Patent Application
20140067795
The disclosed technology relates generally to searching, filtering and matching and, more particularly, to sophisticated tools and techniques to be used in connection with various types of searches, filterings and matches including, but not limited to, operations spanning multiple data sources and/or online operations.
For many people, Google is synonymous with online searching due to its impressive 60% share of the Internet search market. What is less well understood is the fact that most of the world's approximately 800 billion gigabytes of data is inaccessible to any Internet search engine because it resides in inaccessible databases, is unstructured and therefore not machine readable, or requires more search effort to find than its perceived value, a problem that will be referred to herein as search friction.
Search friction is made worse in several ways: by the rapid increase in searchable data made possible by engines like Google being applied to unstructured data, by the increase in data accessibility via the web, and by the fundamental increase in amount of data being created and stored. Search friction is minimal when a search process works well and becomes intolerably large when the process entirely fails to achieve desired results. The 2010 Nobel Prize for Economics was awarded to scholars who identified search friction as a major economic problem for employment and other important markets.
Existing data construct models underlying user queries and data store design possess inherent limitations impacting how data is captured, retrieved, analyzed and presented. Current systems can only support two types of information: text (and text strings) and numbers combined using very simple logic: AND, OR, NOT, <, >, =, and mathematical functions. These constrained data constructs lead to a number of significant challenges.
Current information systems lack the ability to express and model targets, preferences, importance, beliefs, and estimates—the fundamental parts of human discourse—when leveraging technology to search, match, filter, forecast, evaluate and decide. Further, with increasing movement toward crowd sourcing and social networks, there is a correspondingly growing lack of ability to express and fuse multiple targets, preferences, importance, beliefs, and estimates.
E-commerce solutions have suffered many limitations over the years. For example, many search tools only allow searches for specifically defined items. Said search tools provide no mechanism for the user to express uncertainty in their query and provide no mechanism to express or capture uncertainty in the stored data. Indeed, search tools that do not recognize these uncertainties provide significantly fewer useful results, if any, to users.
For example, employers typically want to find the best possible job candidates. If they set search requirements too tight, the search result is usually empty. If they set search requirements too loosely, however, then there are often too many resumes to read and too little guidance for closure. On the flip side, job seekers generally want to find the best possible position. If they set their desires to high, they will likely find no positions. If they set their desires too low, however, they generally undervalue themselves.
Current systems tend to limit both employers and prospective employees to expressing information as deterministic, i.e., single-valued values. While some information includes single-valued, demographic facts, e.g., age, years in last job, etc., much important information that expresses both parties' preferences, targets, importance, beliefs, and estimates, are not expressed well, if at all, in this manner. Similar limitations persist in other areas such as social networking applications and websites, for example.
Thus, there remains a need for improved searching and search-related tools and techniques, particularly with regard to matching engines as pertaining to areas such as employment and social networking.
Today's social networking sites often provide users with an overwhelming amount of information via a social stream. Unfortunately, current filters typically fail to perform in accordance with users' desires. For example, some filters only put the most recent entries first and other filters only put the most popular items at the top. Certain filters merely apply simple keyword filtering techniques. These mechanisms, whether performed singly or in combination with each other, do not meet today's user demands and expectations because they lack the ability to enable users to filter and order social streams based on user targets, preferences, and importance.
Email is another area that would greatly benefit from improved filter mechanisms. For example, email providers and applications such as Yahoo and Microsoft Outlook merely provide simple filters such as filtering messages by information in the “From” and/or “To” fields as well as “begins with,” “ends with,” or “contains” with regard to the body or other portions of the message. Google may allow a user to personalize his or her news as received from Google but Google does not order a newsfeed by users' choices. Email applications lack the ability to enable users to filter and order messages based on user targets, preferences, and importance.
Thus, there remains a need for improved filtering tools and techniques, particularly with regard to social networking applications, email applications, and websites.
Traditional situation awareness systems are configured to allow users and companies to manage sales opportunities by creating and linking opportunity records to specific accounts, and track opportunities, including percent chance of close, opportunity value, and all associated data. User-defined fields may be added to each opportunity record and interactions, such as follow-up phone calls and onsite meetings, may be scheduled into an existing calendar. Such systems may maintain a complete history of activities with specific notes about each opportunity. Various items such as appointments, tasks, notes, documents, e-mails, and activities to specific sales opportunities may all be created and linked, and reports on sales funnel and opportunity progress may be generated and disseminated accordingly.
However, true situation awareness requires a complete perception of the environment, something that can only be attained by searching for information in data stores described by preferences, targets, and importance, and capturing estimates and beliefs of subject matter experts, colleagues, and other available information sources. Also, obtaining such information is only the beginning—systems and people must be able to adequately analyze and comprehend the meaning of such information by understanding uncertainty and risk inherent in the results and understanding what additional information is needed and the associated cost/benefit trade-off, for example. Current systems fail to do any of this, let alone project understanding into the future to anticipate what might happen and using the information to support choosing a course of action.
Thus, there remains a need for improved situation awareness tools and techniques.
Thus, there remains a need for improved searching and search-related tools and techniques, particularly with regard to unstructured data.
