi-player gaming platforms, electronic communications associated with a particular channel (e.g., a particular collaborative project or multi-player game) may be identified as being reply-linked to each other.
Furthermore, in this example, electronic communications 530-540, 554-558, 560-564, 570, and 594 are unreviewed electronic communications of the recipient. By contrast, electronic communications 550, 552, 580, and 590 are previously reviewed electronic communications of the recipient. In an example, an electronic communication may be referred to as an unreviewed electronic communication if its message (e.g., message 320 of
As described with reference to the example dialog between user 110 and device 120 of
In a first example presentation order, conversation threads may be presented according to a reverse chronological-sequential order based on a latest unreviewed electronic communication of each conversation thread. In the example timeline of
Returning to
In a third example presentation order, conversation threads may be presented according to a reverse chronological-sequential order based on a timing of an earliest unreviewed electronic communication of each conversation thread. In the example timeline of
In a fourth example presentation order, conversation threads may be presented according to a chronological-sequential order based on a timing of an earliest unreviewed electronic communication of each conversation thread. In the example timeline of
In a fifth example presentation order, conversation threads that include a reply by the recipient at some point within the thread may be prioritized in the presentation order over conversation threads that do not include a reply by the recipient. In the example timeline of
In a sixth example presentation order, prioritization of conversation threads having a reply by the recipient, such as described above with respect to the fifth example presentation order, may consider only those replies by the recipient for which an unreviewed electronic communication is a reply directly to that reply of the recipient. This presentation order may be used to prioritize conversation threads that include unreviewed electronic communications that are directly reply-linked to replies of the recipient over other conversation threads.
In a seventh example presentation order, conversation threads may be prioritized based on one or more factors, including the content of the subject, message, or attachment of the electronic communications, the senders of the electronic communications, the quantity of electronic communications per conversation thread, the frequency of electronic communications per conversation thread, or the presence of importance indicators (e.g., flags) associated with the electronic communications. In an example, conversation threads may be ranked according to the one or more factors, and may be presented in an order that is based on the ranking of the conversation threads. Such ranking may be based on any desired heuristics, machine learning algorithms, or other ranking methodologies.
At 710, electronic communications are obtained for a recipient. In an example, the electronic communications may be obtained at a computing device of a user from a remote server system via a communications network. The electronic communications obtained for the recipient at 710 may span one or more types of electronic communications, and may be collected from one or more communications services and/or applications. Furthermore, the electronic communications obtained at 710 may refer to a subset of all electronic communications of the recipient. For example, the electronic communications obtained at 710 may include a primary or focused inbox or folder of the recipient, and may exclude other inboxes or folders such as junk mail, promotions, etc.
At 712, unreviewed electronic communications are identified for the recipient among the electronic communication obtained at 710. As previously described with reference to
At 714, electronic communications obtained at 710 are organized according to a schema. The schema may be programmatically defined by one or more of a communications application of the user's computing device, by a communications service of a server system, or by a personal assistant machine, depending on implementation. For example, some communications services or applications may organize or partially organize electronic communications into conversation threads, whereas other communications services or applications may not support the use of conversation threads.
At 716, electronic communications obtained at 710 may be grouped into a plurality of conversation threads containing two or more reply-linked electronic communications. As previously described, two or more electronic communications are reply-linked if an electronic communication is a reply to an earlier electronic communication, and that an electronic communication may be reply-linked to an earlier electronic communication by one or more intermediate reply-linked electronic communications. Following operation 716, each conversation thread includes two or more electronic communications for the recipient that are reply-linked to each other. However, it will be understood that at least some conversation threads may include an individual electronic communication. At 718, data representing the grouping of electronic communications may be stored for each conversation thread. For example, data representing the grouping from operation 716 may be stored in a storage subsystem of a computing device, including locally at the user's computing device and/or at a remote server system.
At 720, electronic communications of each conversation thread may be ordered in chronological order according to a timestamp indicating a timing of each electronic communication. At 722, data representing the ordering of electronic communications may be stored for each conversation thread. For example, data representing the ordering from operation 722 may be stored in a storage subsystem of a computing device, including locally at the user's computing device and/or at a remote server system.
At 724, the conversation threads may be ordered based on a rule to obtain a presentation order among the conversation threads. As previously described with reference to the presentation order examples of
At 728, an instruction to initiate audio presentation of the electronic communications for the recipient is received. The instruction may take the form of a spoken command by a user, such as previously described with reference to
At 730, responsive to the instruction received at 728, an audio presentation of the conversation threads is output according to the presentation order obtained at operation 724. The presentation order may be defined by one or more of the grouping of electronic communications at 716, the ordering of electronic communications at 720, and the ordering of conversation threads at 724, and may be based on the data stored at 718, 722, and 726.
In an example, the audio presentation includes unreviewed electronic communications of each conversation thread in a chronological-sequential order beginning with an oldest unreviewed electronic communication and continuing to a most recent unreviewed electronic communication of the conversation thread before another of the plurality of conversation threads that includes an unreviewed electronic communication that is interspersed in time between the oldest unreviewed electronic communication and the most recent unreviewed electronic communication of the conversation thread. For example, at 732, two or more unreviewed electronic communications of a first conversation thread are audibly output according a chronological sequential order before unreviewed electronic communications of a second conversation thread at 734.
Furthermore, in an example, the presentation order of the conversation threads may be a reverse chronological-sequential order based on a most recent unreviewed electronic communication of each of the plurality of conversation threads such that the first conversation thread having a first most recent unreviewed electronic communication is presented at 732 before the second conversation thread having a second most recent unreviewed electronic communication that is older than the first most recent unreviewed electronic communication of the plurality of conversation threads. An example of this reverse chronological-sequential order is described with reference to
The audio presentation output at 730 may include, for each unreviewed electronic communication, at least a portion of text content of a message of the unreviewed electronic communication presented as an audible output. In an example, all text content of the message of the unreviewed electronic communication may be presented as an audible output. Furthermore, in at least some examples, the audio presentation further includes, for each conversation thread of the plurality of conversation threads, a thread summary of the conversation thread presented as an audible output before the text content of the conversation thread. Examples of thread summaries presented before message content are described with reference to
At 740, a second instruction to advance the audio presentation may be received. The instruction received at 740 may take the form of a spoken command of a user, such as previously described with reference to
At 742, responsive to the second instruction, the audio presentation of the plurality of conversation threads may be advanced from a current conversation thread to a subsequent conversation thread of the presentation order. It will be understood that other forms of navigation within the audio presentation may be supported by the personal assistant device, including ending the audio presentation, restarting the audio presentation, skipping to a next conversation thread, skipping to a particular conversation thread identified by the user, skipping a next unreviewed electronic communication, skipping to a particular unreviewed electronic communication identified by the user, etc.
