Patent Application
20130344468
This invention generally pertains to research polling, and more specifically to obtaining structured data from freeform text entered via text boxes in a research poll.
When conducting a research poll, multiple choice questions allow respondents to answer a question given a set of possible different answers. The main strength of this type of question is that the form is easy to fill in and the answers can be checked and easily quantified. But multiple choice questions can also bias the results of a poll, since the allowable answers and the way they are worded may not be in line with how someone would naturally answer the question. For this reason, open-ended questions, where a user is free to provide any answer without being prompted by multiple choice, may yield better responses in many circumstances.
A downside of open-ended questions, however, is that they can be very difficult to quantify. One major problem lies in the designing of a numerical way for analyzing and statistically evaluating distinct responses and responses that are differently worded by are intended to mean the same thing. To process multiple choice questions, answer choices are counted and statistics used to analyze the results. But for open-ended questions, answers are sometimes manually mapped to certain numerical values to be judged quantitatively. Computer programs can be designed to pre-process the open-ended responses. However, unstructured data processing is still a challenging task and may cause significant errors. In particular, it can be difficult to disambiguate open-ended answers that should be treated as the same from those that should be treated as distinct.
Embodiments of the invention provide a system for obtaining structured data from freeform text answers in a research poll. The system includes a database of objects that may represent answers to a research poll. The system presents a research poll to a user, where the research poll includes at least one freeform text field among the answers in the poll. A user answering the poll provides a partial user input to a research poll question in the text field. In response, the system searches for objects in the database that match the user's text input, and optionally also based on the question. If one or more matching objects are found, the system presents the matching objects in a listing interface, from which the user may select an object for the answer to the poll question. In one embodiment, this process is repeated as the user provides each character of user input, thereby narrowing the matching objects via a prefix query of the database using the user input. Upon selection of an object, the system marks the selected object as the user's answer to the corresponding poll question.
In various embodiments, the matching object is presented as an auto-fill to the partial user input. Alternatively, the matching object may be presented as a list of candidate answers to complete the partial user input. In response to an unsuccessful match, the system may receive a freeform text answer from the user and update the object database with the freeform text answer. The objects in the database may include objects collected from at least one of: input from other users, user profiles, advertisements, product reviews, user comments, and social networking system pages.
In various embodiments, the system ranks the matching objects obtained from the database and orders the matching objects in a list for the user based on the ranking. The system may compute the rankings based on how well the objects fit with a category of the question. For example, if a question asks for a favorite food and the user types “bru” in the text field, the system may rank the matching object associated with the food item “Brussels sprouts” higher than the matching object associated with the city “Brussels.” Alternatively, the system may filter the matching objects based on whether they also match the category, thereby preventing users from selecting irrelevant objects for the answer. The category of the question may be provided by the creator of the research poll, or the category may be learned over time based on other users' answers to the same question.
The figures depict various embodiments of the present invention for purposes of illustration only. One skilled in the art will readily recognize from the following discussion that alternative embodiments of the structures and methods illustrated herein may be employed without departing from the principles of the invention described herein.
An online research poll system offers its customers the ability to collect opinion and feedbacks effectively and affordably than paper forms. People respond to the questions on a number of client devices with their answers, which can be instantly transferred to a poll server for processing. The polling software on the poll server can be easily maintained and updated with great flexibility. Security mechanisms can also be deployed in the polling system to ensure users' privacy.
Embodiments of the invention provide include a research poll system that allows a user to enter freeform text in a text field as an answer to one or more questions in the poll. However, sometimes even if two users give the same answer to a question, they may spell the answer differently, or write the answer in different order. To avoid this ambiguity, the system gathers similar answers that are intended to refer to the same thing and stores a structured answer in a database. This enables the system to provide selections for the users as they type at least a portion of their answer into the text field. For example, to assist users in answering a question about their favorite soda, the system may search the database and display a list of brands that match the text that the user has typed in the text field. Once the user chooses one of the candidates, the answer is complete with a unified spelling and format. At the same time, the users may still have the freedom to ignore the assistance and write their own answers that are not included on the list.