Implementations of the disclosed technology pertain to various types of searching and search-related tools and techniques. Embodiments may serve to revolutionize how information systems capture, retrieve, analyze, and present quantitative and qualitative data and also offer unique and significant value across a number of vertical segments that may be very large. In contrast to traditional information systems that are capable of analyzing only text and numbers in binary fashion with the constraints of AND, OR, NOT, =, <, > relationships between data elements, implementations of the disclosed technology may be leveraged to manage uncertain data constructs including targets, preferences, importance, beliefs and estimates in addition to text and numbers, and to include the input or objectives of multiple people.
The system 100 also includes three mobile electronic devices 108-112. Two of the mobile electronic devices 108 and 110 are communications devices such as cellular telephones or smartphones. Another of the mobile devices 112 is a handheld computing device such as a personal digital assistant (PDA) or tablet device. A remote storage device 114 may store some of all of the data that is accessed and used by any of the computers 104 and 106 or mobile electronic devices 108-112.
The electronic device 200 includes a housing 202, a display 204 in association with the housing 202, a user interaction mechanism 206 in association with the housing 202, a processor 208 within the housing 202, and a memory 210 within the housing 202. The user interaction mechanism 206 may include a physical device, such as a keyboard, mouse, microphone, speaking, or any combination thereof, or a virtual device, such as a virtual keypad implemented within a touchscreen. The processor 208 may perform any of a number of various operations. The memory 210 may store information used by or resulting from processing performed by the processor 208.
Certain embodiments of the disclosed technology directly address the issue of search friction by closing the gap between search objectives and the results generated by the search process. This may be accomplished by enabling users to accurately describe their search objectives and by assuring that those objectives are accurately reflected in the search results. Such embodiments may also provide superior record identification, search analysis, results visualization, feedback, and tuning capabilities that support an improved discovery process. This may serve to address the search friction problem of efficiency of information discovery when the desired data represents a small portion of all data searched.
As used herein, a structured database generally refers to a database of records that is either created directly by human, machines, or techniques that impose structure upon unstructured data. These records may describe information or data about an object (e.g., a person, his or her beliefs, a physical entity, or an organization) or they may describe a search objective, e.g., a description of the data desired to be found. In some cases, a record may have both data and search objective(s) as described below.
In a standard search, the typical effort is to find data records that match a search objective. Each record is usually composed of fields of information where each field describes a characteristic of the record that is either textual or numerical. A search objective generally refers to a description of a field in terms of text strings or numerical values that describe the records that are being sought. In traditional database searches, a record's value must perfectly match one or more of the search objective values for the record to be identified as a match. If there are multiple fields, then matches must occur in all of the fields for the record to be identified as completely matching the search objective. This type of “match” search is generally brittle resulting in search friction.
As used herein, the term “brittle” generally refers to the inability of a search process to comprehensively and/or accurately describe the desired search result. This typically leads to any of a number of undesirable situations such as the generation of a null set of search results, only partially desired search results, or a large number of search results that must be manually inspected or otherwise processed to find desired results.
The more brittle a search process is, the larger the gap will be between the desired search results and the actual search results, and the more search friction there will be. Certain embodiments of the disclosed technology may reduce or eliminate the brittleness of search processes and, therefore, significantly reduce search friction.
It should be noted that the data in a record or a search objective may be the result of the fusion of multiple people's input or the un-fused search objectives of a group of people. This concept is described in detail below.
Certain embodiments of the disclosed technology include the calculation of a field satisfaction (e.g., a value between 0.0 and 1.0 for each field and subsequent application of algorithms to combine these into an overall record satisfaction. Record satisfactions may range between 0.0 and 1.0, where 1.0 generally refers to a complete match and 0.0 generally refers to a complete non-match or miss. The record(s) may be ranked by record satisfaction.
In situations where there is no complete match, i.e., each record has a satisfaction less than 1.0, the record(s) with the highest satisfaction may be selected. These embodiments typically result in better defined search objectives and enable better feedback to refine search objectives. In cases where search objectives are not well known, such embodiments may provide results that enable fast feedback for every stage of the information discovery process, thus reducing or eliminating search friction by removing the brittle search problem.
In structured data, there are four main categories of information. A first type of data as used herein will be referred to as textual data and may include, but is not limited to, one-option data, multiple-option data, independent N-option data, and dependent N-option data. One-option data generally refers to a single idea or statement of fact. Multiple-option data, e.g., two-option data, generally refers to multiple ideas or statements of fact, such as “male or female,” “yes or no,” and “like or dislike,” for example. Independent N-option data generally refers to N options that are independent of each other, e.g. list of cities, zip codes, type, or other cataloging characteristic. Dependent N-option data generally refers to N options that are linearly related by a generally understood and/or communicated characteristic such as goodness, probability, agreement, frequency, quality, etc., e.g., “Very Bad,” “Bad,” “Normal,” “Good,” and “Very Good.”
Textual data may be captured by receiving a selection of structured options, e.g., instructing a user to “check Male or Female,” “select Yes or No,” or “enter zip code.” Textual data may also be captured by the input of recognizable test strings such as a zip code, for example, or by the result of categorizing unstructured data.
Numerical data generally refers to numerical information or characteristics having units associated therewith such as age, length, or cost, for example. Numerical data may be captured by receiving a selection of structured options, e.g., instructing a user to “select height” and providing selections such as 52-56, 57-59, etc., for example. Numerical data may also be captured by the input of recognizable test strings such as height or weight, for example, or by the result of categorizing unstructured data.