The action of advancing audio presentation with respect to a conversation thread is one of a plurality of actions that may be supported by the personal assistant device. For example, operation 740 may instead include an instruction to perform a different action, such as replying to, forwarding on to another recipient, storing, or deleting the conversation thread, or marking the conversation thread as important (e.g., flagging the conversation thread or an electronic communication thereof). For at least some types of action, responsive to the instruction to perform the action, the action may be applied to each electronic communication of the conversation thread by the personal assistant device at 742. A spoken command used to initiate a particular action by the personal assistant device may include one or more keywords that are predefined at and recognizable by the personal assistant device, or an intent of a spoken utterance may be inferred by the personal assistant device from context, such as previously described with reference to the instruction received at 728.
At 752, an instruction may be received. For example, the instruction received at 752 may correspond to the instruction received at 728 of
At 758, each unreviewed electronic communication in the most-recent conversation thread may be audibly output in a chronological-sequential order beginning with an oldest unreviewed electronic communication at 760. Audibly outputting the oldest unreviewed electronic communication at 760 may include audibly outputting the communication summary at 762 and audibly outputting some or all of the text content of the message at 764. However, the communication summary may not be audibly output in other examples.
At 766, if more unreviewed electronic communications are in the conversation thread, the method returns to 760 where the oldest unreviewed electronic communication is audibly output. Accordingly, the method continues to a most-recent unreviewed electronic communication, such as previously described with reference to the example presentation order of
At 766, if there are no more unreviewed electronic communications in the conversation thread, the method proceeds to 768. At 768, if there are more conversation threads that include unreviewed electronic communications, the method may return to 754 where the next-most recent conversation thread is audibly output at 754. Accordingly, responsive to completing audible output of the most-recent unreviewed electronic communication from a conversation thread, the method includes audibly outputting each unreviewed electronic communication in a next-most-recent conversation thread including a next-most-recent set of unreviewed, reply-linked electronic communications for the recipient. Each unreviewed electronic communication in the next-most-recent conversation thread is audibly output at 758 in a chronological-sequential order beginning with an oldest unreviewed electronic communication and continuing to a most-recent unreviewed electronic communication.
As described, for example, with reference to
At 810, the method includes receiving an instruction to initiate audio presentation of electronic communications for a recipient. As previously described with reference to operation 728 of
At 812, electronic communications for the recipient are obtained. As previously described with reference to operation 710 of
At 814, unreviewed electronic communications for the recipient are identified. As previously described with reference to
At 816, an estimated time is determined to present a portion of an audio presentation in which the portion includes audible output of text content of the unreviewed electronic communications for the recipient. The text content may include the text content of a message of each unreviewed electronic communication. As an example, the estimated time is determined based on a feature of the text content of the plurality of unreviewed electronic communications. The feature of the text content may include a word count or a character count of the text content, as examples; and the time estimate may be algorithmically computed based on the word or character count (e.g., 0.7 seconds per word). As another example, the method may further include converting the text content of the plurality of unreviewed electronic communications into audio data representing the audible output of the text content, determining the estimated time to present the subsequent portion of the audio presentation based on a feature of the audio data. The feature of the audio data may include an amount (e.g., a byte count) of the audio data or a duration of the audio data at a target presentation rate, as examples.
The estimate time may be determined based on other information contained in the audio presentation that is to be audibly output by the personal assistant device in the subsequent portion. For example, where the audio presentation includes thread summaries for each conversation thread, the estimated time may be determined further based on the duration of the thread summaries within the subsequent portion of the audio presentation.
In at least some examples, the estimated time identified by the presentation road map may take the form of a generalized time estimate.
At 818, the audio presentation is output responsive to the instruction. Outputting the audio presentation includes outputting an initial portion of the audio presentation that includes a presentation road map 820, and a subsequent portion that includes the audible output of the text content of the plurality of unreviewed electronic communications for the recipient. In an example, the presentation road map output at 820 identifies the estimated time to present the subsequent portion of the audio presentation output at operation 822, which corresponds to the portion for which the estimate time was determined at operation 816.
The presentation road map output at 818 may identify other features of the audio presentation, such as previously described with reference to
Aspects of method 800 may be similarly performed to present an estimated time in a thread summary for a conversation thread containing one or more reply-linked electronic communications or for a communication summary of an individual electronic communication, such as described with reference to
As introduced above, there are many scenarios in which it may be beneficial for a user to audibly review electronic communications. However, most electronic communications are not designed for audible presentation. To the contrary, electronic communications have conventionally been visually presented using a graphical user interface. Merely outputting a verbatim audible reproduction of certain types of message content may not provide a favorable user experience. Verbatim reproductions may be incomprehensible and/or longer in duration than desired. For example, large data tables may take an impractically long time to read verbatim and may be difficult to comprehend. As another example, long Uniform Resource Locators (URLs) may not provide useful information when read verbatim. As described herein, certain types of message content may be summarized, simplified, and/or omitted to improve the user experience of consuming the message content in audible form. Furthermore, a user may optionally be provided with a notice of features that may not be suitable for audio presentation, thus allowing the user to make an informed decision as to whether a particular electronic communication instead should be visually reviewed at a later time.
Personal assistant machine 910 obtains an electronic communication 920 for a recipient, such as previously described with reference to operations 710 of
Electronic communication 300 of
Personal assistant machine 910 may perform feature identification 930 using feature identification machine 932 to identify presence of a predefined feature 924 within data 922 of electronic communication 920. The feature identification machine may be configured to identify features that can be summarized, simplified, omitted, and/or otherwise modified in an audible narrative so as to improve a user experience when listening to the audible narrative. In at least some examples, feature identification machine 932 may use one or more feature definitions 936 to search for and identify one or more instances of predefined feature 924 within electronic communication 920. As an illustrative example, feature definition 936 may define file formats of media content that are to be identified within a message portion of electronic communication 920. As described in more detail below, a feature identification machine 932 may use a plurality of different feature definitions 936 to identify a corresponding plurality of different types of features.
Each feature definition may describe one or more characteristics of a corresponding predefined feature, such as a location within a data structure of the electronic communication where the predefined feature may be found (e.g., message body, subject, sender, recipient(s), attachments, and/or other data), a type or data format of the predefined feature (e.g., an image, a video, a weblink, or a text object represented by a particular markup language or file extension), a data size of the predefined feature, a data signature or structure of the predefined feature, or other suitable features.