In addition to the interactions with the users, the research poll system can interact with various types of objects supported by the system including but not limited to: user profiles, advertisements, user-generated content (e.g., user posts), events (e.g., a sale that users are interested in), entity hubs (e.g., a particular entity's presence in social networks), etc. The poll system can associate a research question with matched objects from the database based on user's partial input to provide assistance to the user. For example, the poll system can provide a typeahead, i.e., displaying a matched object from the query results in grey letters, as the user types each character. The poll system can also display a list of candidate answers from objects that match the text input mined from other users' answers, user profiles and advertisements. These are just a few examples of the objects that match the text input upon which a user may act on in a research poll system, and many others are possible. An object can also include an item of user generated content. For example, a user may post on a company's fan page. The post can include a user generated comment providing the user's opinion of the company's products. In one embodiment, a research poll system provides a matching object for a sponsored object. For instance, the sponsored object from an advertisement, from a “liked” product page and/or the like.
In
The user interface 100 may also include a privacy element 110. The privacy element enables poll users to limit the use of their interaction with and/or information provided via the text field 106. For example, a user can indicate that his or her answer to the question 104 not be shared with others. On the other hand, if the user decides to share his or her favorite drink choice, the research poll environment can interface with social networks to add the information to his or her public profile, review and fan page of the specific product, and group of users sharing the same choice.
In general, the poll server 210 links the research poll system 200 via networks to one or more of the clients and users to conduct online poll, collect answers, and generate poll reports. The poll server 210 can optionally connect to one or more third party websites that launch and manage market research polls to design, generate and collect questionnaire, as well as to analyze poll results. During the polls, the poll server 210 communicates with various data stores, such as the profile 205, the ad store 215, and the object store 225, which store data structures corresponding to their respective objects maintained by the poll system 200. For example, the profile store 205 contains data structures for describing users' profiles, such as demographical information for personal users, or product and brand information for business users. Similarly, the ad store 215 maintains data related to advertisements, such as advertisers, product specifications, campaign plans, advertisement contents, and targeting users.
Before conducting the research poll, the poll server 210 can assist in selecting groups of user for the poll. For example, a market research may require a control group of users that has been exposed to promotional sales. This group of users can be identified from those following in the previous sale events from the ad store 215. By querying user profiles from the profile store 205, the poll server 210 can also identify users based on demographical data, such as gender, race, age, employment, hobby, and location, among other information. Alternatively, users can also be categorized according to their interest level in the poll product. To estimate a user's interest in a particular product, for example, the poll server 210 can retrieve data from the profile store 205 and the ad store 215 to compute a weighted sum of the user's affinities with the product including the user's review, comments, interactions with friends and “like” status regarding similar products and associated advertisements.
The input matching engine 220 searches for objects that match the user input received by the poll server 210. In one embodiment, the input matching engine 220 first determines whether a previous search for the research question has been performed. If so, the input matching engine 220 retrieves matching objects from the previous search result. Otherwise, a new matching object search is performed by the input matching engine 220. Since the user input may be partially typed answers to a research question, the input matching engine 220 can retrieves a number of objects that match the partially type input and keywords in the research question from the object store 225. The candidate objects can also be retrieved from previously received ad in the ad store 215 for similar products and brands from advertisers, advertising brokers, and/or the like. Alternatively, the input matching engine 220 can search profile store for competitors, user reviews, recommendations, fans, similar business, “like” items for objects that match the text input to the user input.
Once objects that match the text input are retrieved, the input matching engine 220 selects the candidate objects to present to the user. In one embodiment, the input matching engine 220 filters or ranks the matched objects from the object store 225. The filtering and ranking of the matching objects can be computed based on a number of criteria, for instance, the closeness a matching object fits with a category associated with the poll questions. As an example, in the user interface 100 in
In one embodiment, poll questions can be categorized manually by the party that designs, manages, or sponsors the questionnaire. For example, poll question 104 “What is your favorite brand of soda drink?” in
Alternatively, poll questions can be categorized automatically by the poll server 210 through semantic analysis and machine learning. The semantic analysis analyzes relationships among a set of poll questions and terms included in the poll questions to produce a set of categories. Objects mined from the profile store 205 and ad store 215, as well as new poll questions and user answers can be input to a supervised or unsupervised learning algorithm to augment the categories and associated questions and objects. Note that as a result of the semantic analysis and learning, a poll question may be associated with multiple categories. For example, the poll question 104 may be categorized under “soda drink” and “favorite brand.”
After selecting the candidate objects to present, the input matching engine 220 transfers the candidate objects to the poll server 210, which displays the candidate objects on the poll user interface. In one embodiment, the candidate objects can be paired with the research question. As a result, the input matching engine 220 can retrieve the candidate objects associated with the question and questions in the same category.