Two other types of data described herein will be referred to as “beliefs” and “estimates.” Belief and numerical estimate data types generally include a certainty or other qualifying data characteristics along with the primary data characteristic. Current systems reduce data to simple textual and numerical data representations. For example, some market research data capture may describe, and be limited to, demographic characteristics such as age or gender, and other measures that can be represented only as textual and numerical data. As used herein, a primary objective of that data capture may be the beliefs and estimates of respondents so that the response to current or future stimulus may be represented.
As used herein, a belief generally refers to an opinion expressed as data, e.g., a statement of some idea or principal that is represented by textual data and a certainty, knowledge, intensity, or existence characteristic. In an example where a user indicates “I am a Democrat but not a very strong one,” the “not a very strong one” portion of the data string is an N-option classification of certainty. In an example where a user indicates that “I am sure the moon is made of cheese,” the usage of “sure” represents a two-option classification, e.g., the moon is either made of cheese or something else. If a user indicates that “the Big Bang theory explains the origins of the universe and I believe it is true,” the usage of “believe” represents an N-option classification that depends on how many options are available. Consider a numerical data example in which a user indicates that “I estimate the project will cost about $1200, maybe as little as $900, and maybe as much as $1350.” In this example, use of the term “about” represents the user's certainty with respect to the project cost.
As used herein, an estimate generally refers to a forecast or projection of a quantitative measure that is uncertain. An estimate may be represented by numerical data with the addition of a certainty. For example, if one field of data is the cost of a proposed project, the value may be tagged with a certainty of the estimate. Such data is typically a distribution of values that can take any format ranging from simple uniform distributions to complex probabilistic functions that describe the probability of every point in the range of numerical possibilities.
A belief map may be implemented in connection with belief information collection and visualization. Belief maps may be used for both the collection of and visualization of data. In certain embodiments, a user may move one or more dots on his or her belief map to capture his or her opinion on a particular matter. Multiple inputs may then be displayed to enable the user to visualize assessments across option, criteria, or people. This visualization may serve to improve the management of search results. Belief maps are described in U.S. Pat. No. 6,631,362, titled “GENERAL DECISION-MAKING SUPPORT METHOD AND SYSTEM” and issued on Oct. 7, 2003, the content of which is hereby incorporated by reference herein in its entirety.
In the example, the horizontal axis of the belief map 700 represents certainty, knowledge, intensity, and/or existence. The interpretation as to what type of qualifying data is represented may be an application-specific labeling consideration to avoid confusion, but the underlying certainty treatment would remain unchanged. A point at the far right of the belief map 700 would indicate strong certainty, meaning that it is based on good knowledge or believed with strong intensity. If at the far left of the belief map 700, however, the point would be no better than the flip of a coin because it would correspond to uncertainty, e.g., based on weak knowledge or not strongly believed.
In certain implementations, the location of a belief map or slider dot is translated into a single “belief” number. This value thus simplifies the two-dimensional belief map data representation into a certainty-weighted single dimensional data format. Consider an example in which C=a criteria satisfaction that ranges from 0.0 (i.e., no support) to 1.0 (i.e., full support) and K=a knowledge value that ranges from 0.5 (i.e., the flip of a coin) to 1.0 (i.e., high certainty). The belief B may be determined using the following equation:
B=K*C+(1−K)*(1−C)
B=1−2*(X−X2) for X>0.5
B=2*(X−X2) for X<=0.5
In the example, X is the value shown on the corresponding slider 1100 of
Traditional searches often include boxes that can be checked to indicate whether a specific field value is to be searched for or not. For each box that is checked, the corresponding value is searched for and, if the box is not checked, the corresponding value will not be searched for. Consider an example in which a user wants to search a database for a specific person who is believed to live in Portland but may live in Vancouver and probably does not live in Seattle.
In the example, each box in the interface 1200 is either checked or not with a search result showing everyone that matches the checked characteristic. This search is “brittle” because if “Portland” is the only box that is checked, for example, then only data records that have “Portland” will be found. There is no way to indicate the search objective preference for Portland followed by Vancouver and a lack of search desire for Seattle. The contrast between desired search objective and that which is possible to input illustrates why this type of match search is brittle.
With the format provided by the interface 1300, the search objective for city can be described by moving the sliders to represent the preference for each option. In the example, the default input is placed at “don't care,” which may be equivalent to not checking the corresponding box in the traditional interface 1200. Moving a slider to the right end is essentially equivalent to checking the corresponding box in the traditional interface 1200. This example illustrates how the search objective of a preference for Portland followed by a preference for Vancouver and a lack of preference for Seattle may be defined.
Implicit in the example 1300 is that all “don't care” cities will have higher search satisfaction than Seattle. When all of the search results are ranked, for example, results having a “don't care” preference will typically have higher satisfaction than results with Seattle because of the negative preference for Seattle as described in the present example.
Traditional numerical data searches tend to be for a specific value, less than a value, more than a value, or across a fixed range of values, e.g., “find all people who are 55 years old,” “find all people less than 5′7″ tall,” “find all cars that get greater than 50 mpg,” and “find all cameras that cost between $101 and $200.” These types of search objectives may be sufficient at times but often they are not. In reality, the search objective values given in search objective definitions are not hard edges. For example, if a user shopping for a camera provides the search objective as a range, e.g., $101 to $200, potentially good choices will likely be missed. For example, a camera that meets other search objectives, e.g., for size, capacity, and function, but costs $210 would probably be worth considering by the user but would not be listed as a result of the search, an example of search friction or brittleness. In other words, for measured search objectives there should be at least two values: a search objective comprise of target values and threshold values. In the case of the camera the search objective might be reworded to read “find all cameras that cost less than $230, ideally below $200.”