Feature definition 936 may form part of a feature library 934 that includes a plurality of feature definitions 938. Feature library 934 may be included as part of feature identification machine 932 or may be referenced by the feature identification machine from a data storage subsystem. Each of the plurality of feature definitions 938 may be associated with a corresponding predefined feature that is identifiable by feature identification machine 932. For example, one feature definition may be directed to identifying the presence of hyperlinks within the message body of the electronic communication, while another feature definition may be directed to identifying a particular type of sender of the electronic communication for which other data may be extracted from the message body. Feature identification 930 may be performed to identify multiple predefined features 926 within electronic communication 920. Examples of predefined features that may be identifiable by feature identification machine 932 include graphical media content, text content, a language (e.g., English, Spanish, or Japanese) of the text content, spatial arrays of text content (e.g., tables and charts), URLs and/or other network addresses of corresponding network resources, shipment confirmation content, event scheduling content, or other suitable features.
In at least some examples, each of a plurality of predefined features defined by feature library 934 may be used to find different types of features. However, feature definitions of feature library 934 optionally may be filtered to obtain a subset of feature definitions used to perform a feature search. Additionally or alternatively, feature identification machine 932 may implement a hierarchy among the feature definitions. For example, the feature identification machine may utilize one or more hierarchically-lower feature definitions only if a hierarchically-higher feature definition successfully identifies a corresponding predefined feature. In at least some examples, a separate processing pipeline thread may be created for each feature definition of feature library 934, thereby enabling multiple processing pipeline threads to coexist.
In some examples, feature identification machine 932 may include one or more artificial intelligence and/or machine learning classifiers configured to identify a particular type of feature based on previous machine-learning training. As a nonlimiting example, the feature identification machine may include an artificial neural network configured to identify a particular feature, such as foreign language text. In some examples, two or more separately trained classifiers may be configured to look for a corresponding two or more different types of features within data 922.
Personal assistant machine 910 may perform data extraction 940 using data extraction machine 942 for each predefined feature identified in electronic communication 920 to extract a corresponding data subset from the electronic communication for that predefined feature. The data subset extracted for predefined feature 924 may include the predetermined feature (or a portion thereof) and/or other data 928 of the electronic communication. For example, upon identifying a predefined feature that includes a predefined sender of a shipment confirmation, shipping information contained in other data 928 may be extracted by data extraction machine 942.
In at least some examples, the data subset may be identified within the entire collection of data that defines electronic communication 920 by using a data extraction template 946 selected by data extraction machine 942 or other component of personal assistant machine 910. Data extraction templates such as template 946 may be configured to prioritize data extracted from the electronic communication in a manner that reduces or minimizes an amount of time needed to audibly present the corresponding narrative as compared to the original electronic communication. Data extraction template 946 may identify a plurality of data items for the data subset to be extracted from data 922 of electronic communication 920 and may further include a definition for each data item that is similar to the previously described feature definition 936. For example, a definition in the extraction template for each data item to be extracted may describe one or more characteristics of a corresponding feature of that data item, such as a location within a data structure of the electronic communication where the data item may be found, a type or format of the data item, a size of the data item, a signature or structure of the data item, or other suitable features.
Data extraction template 946 may be one of a plurality of data extraction templates 948 of a data extraction template library 944. Data extraction template library 944 may be included as part of data extraction machine 940 or may be referenced by the data extraction machine from a data storage subsystem. Each of the plurality of data extraction templates 948 may be associated with a corresponding predefined feature that is identifiable by feature identification machine 932. In other words, a predefined feature 924 (e.g., a URL) identified using a particular feature definition 936 (e.g., URL definition) may be subsequently processed with the data extraction template 946 (e.g., URL extraction template) associated with that feature definition 936. Accordingly, data extraction 940 may be performed to identify a plurality of data subsets within electronic communication 920 that each correspond to one or more of the plurality of predefined features 926 identified within the electronic communication.
In some examples, data extraction machine 942 may include one or more artificial intelligence and/or machine learning models configured to extract, from an electronic communication, a data subset corresponding to an identified feature based on previous machine-learning training. As a nonlimiting example, the data extraction machine may include an artificial convolutional neural network configured to extract a particular data subset, such as a string of foreign language text.
Personal assistant machine 910 may perform data analysis 950 using data analysis machine 952 to obtain a processed form of the data subset extracted by data extraction 940. The processed form of the data subset may be used to select and/or populate a downstream audio presentation template as part of narrative creation 960. Data analysis 950 may include combining one or more data subsets obtained by data extraction 940 with third-party sources of data, such as user preferences, user profile information, or other suitable data. Myriad different forms of data analysis may be performed by data analysis machine 952. For example, data analysis may include determining a number of instances of a particular type of predefined feature within the electronic communication and assigning a score to the electronic communication based on the number of instances. As another example, data analysis 950 may include determining an incomprehensibility score for electronic communication 920 that is based on one or more data subsets extracted from the electronic communication. An example of this incomprehensibility score is described in further detail with reference to
In at least some examples, the data subset may be analyzed using a data analysis template 956 selected by data analysis machine 952 or other component of personal assistant machine 910 for the predefined feature 924 that was identified within electronic communication 920 by feature identification 930. For example, each of a plurality of data analysis templates may be associated with a corresponding predefined feature that is identifiable by feature identification machine 932. In other words, data extracted for a predefined feature 924 (e.g., a URL) identified using a particular feature definition 936 (e.g., URL definition) may be subsequently processed with the data analysis template 956 (e.g., a URL analysis template) associated with that feature definition 936. For example, a data analysis template for a URL may define an algorithm for identifying a primary domain within the URL to be included in the narrative. Data analysis template 956 may be further selected based on the data subset extracted from the electronic communication by data extraction 940 for predefined feature 930. For example, extracted data that includes images within a message body may be analyzed using a different data analysis template than other visual items, such as data tables. Data analysis template 956 may include one or more algorithms for processing one or more data subsets extracted by data extraction 940 to obtain a processed form of that data.
Data analysis template 956 may be one of a plurality of data analysis templates 958 of a data analysis template library 954. Data analysis template library 954 may be included as part of data analysis machine 952 or may be referenced by the data analysis machine from a data storage subsystem. Each of the plurality of data analysis templates 958 may be associated with a corresponding predefined feature that is identifiable by feature identification machine 932. Accordingly, data analysis 950 may be performed to obtain a processed form of one or more data subsets corresponding to one or more predefined features identified within electronic communication 920.