The data logger 230 is capable of storing user answers to the research questions so that the poll server can process the data and report poll results after the research poll is finished. The data logger can also store all the objects in the matching object search results associated with the research questions and the question categories. The data logger monitors communications at the poll server 210 regarding different interactions users may have with different types of research poll objects in the research poll system 200. The data logger 230 can maintain such data in any suitable manner. In one embodiment, each of the profile store 205, the ad store 215, and the object store 225, stores data structures to manage the data for each instance of a corresponding type of research poll object maintained by the system 200. The data structures include information fields that are suitable for the corresponding type of object. For example, the ad store 215 contains data structures that include the product descriptions, target audiences, and expiration time for an advertisement, whereas the profile store 205 contains data structures with fields suitable for describing a user's profile. When a new object of a particular type is created, the data logger 230 initializes a new data structure of the corresponding type, assigns a unique object identifier to it, and begins to add data to the object as needed. This might occur, for example, when a new matching object search is received, and input matching engine 220 collects a new group of objects that match the text input in response to a research question, ranks the candidate objects, and selects the top ranked objects.
In one embodiment, the data logger 230 further processes user answers to the research questions to discover candidate objects. If certain freeform answers occur at a number greater than a predetermined threshold, the data logger 230 adds the freeform answers to the object store 225 as new candidate objects for the corresponding research questions and the question categories. The threshold can be defined using either absolute (e.g., five occurrences) or relative (e.g., 5% of the freeform answers) number of occurrence. For example, in
In one embodiment, objects that match the text input can be searched and matched by the input matching engine 220 from the object store 225, as described above with reference to
In one embodiment, the system processes the poll data by aggregating the answers that select the same matching object. Since the matching object normalizes the answer formats, the processing of the research poll is significantly simplified. For example, potential user inputs to answer the poll question 104, such as of “Coke”, “Coca-Cola”, or “coca cola” are normalized to a standard answer “Coca Cola®” by the matching object 108A. Aggregating users who select the answer “Coca Cola®” can be implemented by an exact string comparison, which introduces no false positive or false negative. In addition, the report of the poll result can also include free text when users do not select any matching object. These freeform text answers may be processed and stored in the object store 225.
The foregoing description of the embodiments of the invention has been presented for the purpose of illustration; it is not intended to be exhaustive or to limit the invention to the precise forms disclosed. Persons skilled in the relevant art can appreciate that many modifications and variations are possible in light of the above disclosure.
Some portions of this description describe the embodiments of the invention in terms of algorithms and symbolic representations of operations on information. These algorithmic descriptions and representations are commonly used by those skilled in the data processing arts to convey the substance of their work effectively to others skilled in the art. These operations, while described functionally, computationally, or logically, are understood to be implemented by computer programs or equivalent electrical circuits, microcode, or the like. Furthermore, it has also proven convenient at times, to refer to these arrangements of operations as modules, without loss of generality. The described operations and their associated modules may be embodied in software, firmware, hardware, or any combinations thereof.
Any of the steps, operations, or processes described herein may be performed or implemented with one or more hardware or software modules, alone or in combination with other devices. In one embodiment, a software module is implemented with a computer program product comprising a computer-readable medium containing computer program code, which can be executed by a computer processor for performing any or all of the steps, operations, or processes described.
Embodiments of the invention may also relate to an apparatus for performing the operations herein. This apparatus may be specially constructed for the required purposes, and/or it may include a general-purpose computing device selectively activated or reconfigured by a computer program stored in the computer. Such a computer program may be stored in a tangible computer readable storage medium or any type of media suitable for storing electronic instructions, and coupled to a computer system bus. Furthermore, any computing systems referred to in the specification may include a single processor or may be architectures employing multiple processor designs for increased computing capability.
Embodiments of the invention may also relate to a computer data signal embodied in a carrier wave, where the computer data signal includes any embodiment of a computer program product or other data combination described herein. The computer data signal is a product that is presented in a tangible medium or carrier wave and modulated or otherwise encoded in the carrier wave, which is tangible, and transmitted according to any suitable transmission method.
Finally, the language used in the specification has been principally selected for readability and instructional purposes, and it may not have been selected to delineate or circumscribe the inventive subject matter. It is therefore intended that the scope of the invention be limited not by this detailed description, but rather by any claims that issue on an application based hereon. Accordingly, the disclosure of the embodiments of the invention is intended to be illustrative, but not limiting, of the scope of the invention, which is set forth in the following claims.