In difference objective belief maps, responses that show a selected amount of difference between two belief maps may be selected. For example, a user may select responses that have changed due to some event or new information. This difference allows for a comparison of a large number of responses to discover the beliefs and demographics of survey takers who have changed their responses.
As used herein, a target generally refers to an objective for quantitative information such as “more is better,” “less is better,” or “target is best.” In one example, a $200 item may be the objective for a user but an item priced up to $250 may be acceptable. In another example, a departure time of 3:00-3:30 may be the objective for a user but any time between 1:30 and 4:00 may be acceptable.
As used herein, a preference generally refers to an objective for qualitative information that includes certainty qualifications. Examples of a preference include: (1) “yes” is preferred over “no,” (2) a trip to “Seattle” may be preferred over one to “Portland” but is not as good as one to “Vancouver,” and (3) “a point and shoot camera” is preferred over an “SLR” but a “combination camera” is probably acceptable if no other type exists.
As used herein, an importance generally refers to an expression of degree of relative significance of measures where some are more highly valued than others. For example, a user may consider a zoom feature for a camera to be twice as important as megapixels and three times as important as price. In another example, a user may consider layover time to be 20 percent more important than price and price to be 60 percent more important than airline.
In traditional searches, the satisfaction for each field, Sf, are related in an “or” fashion. That is, if the search objective matches the data in a field then Sf=1, if not then Sf=0. The record satisfaction Srecord for a record is simply the product of these satisfactions:
S
record=ΠiSfi
The counter i goes from 1 to N, the number of fields searched. With this formula, Srecord is either 0 or 1 as either all the fields in the record are matched, i.e., Srecord=1.0, or at least one of them does not, i.e., Srecord=0.0. Thus, if the data in any one field does not match the search objective then the total satisfaction is 0.0.
In contrast, certain implementations of the disclosed technology use a formula to find the record satisfaction that is common in decision making but not in search. In these embodiments, the record satisfaction is the sum of the individual field satisfactions:
S
record=(ΣSfi)/N
where N refers to the number of search objectives and i=1 to N. Each “field satisfaction” is a probabilistic measure of value of the contribution of field data to the search objective and has a value between 0.0, i.e., no satisfaction, and 1.0, i.e., full satisfaction. This formulation allows the satisfaction with one measure to be traded off for satisfaction with another. Current systems' lack of being able to accommodate tradeoffs, which are implicit in virtually every search, leads to search friction.
The effect of this definition is to calculate a satisfaction for each record in the database, which can then be rank ordered to find which records are most satisfactory relative to the search objective. The range of evaluated records may be reduced by filtering and prior indexing. Field satisfactions may be developed for virtually all types of data and search objectives.
One of the many benefits of searching in satisfaction space is the ability to have a value for each field satisfaction that is a distinct value between 0.0 and 1.0 rather a discrete value of 0 or 1. This range of satisfaction values provides a user with the ability to tune a search by weighting the importance of one field match relative to another. For example, a user may indicate that “I want to find a date and she should be taller than 5′6″ with red hair” and that “Her height is more important than the color of her hair and I will even accept a brunette.” In the example, the first sentence provides clear criteria for the search and the second sentence provides that: 1) the user has a preference for red hair followed by brunette hair and 2) height is more important than hair color to the user.
The combination of preference and importance allows for detailed tailoring of searches. In equation form, an example of the addition of tuning to a search may be represented by:
S
record=(ΣTi*Sfi)/N
where the tuning or weighting factors are normalized to sum to 1, ΣTi=1.
The ability to adjust tuning allows for a user to search data through different value structures. The relative weightings defined by tuning can reflect an individual's values. In the example above, the user values height over hair color and the extent to which the user values one over the other may be reflected in the tuning values, e.g., the user could put Theight=0.7 and Tcolor=0.3. Should the user decide to tune out hair color altogether, he or she could put 1.0 and 0.0 for the weight and hair color, respectively.
Alternatively or in addition, tuning may provide a user with what-if tradeoff analysis, e.g., the effect of changing what is important allows for ready exploration of the search results to see how sensitive they are to slight changes in the relative importance of the field satisfaction results. For instance, removing the desire for hair color entirely in the present example would result in the maximum possible satisfaction associated with other search objectives. Knowing this value would make it possible to more quickly understand the boundaries of possible satisfaction associated with a set of search data.
Alternatively or in addition, tuning may provide a user with real-time results visualization. For example, continuous tuning may enable minor or major continuous changes in importance to generate equally continuous changes in search results. In this way, a user may adjust tuning as specific record satisfaction values change and use this source of informational feedback to track a course to the subset of records that best describe both the search objective and the best possible search objective given the range of possible data.
The data collected in a database may be factual demographic information, e.g., gender or height, in which case there is generally no disagreement of ambiguity. On the other hand, the data may include estimates and beliefs and there may be a range of them depending on who contributed to the database. For example, consider an example in which a user wishes to assess the potential for sales in new regions. The user may ask his or her sales team to provide data for belief that sales will be successful in these different regions. They may each provide a dot on a belief map and such beliefs may be combined.