Personal assistant machine 910 may perform narrative creation 960 using narrative creation machine 962 to derive a narrative 972 that describes an aspect of one or more data subsets extracted from electronic communication 920 by data extraction 940 and/or processed forms of such data obtained by data analysis 950 for one or more predefined features of the electronic communication. In at least some examples, narrative 972 may be derived by narrative creation machine 962 using an audio presentation template 966 that is configured to translate an aspect of the data subset and/or processed forms thereof into narrative form. Each of a plurality of audio presentation templates may be associated with a corresponding predefined feature that is identifiable by feature identification machine 932. In other words, data extracted for a predefined feature 924 (e.g., a URL) identified using a particular feature definition 936 (e.g., URL definition) that has been processed with the data analysis template 956 (e.g., a URL analysis template) to obtain a processed form of the extracted data (e.g., the primary domain within the URL) may be incorporated into a narrative using the audio presentation template associated with the feature definition. For example, an audio presentation template for a URL, as an example of a predefined feature identified within an electronic communication, may define a component of the narrative as including: “The message includes a link to [insert the primary domain within the URL]”. The audio presentation template, in cooperation with corresponding data extraction, is configured to prioritize salient aspects of a communication so that the salient aspects may be presented in a comprehensible and time-efficient manner. As such, in cooperation with corresponding data extraction, the audio presentation template may functionally redact some to most of the original message content to produce the narrative for audible presentation and/or present words and/or phrases not included in the original message content.
As an illustrative example, narrative 972 may include the natural language phrase—“An hour ago you received a message confirming that your shipment from ACME Corporation will be delivered to your home tomorrow, July 30th”. In this example, predefined feature 924 identified by feature identification 930 may include a sender of electronic communication 920 that is one of a predefined list of shipment confirmation senders defined by feature definition 936. A data subset extracted by data extraction 940 for the predefined feature may include a name identifying the shipment confirmation sender (“ACME Corporation”), a delivery date (“July 30th”), and a delivery location (a mailing address associated with the recipient's home) located within a message portion of electronic communication 920. Processed forms of the data subset obtained by data analysis 950 of this data subset may incorporate data from third-party sources, such as user preferences, user profile information, or other suitable data to include a delivery location (“home”) that is associated with the mailing address of the user within the data subset and a relative delivery date (“tomorrow”) that is associated with a delivery date of the data subset (“July 30th”) extracted from the electronic communication, user profile information, and calendar information. In other words, the processed data may use one or more additional signals (e.g., a user profile including the user's home address) to convert data (e.g., a physical address) into a more user-friendly narrative (e.g., saying “home” in the narrative instead of saying “11222 Dilling St, Studio City, Calif. 91602”).
In at least some examples, audio presentation template 966 may include one or more predefined natural language statements and one or more data fields defined in relation to the natural language statements that collectively form narrative 972. Narrative creation machine 962 may incorporate, into the one or more data fields, one or more data subsets extracted from electronic communication 920 by data extraction 940 and/or processed forms of such data obtained by data analysis 950. Continuing with the above shipping confirmation example, audio presentation template 966 may include the predefined natural language statement “you received a message confirming that your shipment from” followed by the data field represented in the above example by a name (“ACME Corporation”) identifying the shipment confirmation sender extracted by data extraction 940.
Audio presentation template 966 may be one of a plurality of audio presentation templates 968 of a data presentation template library 964. Data presentation template library 964 may be included as part of narrative creation machine 960 or may be referenced by the narrative creation machine from a data storage subsystem. Each of the plurality of audio presentation templates 968 may be associated with a corresponding predefined feature that is identifiable by feature identification machine 932. Audio presentation template 966 may be selected by narrative creation machine 962 or other component of personal assistant machine 910 for the predefined feature 924 that was identified within the electronic communication by feature identification 930.
In at least some examples, audio presentation template 966 may be further selected from two or more audio presentation templates associated with predefined feature 924 for the data subset that was extracted by data extraction 940 and/or the processed forms of such data that was obtained by data analysis 950. Each audio presentation template may include one or more conditions against which the data subset extracted by data extraction 940 and/or the processed forms of such data obtained by data analysis 950 may be judged for selection of that audio presentation template.
As an example, data analysis 950 may include determining an incomprehensibility score for an electronic communication, and an audio presentation template may be selected from two or more audio presentation templates based on the value of the incomprehensibility score being below or above a threshold value. In this example, a first audio presentation template may define a first narrative that includes a qualitative description of the incomprehensibility score, such as “this message will be difficult to understand”, while a second audio presentation template may instead define a second narrative that audibly describes an aspect of the content without providing the qualitative description, such as “Tom wishes you ‘happy birthday’ and includes an image”.
Personal assistant machine 910 outputs narrative 972 in an electronic format for inclusion in an audio presentation 970. Audio presentation 970 is an example of the audio presentation previously described with reference to
At 1010, the method includes obtaining an electronic communication for a recipient, such as previously described with reference to operations 710 of
At 1012, the method includes identifying presence of a predefined feature (e.g., 924 of
As an illustrative example, the predefined feature may include graphical media content within the electronic communication. The graphical media content may include one or more graphical media content items, such as images and/or videos. Graphical media content may further include spatial arrays of text content, such as tables or charts. Graphical media content items may be included in a message portion of the electronic communication, such as in-line with text content of a message body. Additionally or alternatively, the graphical media content items may be included as attachments to the electronic communications.
At 1014, the method includes extracting a data subset from the electronic communication that is identified by a data extraction template selected for the predefined feature identified within the electronic communication. In at least some examples, the data extraction template is selected for the predefined feature from a plurality of data extraction templates, such as previously described with reference to data extraction template library 944.
At 1016, the method includes analyzing the data subset extracted at 1014. In at least some examples, analyzing the data subset at 1016 may include determining an incomprehensibility score for the electronic communication.
Continuing with the above example where the predefined feature includes graphical media content within the electronic communication, the incomprehensibility score may be determined based, at least in part, on a relative amount of graphical media content as compared to text content identified within the electronic communication. For example, the incomprehensibility score may be based on a quantity of one or more graphical content items (e.g., images or videos) and a quantity of one or more text objects (e.g., words) identified as being present within a message portion of the electronic communication by identifying presence of each graphical media content item and each text object within the message portion of the electronic communication. In at least some examples, the quantity of text objects has a relationship to the incomprehensibility score that is an inverse of the quantity of the graphical media content items. For example, as the quantity of graphical media content items increases in relation to the quantity of text objects, the incomprehensibility score may increase to indicate that audio presentation of the electronic communication will be less comprehensible to users. Conversely, as the quantity of graphical media content items decreases in relation to the quantity of text objects, the incomprehensibility score may decrease to indicate that audio presentation of the electronic communication will be more comprehensible to users.
At 1018, the method includes deriving a narrative based on the data subset and/or processed forms of such data using an audio presentation template selected for the predefined feature identified within the electronic communication. As previously described with reference to
In at least some examples, the aspect described by the narrative may include a notice that at least a portion of the electronic communication cannot be audibly presented or is unlikely to be comprehended by the user if output in audible form. For example, visual media content, network addresses for network resources, text content in a language other than the recipient's preferred language, etc. may not be capable of being audibly output as part of an audio presentation of the electronic communication or may be unlikely to be audibly comprehended by the user. The narrative may alternatively or additionally identify the portion of the electronic communication that is not included in the audio presentation by a class identifier (e.g., image, video, network address, or foreign language text) and/or a title of that portion (e.g., image title, video title, or primary domain of the network address).
In a first example where the predefined feature identified at 1012 includes graphical media content within the electronic communication, the aspect described by the narrative may identify a media type of the graphical media content, such as an image, video, or spatial array of text content such as a table or chart. An example narrative that describes the presence of graphical content in a message body of an electronic communication includes: “A few hours ago, Tom sent an email about ‘Status report’ that includes some visual content in the message, so I'll read what I can. It says, ‘Hi team, I need to get this broadly distributed. Look for your name and assignment here.’ Then there's a table, then it goes on, ‘Please let me know if you have any additional questions.’ That's the entire message.” By providing a notice of the presence of the table in the above example rather than audibly outputting the text content of the table, the time needed to audibly review the electronic communication may be reduced and/or the message content may be more clearly understood by the user. Continuing with the above example, salient information that is relevant to the user (as the recipient of the electronic communication) may be extracted from the data table, and a portion of that information may be included in the narrative that is audibly presented to the user. For example, a data table may include text in the form of a list of work shifts for multiple people, and data analysis performed at 1016 may include identifying the name of the user within a cell of the data table and extracting a row and/or column of data from the data table that contains that cell. The data extracted from the row and/or column may be incorporated into a narrative that is then audibly output to the user, such as “The table included in the message includes your name followed by ‘Saturday’ within a day field of the table, and ‘3 pm to 6 pm’ within a time field of the table”.
Alternatively or additionally, the aspect described by the narrative may identify an estimated audio comprehensibility for graphical content that is based on the incomprehensibility score determined at 1016. For example, the estimated audio comprehensibility may include a quantitative and/or a qualitative description of the incomprehensibility score determined at 1016. An example of a qualitative description of the incomprehensibility score may include a notice, such as a warning or a suggestion that the message portion of the electronic communication contains visual content or that the message portion should be reviewed visually via a graphical display device rather than being audibly output as part of the audio presentation. For example, a narrative may include: “About an hour ago, John emailed about ‘puppy extravaganza’ to you and Erik. There's a lot of visual content, so you'll want to view this on a screen.” Further examples of the use of an incomprehensibility score are described with reference to
In a second example where the predefined feature identified at 1012 includes text content, the aspect described by the narrative may identify text content that is not in a preferred language of the recipient. For example, the predefined feature may include text content having a different language (e.g., Japanese) than the preferred language (e.g., English) of the recipient, and the aspect described by the narrative may identify presence of the text content having the different language and/or may indicate the different language. Spelling errors within text content that are indecipherable by the computing system may be characterized as a different language from the preferred language of the recipient.
The incomprehensibility score described in further detail with reference to
In an example, a narrative may include: “You received a message from Tim, but it includes a lot of Japanese text, so I'll flag this message so you can review it later.” In this example, the relative proportion of Japanese text to the user's preferred language text may be judged to be sufficiently high to avoid audibly outputting the text content of the message. In another example, a narrative may include: “You received a message that includes only a few words in Japanese, so I'll attempt to read the message. Please interrupt if you want me to skip this message”. In this example, the relative proportion of Japanese text to the user's preferred language text may be less than the preceding example, and the narrative may inform the user of the presence of content that may make an audible presentation of such content incomprehensible.
By reducing an amount of text content that is audibly output to the user by removal of foreign language or indecipherable content from the audio presentation, an amount of time needed to audibly review the electronic communication may be reduced. However, in other examples, where translation to the recipient's preferred language is supported, the narrative may indicate that the translated portion of the text that is audibly output in the audio presentation has been translated from the foreign language, and may further identify the foreign language. For example, a narrative may include “You received a message from Tim that includes a portion of text that was automatically translated for you from Japanese, it says . . . .”
In a third example where the predefined feature identified at 1012 includes a network address (e.g., weblink or file path) of a network resource, the aspect described by the narrative may identify a primary domain of the network address while excluding one or more subdomains and/or a preamble portion of the network address from the narrative or from the audio presentation. For example, given the network address “https://products.office.com/en-us/outlook/email-and-calendar-software-microsoft-outlook”, the narrative may include “office.com” or “products.office.com” while excluding one or more subdomains “en-us/outlook/email-and-calendar-software-microsoft-outlook” and/or the preamble portion “https://”. Examples of narratives for file paths related to shared documents include: “Amy shared ‘latest mocks for Sam’ with you on OneDrive”, where OneDrive™ represents the primary domain of the sharing service; and “OneDrive sent you a reminder, Amy shared ‘latest mocks for Sam’ with you.’. By reducing a length of the network address audibly output to the user, the network address included in the message content may be more clearly understood by the user and/or an amount of time needed to audibly review the electronic communication may be reduced. Additionally or alternatively, the incomprehensibility score described in further detail with reference to
In a fourth example where the predefined feature includes a sender of the electronic communication that is one of a predefined list of shipment confirmation senders, the aspect described by the narrative may identify the shipment confirmation sender, and one or more of a delivery date, delivery status, delivery location, contents of a shipment, and/or other suitable shipment-related information. As previously described with reference to
In a fifth example where the predefined feature includes event scheduling content, the aspect described by the narrative may identify an organizer of an event indicated by the event scheduling content, and may further identify one or more of an event date, an event name, an event location, and/or other suitable event-related information. Event scheduling content may include calendar data of the electronic communication and/or text content data in a message portion of the electronic communication, as examples. Event scheduling content may be identified by a sender of the electronic communication in at least some examples. Event scheduling content may include transportation-related events, such as airline flight reservations, train or bus reservation, etc. Examples of narratives describing event scheduling content of an electronic communication in simplified form include: “Conference Room 10005 is reserved for “Team sync” at 10 AM tomorrow.”; “Tom declined your invitation to ‘customer review’ at 9:30 AM on Friday.”; “Sam invited you to ‘Let's get lunch!’ at 12:30 PM tomorrow, and commented, ‘Can't wait to see you!’.”; “Paul extended ‘Brainstorming session’, which is happening tomorrow at 2 PM, to two hours, and added ‘Conference Room 125’ as the location.”; and “Your flight to Boston tomorrow is delayed one hour, and will be leaving at 1 pm instead of 12 pm”. By extracting the event scheduling content from the electronic communication, event scheduling information may be more clearly conveyed to the user by audio presentation and/or an amount of time needed to audibly review the electronic communication may be reduced.