An additional fusion method includes the utilization of characteristics of an individual group of objectives to generate individually optimized search results that may then be fused at the results level. Such a technique may be referred to as “crowd sourcing” where the search desires of a group of individuals are used to find the best possible combination of results for a larger group as a whole. In this method, each person's search objective results in a satisfaction result that may be used as data to calculate the highest satisfaction for everyone. This method may result in the generation of a crowd sourced collection of highest satisfaction results at the group level and use satisfaction results as data for overriding objectives.
These fusion techniques may be particularly useful for crowd source solutions in environments where individuals have different stimulus and individual objectives, such as on a battle field or a political campaign, for example, but share a common objective that is used to determine the format of the crowd source result. Because of its individual search objective accuracy and ability to fuse results according to commonly held or enforced relevance, such techniques tend to result in a more accurate methodology for crowd source solutions and market research analysis as compared to current techniques.
In certain embodiments, the stored information may include belief data that includes at least one representation of a statement corresponding to a user, the representation having associated therewith a belief certainty. The belief certainty may be provided by a user via a user interface or by an automated process such as automatic tagging, for example. The belief certainty may be based on at least one other representation of a statement corresponding to another user.
Alternatively or in addition thereto, the stored information may include estimation data that includes at least one characteristic having units associated therewith, the characteristic having associated therewith an estimation certainty. The estimation certainty may be provided by a user via a user interface or by an automated process such as automatic tagging, for example. The estimation certainty may be based on at least one other representation of a statement corresponding to a user. The estimation certainty may be further based on at least one other representation of a statement corresponding to another user.
At 2704, a machine executes a query against the information stored by the data store(s). The query may incorporate virtually any of the pertinent techniques described above. A detailed example of such a query is presented below with regard to
At 2706, a search request is received. The search request may be provided by a user using a user interface (UI), for example. Alternatively, the search request may be received from a third party or the request may be automatically generated by the system.
Responsive to the search request received at 2706, the machine performs a search operation as indicated at 2708. The search operation performed at 2708 may incorporate at least one result of the querying performed at 2704. Performing the search operation may include a determination as to whether any of the stored information meets a user-specified preference condition that indicates a user's preference for a first aspect of the stored information over at least a second aspect of the stored information. The user-specified preference condition may indicate a first level of preference of the user for the first aspect of the stored information and a second level of preference of the user for the second aspect of the stored information.
Alternatively or in addition thereto, performing the search operation at 2708 may include determining whether any of the stored information meets one or more user-defined importance conditions that each indicate whether a certain aspect of the stored information meets or exceeds a corresponding level of importance to the user.
Alternatively or in addition thereto, performing the search operation at 2708 may include determining whether any of the stored information meets a user-established target condition that indicates a target for a certain aspect of the stored information and a threshold range corresponding to said target. For example, the target may include a specific numerical value for a certain characteristic and the threshold range may include two additional values: one higher than the target and one lower than the target. In these embodiments, at least one search result may have a value that is within the range and may even be equal or substantially equal to the target itself.
At 2710, the machine provides at least one search result based on the search operation performed at 2708. In certain embodiments, the one or more search results are based on at least one subset of the stored information that corresponds to multiple users. The result(s) may be presented visually to a user via a graphic user interface (GUI) and a display device, for example. Alternatively or in addition thereto, the result(s) may be presented to the user by way of an audio device. The user may perform any of a number of subsequent actions with regard to the result(s) provided at 2710, such as the filtering, CDM, matching, and actionable intelligence operations described herein, for example.
At 2806, a filtering request is received. The filtering request may be provided by a user using a UI, for example. Alternatively, the filtering request may be received from a third party or the request may be automatically generated by the system.
Responsive to the filtering request received at 2806, the machine performs a filtering operation as indicated at 2808. The filtering operation performed at 2808 may incorporate at least one result of the querying performed at 2804. Performing the filtering operation may include a determination as to whether any of the stored information meets a user-specified preference condition that indicates a user's preference for a first aspect of the stored information over at least a second aspect of the stored information. The user-specified preference condition may indicate a first level of preference of the user for the first aspect of the stored information and a second level of preference of the user for the second aspect of the stored information.
Alternatively or in addition thereto, performing the filtering operation at 2808 may include determining whether any of the stored information meets one or more user-defined importance conditions that each indicate whether a certain aspect of the stored information meets or exceeds a corresponding level of importance to the user. In certain embodiments, at least one user-defined importance condition corresponds to a key word such as a topic, a subject, an email address, a website, a blog, a person, an entity, and a location, for example.
Alternatively or in addition thereto, performing the filtering operation at 2808 may include determining whether any of the stored information meets a user-established target condition that indicates a target for a certain aspect of the stored information and a threshold range corresponding to said target. For example, the target may include a specific numerical value for a certain characteristic and the threshold range may include two additional values: one higher than the target and one lower than the target. In these embodiments, at least one search result may have a value that is within the range and may even be equal or substantially equal to the target itself.
At 2810, the machine provides at least one filtering result based on the filtering operation performed at 2808. In certain embodiments, the one or more filtering results are based on at least one subset of the stored information that corresponds to multiple users. The result(s) may be presented visually to a user via a GUI and a display device, for example. Alternatively or in addition thereto, the result(s) may be presented to the user by way of an audio device. The user may perform any of a number of subsequent actions with regard to the result(s) provided at 2810, such as the search, CDM, matching, and actionable intelligence operations described herein, for example.