At 1020, the method includes outputting the narrative in an electronic format for audio presentation via an audio output interface in which the narrative describes the aspect of the data subset or processed form thereof that was extracted from the electronic communication. In at least some examples, the electronic format including the narrative may be stored in a data storage subsystem from which the narrative may be later accessed and converted into audible natural language speech at the time of the audio presentation, such as by speech output machine 232 of
At 1022, the method includes receiving an instruction to initiate the audio presentation of the electronic communication for the recipient via a client computing device of the computing system, such as previously described with reference to operations 728 of
At 1024, the method includes transmitting the narrative in the electronic format to a remote client computing device via a communications network to output the audio presentation including the narrative via an audio output interface of the remote client computing device. Transmitting the narrative in an electronic format at 1024 may be performed by a server system or other computing device responsive to the instruction received at 1022. However, in other examples, operation 1024 may be omitted, such as where operation 1020 is performed locally at the client computing device.
At 1026, the method includes outputting the audio presentation including the narrative via the audio output interface of the client computing device. Outputting the audio presentation at 1026 may be performed responsive to the instruction received at 1022. In at least some examples, the narrative may be presented as part of the communication summary of the electronic communication, such as example communication summary 422 of
During an audio presentation of electronic communications, a user may not be looking at a graphical display or a graphical display may be inaccessible. However, the contents of some electronic communications may make little or no sense to users when graphical content or other features of the electronic communication are omitted from the audio presentation. In some cases, the conversational computing interface disclosed herein may determine an incomprehensibility score for an electronic communication, and may notify the user of electronic communications having contents that may be more suitably reviewed using a graphical display device as opposed to being audibly output as part of an audio presentation.
At 1110, the method includes quantifying visual objects in a message portion (i.e., message body) of the electronic communication and optionally attachments to the electronic communication. Visual objects may include graphical media content items, such as images or videos, and/or spatial arrays containing text content, such as tables or charts. For example, a quantity of graphical media content items (imageCount) may be determined at 1112 for images and/or videos within the message portion. In at least some examples, graphical media content items of larger than a predefined pixel area (e.g., 50×50 pixels) are counted, while graphical media content items that do not exceed the predefined pixel area may not be counted. Graphical media content items, such as icons or graphical signature blocks that are programmatically applied on behalf of the sender of the electronic communication are often smaller than other graphical media content items incorporated as message content or attachments. Therefore, a size of the predefined pixel area may be selected to exclude graphical media content items such as icons or graphical signature blocks from being considered when determining an incomprehensibility score, since these items need not be audibly output for the recipient to comprehend the content of the electronic communication, as intended by the sender. At 1114, a quantity of spatial arrays of text content (tableCount) may be determined for spatial arrays within the message portion. In an example, a spatial array may be defined as being larger than a 1×1 cell.
In at least some examples, the method may further include, at 1116, determining a quantity of network addresses of network resources (addressCount), such as weblinks, file paths, etc. that are contained within the message portion.
At 1118, an amount of text content contained in the message portion of the electronic communication may be quantified. For example, a quantity of text items (textCount), such as words within the message portion may be determined. However, other suitable techniques may be used to quantity text content, such as character count, phoneme count, etc. In at least some examples, textCount may include words that are in a preferred language of the recipient and omit words that are not identified as being in the preferred language.
At 1120, an incomprehensibility score (IncomprehensibilityScore) is calculated based on the textCount and one or more of the imageCount, tableCount, and addressCount values identified at 1112-1118. In an example that assumes a range for IncomprehensibilityScore is bounded by 0 and 100, in which 100 represents an incomprehensible audio presentation of the electronic communication, if imageCount>2, or tableCount>1, then IncomprehensibilityScore may be defined to equal 100, and the electronic communication is thereby classified as being inappropriate for audio presentation. However, other suitable values or thresholds may be used. Otherwise, if imageCount≤2 and tableCount≤1, then an IncomprehensibleLength score may be calculated using the following expression: 10*tableCount+10*imageCount+10*addressCount=IncomprehensibleLength. The incomprehensibility score may be calculated using the following expression: 100*IncomprehensibleLength/(IncomprehensibleLength+textCount)=IncomprehensibilityScore. However, other suitable values or thresholds may be used.
At 1122, the method includes determining if the incomprehensibility score exceeds one or more thresholds. As previously described with reference to
Aspects of method 1100 may be used to determine an incomprehensibility score for text content of a message portion of an electronic communication that is in a language that differs from a preferred language of the recipient. For example, the recipient may set a preferred language (e.g., English) within a settings field of a communications application or communications service, such as application 236 or service 262 of
In at least some examples, the personal assistant device may utilize one or more conversation templates configured to implement the logic of method 700. For example, the timeline of
In
In at least some examples, the methods and processes described herein may be tied to a computing system of one or more computing devices. In particular, such methods and processes may be implemented as a computer-application program or service, an application-programming interface (API), a library, and/or other computer-program product.
Referring again to
Logic subsystem 212 may include one or more processors configured to execute software instructions. Additionally or alternatively, the logic subsystem may include one or more hardware or firmware logic circuits configured to execute hardware or firmware instructions. Processors of the logic subsystem may be single-core or multi-core, and the instructions executed thereon may be configured for sequential, parallel, and/or distributed processing. Individual components of the logic subsystem optionally may be distributed among two or more separate devices, which may be remotely located and/or configured for coordinated processing. Aspects of the logic subsystem may be virtualized and executed by remotely accessible, networked computing devices configured in a cloud-computing configuration.
Storage subsystem 214 may include removable and/or built-in devices. Storage subsystem 214 may include optical memory (e.g., CD, DVD, HD-DVD, or Blu-Ray Disc), semiconductor memory (e.g., RAM, EPROM, or EEPROM), and/or magnetic memory (e.g., hard-disk drive, floppy-disk drive, tape drive, or MRAM), among others. Storage subsystem 214 may include volatile, nonvolatile, dynamic, static, read/write, read-only, random-access, sequential-access, location-addressable, file-addressable, and/or content-addressable devices. It will be appreciated that storage subsystem 214 includes one or more physical devices and is not merely an electromagnetic signal, an optical signal, etc. that is not held by a physical device for a finite duration.
Aspects of logic subsystem 212 and storage subsystem 214 may be integrated together into one or more hardware-logic components. Such hardware-logic components may include field-programmable gate arrays (FPGAs), program- and application-specific integrated circuits (PASIC/ASICs), program- and application-specific standard products (PSSP/ASSPs), system-on-a-chip (SOC), and complex programmable logic devices (CPLDs), for example.
When the methods and operations described herein are implemented by logic subsystem 212 and storage subsystem 214, a state of storage subsystem 214 may be transformed—e.g., to hold different data. For example, logic subsystem 212 may be configured to execute instructions 222 that are part of one or more applications, services, programs, routines, libraries, objects, components, data structures, or other logical constructs. Such instructions may be implemented to perform a task, implement a data type, transform the state of one or more components, achieve a technical effect, or otherwise arrive at a desired result.