At 2906, a collaborative decision making (CDM) request is received. The CDM request may be provided by a user using a UI, for example. Alternatively, the CDM request may be received from a third party or the request may be automatically generated by the system. Responsive to the CDM request received at 2906, the machine performs at least one CDM operation as indicated at 2908. The CDM operation(s) performed at 2908 may incorporate at least one result of the querying performed at 2904.
At 2910, the machine provides at least one CDM result based on the CDM operation performed at 2908. The result(s) may be presented visually to a user via a GUI and a display device, for example. Alternatively or in addition thereto, the result(s) may be presented to the user by way of an audio device. The CDM result may include a hiring decision, a procurement decision, a supply chain management (SCM) decision, a customer relationship management (CRM) decision, a business intelligence (BI) decision, and a product lifecycle management (PLM) decision, or any combination thereof. In situations where there are multiple CDM results, the machine may further provide an indication that a certain one of the CDM results represents a best alternative, as indicated by the optional operation at 2912.
The user may perform any of a number of subsequent actions with regard to the result(s) provided at 2910, such as the search, filtering, matching, and actionable intelligence operations described herein, for example.
At 3006, a matching request is received. The matching request may be provided by a user using a UI, for example. Alternatively, the matching request may be received from a third party or the request may be automatically generated by the system.
Responsive to the matching request received at 3006, the machine performs a matching operation as indicated at 3008. The matching operation performed at 3008 may incorporate at least one result of the querying performed at 3004. Performing the matching operation may include a determination as to whether any of the stored information meets a user-specified preference condition that indicates a user's preference for a first aspect of the stored information over at least a second aspect of the stored information. The user-specified preference condition may indicate a first level of preference of the user for the first aspect of the stored information and a second level of preference of the user for the second aspect of the stored information.
Alternatively or in addition thereto, performing the matching operation at 3008 may include determining whether any of the stored information meets one or more user-defined importance conditions that each indicate whether a certain aspect of the stored information meets or exceeds a corresponding level of importance to the user.
Alternatively or in addition thereto, performing the matching operation at 3008 may include determining whether any of the stored information meets a user-established target condition that indicates a target for a certain aspect of the stored information and a threshold range corresponding to said target. For example, the target may include a specific numerical value for a certain characteristic and the threshold range may include two additional values: one higher than the target and one lower than the target. In these embodiments, at least one search result may have a value that is within the range and may even be equal or substantially equal to the target itself.
At 3010, the machine provides at least one matching result based on the matching operation performed at 3008. In certain embodiments, the one or more matching results are based on at least one subset of the stored information that corresponds to multiple users. The result(s) may be presented visually to a user via a GUI and a display device, for example. Alternatively or in addition thereto, the result(s) may be presented to the user by way of an audio device. The user may perform any of a number of subsequent actions with regard to the result(s) provided at 3010, such as the search, filtering, CDM, and actionable intelligence operations described herein, for example.
At 3106, a situation awareness activity is detected. Responsive to detecting the situation awareness activity at 3106, the machine performs an actionable intelligence operation as indicated at 3108. The actionable intelligence operation performed at 3108 may incorporate at least one result of the querying performed at 3104. Performing the actionable intelligence operation may include a determination as to whether any of the stored information meets a user-specified preference condition that indicates a user's preference for a first aspect of the stored information over at least a second aspect of the stored information. The user-specified preference condition may indicate a first level of preference of the user for the first aspect of the stored information and a second level of preference of the user for the second aspect of the stored information.
Alternatively or in addition thereto, performing the actionable intelligence operation at 3108 may include determining whether any of the stored information meets one or more user-defined importance conditions that each indicate whether a certain aspect of the stored information meets or exceeds a corresponding level of importance to the user.
Alternatively or in addition thereto, performing the actionable intelligence operation at 3108 may include determining whether any of the stored information meets a user-established target condition that indicates a target for a certain aspect of the stored information and a threshold range corresponding to said target. For example, the target may include a specific numerical value for a certain characteristic and the threshold range may include two additional values: one higher than the target and one lower than the target. In these embodiments, at least one search result may have a value that is within the range and may even be equal or substantially equal to the target itself.
At 3110, the machine provides at least one situation awareness activity result based on the actionable intelligence operation performed at 3108. In certain embodiments, the one or more situation awareness activity results are based on at least one subset of the stored information that corresponds to multiple users. The result(s) may be presented visually to a user via a GUI and a display device, for example. Alternatively or in addition thereto, the result(s) may be presented to the user by way of an audio device. The user may perform any of a number of subsequent actions with regard to the result(s) provided at 3110, such as the search, filtering, CDM, and matching operations described herein, for example.
At 3202, a processor applies an importance by asserting at least one user-defined importance condition against the stored information. The at least one importance condition may be provided by a user via a user interface. Each user-defined importance condition may correspond to at least one user-specified preference condition, at least one user-established target condition, both of which are described below, or both.
In addition to applying the importance at 3202, the process further performs either or both of the operations at 3204 and 3206 as described below before advancing to 3208. However, the operations at 3202 and 3204 and/or 3206 may be performed by the processor at least partially concurrently with one another or in a fully sequential manner.
At 3204, the processor applies a preference probability by asserting at least one user-specified preference condition against the stored information. Certain embodiments may include multiple preference conditions that each has a corresponding preference satisfaction value that is no less than 0.0 and no more than 1.0, for example. In situations where there are multiple preference conditions, they may be ranked according to the corresponding preference satisfaction values. The at least one user-specified preference condition may be provided by a user via the user interface.
At 3206, the processor asserts at least one user-established target condition against the stored information. Certain embodiments may include multiple target conditions that each has a corresponding target satisfaction value that is no less than 0.0 and no more than 1.0, for example. In situations where there are multiple target conditions, they may be ranked according to the corresponding target satisfaction values. The at least one user-established target condition may be provided by a user via the user interface. In certain embodiments, the target condition may include a user-provided target value for a first aspect of the stored information and a user-provided threshold range corresponding to the target value.
At 3208, one or more query results are determined. In situations where there are more than one query result at 3208, an optional ranking operation may be performed as indicated at 3210. For example, the machine may provide an indication of a ranking that corresponds to at least one of the results. The ranking may be based on one or more of the user-defined importance conditions, one or more of the user-specified preference conditions, one or more of the user-established target conditions, or any combination thereof.
It should be noted that, while the illustrated example focuses on a certain type of product, i.e., a camera, and the associated user interface, i.e., the interface 3500 of
However, the prior interface 3800 is very “black-and-white” because checking a certain checkbox essentially indicates that the user really desires the corresponding characteristic and not checking a certain checkbox essentially indicates that the user has no desire in the corresponding characteristic. There are various other shortcomings associated with an interface such as the interface 3800 of
In the example, the user may specify (or revise) any of a number of characteristics that pertain to the search. For example, the user may provide (or adjust) certain targets, such as wage and geographic location. The user may also provide (or adjust) any of a number of preferences, such as a variety of different skills associated with desired job listings.
In the example, the interface 4000 allows the user to enter or adjust the targets and preferences by way of multiple sliding mechanisms. While such mechanisms are generally easy to use and even enjoyable by users, such information may be provided by any of a number of different mechanisms. For example, a text box may be provided for any or all of the targets and/or preferences such that the user may enter a numeric value rather than use a slider.
In the example, the user may also enter belief information or estimates, or a combination thereof, the representation having associated therewith a certainty which may be provided by the user via the user interface or by an automated process such as automatic tagging, for example.
The search results may change responsive to changes with respect to other aspects of the second interface 4000. Such changes in the search results may occur in real-time or after a certain action such as a “refresh” request by the user or after a certain specified period of time has passed.
In the example, the user may specify (or revise) any of a number of characteristics that pertain to the search by providing or adjusting targets, preferences, or both. In the example, the interface 4200 allows the user to enter or adjust the targets and preference by way of multiple sliding mechanisms. As with the interface 4000 of
In the example, the user may also enter belief information or estimates, or a combination thereof, the representation having associated therewith a certainty which may be provided by the user via the user interface or by an automated process such as automatic tagging, for example.
The search results may change responsive to changes with respect to other aspects of the second interface 4200, either in real-time or after a certain action such as a “refresh” request by the user or after a certain specified period of time has passed.
There are a number of different industries and fields in addition to employment searches that may take advantage of the tools and techniques described herein, such as social networking, for example.
Consider an example in which a user of a social networking site wants to connect with other users based on any of a number of preferences. In the example, the user may specify whether he or she is merely looking for new people, looking for friends, or something in between. The user may also specify whether he or she is more interested in someone to have fun with or someone to work with. Alternatively or in addition thereto, the user may indicate how important geographic location of the other user(s) is to him or her. A search may be performed using any of a number of implementations as described herein and a listing of results may be presented to the user. Each of the search results presented to the user may have a match value associated therewith and, in certain embodiments, the results may be ranked in accordance with the match values.
In the example, the user may also include belief information or estimates, or a combination thereof, the representation having associated therewith a certainty which may be provided by the user via the user interface or by an automated process such as automatic tagging, for example.
In certain embodiments, a user may indicate that messages from his or her spouse are of greater importance than messages from other senders. The user may indicate that messages from his or her colleagues are of high importance and that messages from his or her boss' work email are of higher importance. The user may also indicate that messages received through his or her work email are of greater importance than from other incoming addresses and that messages with the user's email in the “To” field have a greater importance than messages in which his or her email address is in the “cc” or “bcc” field. The user may also indicate that messages with one or more of a specific set of keywords in the subject or body of the message are of high importance, and, therefore, when said messages are from a sender also indicated as of high importance the combined importance value of said messages is calculated as described above.
With regard to message threads, a user may apply a flag to take advantage of a number of options associated therewith. For example, a user may indicate that flagged messages or message threads are of higher importance than non-flagged messages or threads. The user may also indicate that messages or other items that are related to messages flagged by the user are of high importance.
With regard to news feeds, there are a number of items with which the user may indicate a certain level of importance. For example, the user may indicate that status feed items from friends or family members, e.g., spouse and/or kid(s), should be of higher importance than other items. The user may specify other users or events such that associated status feed items should have a higher importance. The user may indicate that certain items from certain sources, e.g., news sites, should have higher importance.
In certain embodiments, a user may indicate that items with comments from certain friends are of a certain importance and/or that items with more than one “like” are to have a certain importance. The user may further indicate that items having one or more “likes” from certain people, e.g., certain or all family and friends, should have associated therewith a particularly high importance. The user may also direct the system to assign a particular importance to items that the user (or other designated user) has tagged.
The interface 4600 provides a number of other tools to be used by the user. For example, the user may reorder the entries under any or all of the headings in accordance with his or her preferences. In the example, the user has indicated a greater preference for news stories about Iran than President Obama. The user may also add specific information such as targets, e.g., dates or other numerical characteristics, to any or all of the individual keywords or entries under any of the headings. The user may also add items to individual topics or define new topics by dragging keywords or hashtags into one or more of the listings, for example.