The logic subsystem and the storage subsystem may cooperate to instantiate one or more logic machines, such as previously described with reference to personal assistant machine 230, speech output machine 232, speech input machine 234, feature identification machine 932, data extraction machine 942, data analysis machine 952, and narrative creation machine 962. It will be understood that the “machines” as described herein (e.g., with reference to
Machines may be implemented using any suitable combination of state-of-the-art and/or future machine learning (ML), artificial intelligence (AI), and/or natural language processing (NLP) techniques. Non-limiting examples of techniques that may be incorporated in an implementation of one or more machines include support vector machines, multi-layer neural networks, convolutional neural networks (e.g., including spatial convolutional networks for processing images and/or videos, temporal convolutional neural networks for processing audio signals and/or natural language sentences, and/or any other suitable convolutional neural networks configured to convolve and pool features across one or more temporal and/or spatial dimensions), recurrent neural networks (e.g., long short-term memory networks), associative memories (e.g., lookup tables, hash tables, Bloom Filters, Neural Turing Machine and/or Neural Random Access Memory), word embedding models (e.g., GloVe or Word2Vec), unsupervised spatial and/or clustering methods (e.g., nearest neighbor algorithms, topological data analysis, and/or k-means clustering), graphical models (e.g., (hidden) Markov models, Markov random fields, (hidden) conditional random fields, and/or AT knowledge bases), and/or natural language processing techniques (e.g., tokenization, stemming, constituency and/or dependency parsing, and/or intent recognition, segmental models, and/or super-segmental models (e.g., hidden dynamic models)).
In some examples, the methods and processes described herein may be implemented using one or more differentiable functions, wherein a gradient of the differentiable functions may be calculated and/or estimated with regard to inputs and/or outputs of the differentiable functions (e.g., with regard to training data, and/or with regard to an objective function). Such methods and processes may be at least partially determined by a set of trainable parameters. Accordingly, the trainable parameters for a particular method or process may be adjusted through any suitable training procedure, in order to continually improve functioning of the method or process.
Non-limiting examples of training procedures for adjusting trainable parameters include supervised training (e.g., using gradient descent or any other suitable optimization method), zero-shot, few-shot, unsupervised learning methods (e.g., classification based on classes derived from unsupervised clustering methods), reinforcement learning (e.g., deep Q learning based on feedback) and/or generative adversarial neural network training methods, belief propagation, RANSAC (random sample consensus), contextual bandit methods, maximum likelihood methods, and/or expectation maximization. In some examples, a plurality of methods, processes, and/or components of systems described herein may be trained simultaneously with regard to an objective function measuring performance of collective functioning of the plurality of components (e.g., with regard to reinforcement feedback and/or with regard to labelled training data). Simultaneously training the plurality of methods, processes, and/or components may improve such collective functioning. In some examples, one or more methods, processes, and/or components may be trained independently of other components (e.g., offline training on historical data).
Language models may utilize vocabulary features to guide sampling/searching for words for recognition of speech. For example, a language model may be at least partially defined by a statistical distribution of words or other vocabulary features. For example, a language model may be defined by a statistical distribution of n-grams, defining transition probabilities between candidate words according to vocabulary statistics. The language model may be further based on any other appropriate statistical features, and/or results of processing the statistical features with one or more machine learning and/or statistical algorithms (e.g., confidence values resulting from such processing). In some examples, a statistical model may constrain what words may be recognized for an audio signal, e.g., based on an assumption that words in the audio signal come from a particular vocabulary.
Alternately or additionally, the language model may be based on one or more neural networks previously trained to represent audio inputs and words in a shared latent space, e.g., a vector space learned by one or more audio and/or word models (e.g., wav2letter and/or word2vec). Accordingly, finding a candidate word may include searching the shared latent space based on a vector encoded by the audio model for an audio input, in order to find a candidate word vector for decoding with the word model. The shared latent space may be utilized to assess, for one or more candidate words, a confidence that the candidate word is featured in the speech audio.
The language model may be used in conjunction with an acoustical model configured to assess, for a candidate word and an audio signal, a confidence that the candidate word is included in speech audio in the audio signal based on acoustical features of the word (e.g., mel-frequency cepstral coefficients, or formants). Optionally, in some examples, the language model may incorporate the acoustical model (e.g., assessment and/or training of the language model may be based on the acoustical model). The acoustical model defines a mapping between acoustic signals and basic sound units such as phonemes, e.g., based on labelled speech audio. The acoustical model may be based on any suitable combination of state-of-the-art or future machine learning (ML) and/or artificial intelligence (AI) models, for example: deep neural networks (e.g., long short-term memory, temporal convolutional neural network, restricted Boltzmann machine, deep belief network), hidden Markov models (HMM), conditional random fields (CRF) and/or Markov random fields, Gaussian mixture models, and/or other graphical models (e.g., deep Bayesian network). Audio signals to be processed with the acoustic model may be pre-processed in any suitable manner, e.g., encoding at any suitable sampling rate, Fourier transform, or band-pass filters. The acoustical model may be trained to recognize the mapping between acoustic signals and sound units based on training with labelled audio data. For example, the acoustical model may be trained based on labelled audio data comprising speech audio and corrected text, in order to learn the mapping between the speech audio signals and sound units denoted by the corrected text. Accordingly, the acoustical model may be continually improved to improve its utility for correctly recognizing speech audio.
In some examples, in addition to statistical models, neural networks, and/or acoustical models, the language model may incorporate any suitable graphical model, e.g., a hidden Markov model (HMM) or a conditional random field (CRF). The graphical model may utilize statistical features (e.g., transition probabilities) and/or confidence values to determine a probability of recognizing a word, given the speech audio and/or other words recognized so far. Accordingly, the graphical model may utilize the statistical features, previously trained machine learning models, and/or acoustical models to define transition probabilities between states represented in the graphical model.
In at least some examples, I/O subsystem 216 may include or interface with selected natural user input (NUI) componentry. Such componentry may be integrated or peripheral, and the transduction and/or processing of input actions may be handled on- or off-board. Example NUI componentry may include a microphone for speech and/or voice recognition; an infrared, color, stereoscopic, and/or depth camera for machine vision and/or gesture recognition; a head tracker, eye tracker, accelerometer, and/or gyroscope for motion detection and/or intent recognition; as well as electric-field sensing componentry for assessing brain activity.
It will be appreciated that a “service”, as used herein, is an application program executable across multiple user sessions. A service may be available to one or more system components, programs, and/or other services. In some implementations, a service may run on one or more server-computing devices.
According to an example of the present disclosure, a method performed by a computing system comprises: obtaining an electronic communication for a recipient; identifying presence of a predefined feature within the electronic communication; extracting a data subset from the electronic communication, the data subset identified by a data extraction template selected for the predefined feature identified within the electronic communication; deriving a narrative based on the data subset using an audio presentation template selected for the predefined feature identified within the electronic communication, the audio presentation template configured to translate an aspect of the data subset into narrative form; and outputting the narrative in an electronic format for audio presentation via an audio output interface, the narrative describing the aspect of the data subset extracted from the electronic communication. In this example or any other example disclosed herein, the predefined feature is one of a plurality of predefined features identifiable by the computing system; and each predefined feature is associated with a corresponding data extraction template and a corresponding audio presentation template. In this example or any other example disclosed herein, the predefined feature includes graphical media content within the electronic communication; and the method further comprises: identifying presence of each graphical media content item of the graphical media content within the electronic communication; and determining an incomprehensibility score for the electronic communication based on a quantity of one or more graphical media content items identified as being present within the electronic communication; wherein the narrative further describes an estimated audio comprehensibility based on the incomprehensibility score. In this example or any other example disclosed herein, the estimated audio comprehensibility described by the narrative includes a qualitative description of the incomprehensibility score. In this example or any other example disclosed herein, the predefined feature is one of a plurality of predefined features, including the graphical media content and further including text content of a message portion of the electronic communication; and the method further comprises: identifying presence of each text object of the text content within the message portion of the electronic communication; and determining the incomprehensibility score for the electronic communication further based on a quantity of one or more text objects identified as being present within the message portion of the electronic communication; wherein the quantity of the one or more text objects has a relationship to the incomprehensibility score that is an inverse of the quantity of the one or more graphical media content items. In this example or any other example disclosed herein, the predefined feature includes graphical media content within the electronic communication; and the aspect described by the narrative identifies a media type of the graphical media content, the media type including an image, a video, or a spatial array of text content. In this example or any other example disclosed herein, the predefined feature includes a network address of a network resource; and the aspect described by the narrative identifies a primary domain of the network address while excluding one or more subdomains of the network address from the narrative. In this example or any other example disclosed herein, the aspect described by the narrative includes a notice that at least a portion of the electronic communication cannot be audibly presented; and the narrative identifies the portion of the electronic communication by a class identifier. In this example or any other example disclosed herein, the predefined feature includes text content having a different language than a preferred language of the recipient; and the aspect described by the narrative identifies presence of the text content having the different language and/or indicates the different language. In this example or any other example disclosed herein, the method further comprises: receiving an instruction to initiate the audio presentation of the electronic communication for the recipient via a client computing device of the computing system; and responsive to the instruction, outputting the audio presentation of the electronic format including the narrative via the audio output interface of the client computing device. In this example or any other example disclosed herein, the narrative precedes presentation of a message portion of the electronic communication within the audio presentation. In this example or any other example disclosed herein, the narrative forms part of the message portion of the electronic communication within the audio presentation and summarizes at least some of the message portion. In this example or any other example disclosed herein, the narrative forms part of a presentation road map that precedes presentation of a plurality of unreviewed electronic communications for the recipient within the audio presentation, in which the plurality of unreviewed electronic communications include at least the electronic communication obtained for the recipient. In this example or any other example disclosed herein, the narrative forms part of a thread summary that precedes presentation of a plurality of unreviewed, reply-linked electronic communications for the recipient within the audio presentation, in which the plurality of unreviewed, reply-linked electronic communications include at least the electronic communication obtained for the recipient. In this example or any other example disclosed herein, the method further comprises: transmitting the narrative in the electronic format to a remote client computing device via a communications network to output the audio presentation of the electronic format including the narrative via the audio output interface of the remote client computing device.
According to another example of the present disclosure, a computing system comprises: an audio output interface to output audio via one or more audio speakers; a logic subsystem; and a storage subsystem having instructions stored thereon executable by the logic subsystem to: obtain an electronic communication for a recipient; identify presence of a predefined feature within the electronic communication; extract a data subset from the electronic communication, the data subset identified by a data extraction template selected for the predefined feature identified within the electronic communication; derive a narrative based on the data subset using an audio presentation template selected for the predefined feature identified within the electronic communication, the audio presentation template configured to translate an aspect of the data subset into narrative form; and output the narrative in an electronic format for audio presentation via an audio output interface, the narrative describing the aspect of the data subset extracted from the electronic communication. In this example or any other example disclosed herein, the predefined feature is one of a plurality of predefined features identifiable by the computing system; and each predefined feature is associated with a corresponding data extraction template and a corresponding audio presentation template. In this example or any other example disclosed herein, the predefined feature includes graphical media content within the electronic communication; and the instructions are further executable by the logic subsystem to: identify presence of each graphical media content item of the graphical media content within the electronic communication; and determine an incomprehensibility score for the electronic communication based on a quantity of one or more graphical media content items identified as being present within the electronic communication; wherein the narrative further describes an estimated audio comprehensibility based on the incomprehensibility score. In this example or any other example disclosed herein, the predefined feature is one of a plurality of predefined features, including the graphical media content and further including text content of a message portion of the electronic communication; and the instructions are further executable by the logic subsystem to: identify presence of each text object of the text content within the message portion of the electronic communication; and determine the incomprehensibility score for the electronic communication further based on a quantity of one or more text objects identified as being present within the message portion of the electronic communication; wherein the quantity of the one or more text objects has a relationship to the incomprehensibility score that is an inverse of the quantity of the one or more graphical media content items.
According to another example of the present disclosure, a method performed by a computing system comprises: obtaining an electronic communication for a recipient; identifying presence of a predefined feature within the electronic communication, wherein the predefined feature is one of a plurality of predefined features identifiable by the computing system, wherein each predefined feature is associated with a corresponding data extraction template selected from a plurality of data extraction templates, and wherein each predefined feature is further associated with a corresponding audio presentation template selected from a plurality of audio presentation templates; extracting a data subset from the electronic communication, the data subset identified by the corresponding data extraction template; deriving a narrative based on the data subset using the corresponding audio presentation template, the audio presentation template configured to translate an aspect of the data subset into narrative form; and outputting the narrative as part of an audio presentation via an audio output interface, the narrative describing the aspect of the data subset extracted from the electronic communication and preceding message content of the electronic communication audibly presented in the audio presentation.
It will be understood that the configurations and/or approaches described herein are exemplary in nature, and that these specific embodiments or examples are not to be considered in a limiting sense, because numerous variations are possible. The specific routines or methods described herein may represent one or more of any number of processing strategies. As such, various acts illustrated and/or described may be performed in the sequence illustrated and/or described, in other sequences, in parallel, or omitted. Likewise, the order of the above-described processes may be changed.
The subject matter of the present disclosure includes all novel and non-obvious combinations and sub-combinations of the various processes, systems and configurations, and other features, functions, acts, and/or properties disclosed herein, as well as any and all equivalents thereof.