By using interfaces such as the interfaces 4500 and 4600 of
Regardless of the application, the techniques described above take into account certain features of the information that is being filtered by determining its certainty in terms of information belief and estimation. These characteristics of the data feed combined with the query in terms of preference, targets, and importance may provide superior filtering control.
A user may use the interface 4700 to indicate any of a number of topics and an associated importance. In the example, the user has indicated the topics “Contract,” “Acme,” and “Building” as being preferences for enterprise searching. The user may reorder the listing of topics and easily remove topics therefrom and/or add topics thereto by dragging and dropping, for example. In the example, the interface 4700 provides a slider for the user to indicated the importance, i.e., a representation as to how important the specified topics are to the user vis a vis other characteristics such as dates and costs. While the interface 4700 uses slider mechanisms, it should be noted that any of a number of other suitable user interaction mechanisms, e.g., text boxes, may be used in place of or in addition to sliders.
The user may also use the interface 4700 to enter a target date and/or date range. In the example, the user is allowed to enter a “Latest” date and a “Soonest” date to indicate the desired date range. As with the listing of topics described above, a slider mechanism is provided to allow the user to indicate a specific importance with regard to the entered date range.
The illustrated interface 4700 also allows a user to enter various types of value- and cost-related characteristics to capture target information. While the example includes multiple sliding mechanisms for a user to indicate such characteristics, any other suitable user interaction mechanisms, e.g., drag and drop, may be implemented in place of or in addition to the sliding mechanisms. The user may also indicate a measure of importance of such information to the user by way of a sliding mechanism or other suitable technique.
Users may add specific details to certain items such as dates, costs, and other measured topics for targets. As with the listing of topics discussed above, topics here may be modified in a similar manner. For example, topics may be added, deleted, and/or reordered. These operations may be performed by the user dragging keywords or hashtags from elsewhere, e.g., an email message, onto the listing. Systems implementing interfaces such as the interface 4700 of
The user interface 4700 may rely on underlying searching tools and techniques as described above to generate a list of results to be presented to the user. The results may be advantageously presented in an ordered list that best matches the user's or group's preferences, targets, and importance. This list may be updated in real-time as new information is received by the system and specific topics may be tracked. The ordered list may also be updated as preferences, targets, and importance are changed, e.g., in a control area of an interface such as the interface 4700 of
The user interface 4700 may provide the results of the querying to a machine performing a collaborative decision making (CDM) operation and generating at least one CDM result based on the CDM operation. The results can then be combined with possibly one or more interpretations of these results in terms of beliefs and estimations by users and other information gathered from users through an interface such as the interface 4700 of
Each of the listed answers may be displayed in a particular color. For example, red may indicate that there is too little information to support the corresponding answer and, therefore, the user should not necessarily believe the answer to be true and/or accurate. In contrast, an answer highlighted in green may indicate an answer having sufficient information associated therewith to support the answer to a particular extent, e.g., beyond some threshold value.
The dashboard 4800 provides the user with a number of user interaction components. For example, a “Add new issue” control may allow the user to add a new issue to the listing of issues. A “Manage Selected Issue” control, when selected or activated, may provoke the presentation of an issue management interface, such as that illustrated by
The interface 5000 also provides a “Colleagues” panel that enables the user to perform any of a number of actions involving other users or groups. The user may select another user or group by selecting the corresponding email address in a listing of email addresses, for example. The user may then initiate the sending of a request involving the other user(s). In this manner, the system may retrieve and/or generate multi-user belief data to be used in generating responses to the pertinent issue(s), for example. As with the interface 4900 of
The interface 5200 also provides a “Colleague inputs” panel that provides, when applicable, pertinent information associated with one or more of the user's colleagues, e.g., as established by using the information source interface 5000 of
In the example, the interface 5200 also provides an “Archive and Report” control that, when selected or otherwise activated, causes the system to move the issue from an active list to an archived list. The “Archive and Report” control may also allow the user (or others) to provide a rationale for the findings. Selection or activation of the “Archive and Report” control may prompt the presentation of an archive-related interface, such as that illustrated by
This application is a continuation-in-part of U.S. patent application Ser. No. 13/353,181, a continuation-in-part of U.S. patent application Ser. No. 13/353,191, a continuation-in-part of U.S. patent application Ser. No. 13/353,200, a continuation-in-part of U.S. patent application Ser. No. 13/353,207, a continuation-in-part of U.S. patent application Ser. No. 13/353,219, and a continuation-in-part of U.S. patent application Ser. No. 13/353,226, all of which were filed on Jan. 18, 2012, the contents of which are hereby incorporated by reference herein in its entirety.
| Number | Date | Country | |
|---|---|---|---|
| 61461539 | Jan 2011 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 13353181 | Jan 2012 | US |
| Child | 14013342 | US | |
| Parent | 13353191 | Jan 2012 | US |
| Child | 13353181 | US | |
| Parent | 13353200 | Jan 2012 | US |
| Child | 13353191 | US | |
| Parent | 13353207 | Jan 2012 | US |
| Child | 13353200 | US | |
| Parent | 13353219 | Jan 2012 | US |
| Child | 13353207 | US | |
| Parent | 13353226 | Jan 2012 | US |
| Child | 13353219 | US |
