Generation of inferred questions for a question answering system

• US 10,460,239 B2
• Filed: 09/16/2014
• Issued: 10/29/2019
• Est. Priority Date: 09/16/2014
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Petri, John E., Pfeifer, Michael D., Rapp, William C.
• Primary Examiner(s): Afshar, Kamran
• Assistant Examiner(s): Buss, Benjamin J
• Application Number:
  US14/487,566
• Publication Number:
  US 20160078354A1
• Time in Patent Office: 1,869 Days
• CPC Class Codes: G06N 5/02 Knowledge representation; Symbolic representation

Abstract

A system and computer implemented method for generating a set of inferred questions for a question answering system is disclosed. The method may include determining, based on context data, a user state. The method may also include extracting characterization information for an object satisfying an attention criteria. The characterization information may be configured to include sensory data. The method may also include determining a relationship between the object and the user state based on the characterization information for the object and the context data of the user state. The method may also include generating, based on the relationship between the object and the user state, a set of inferred questions for a question answering system.

Claims

1. A computer implemented method for generating a set of inferred questions for a question answering system, the method comprising:

determining, by a user monitoring system and based on context data a user state, wherein determining the user state includes comparing the context data with one or more user state profiles;

identifying, by the user monitoring system, an object satisfying an attention criterion, wherein the attention criterion is defined by an attention threshold value calculated by an attention synthesis algorithm;

extracting, by the user monitoring system using a data extraction algorithm, characterization information for the object satisfying the attention criterion, wherein the characterization information is configured to include sensory data of the object, wherein the characterization information includes data collected to describe or characterize the object;

determining, by the user monitoring system and using the characterization information for the object and a subset of the context data, a relationship between the object and the user state;

generating, by the user monitoring system based on the relationship between the object and the user state, a set of inferred questions for the question answering system; and

generating, using the question answering system, one or more answers for the set of inferred questions; and

using, by the user monitoring system, feedback from the user on the one or more answers generated for the set of inferred questions in one or more subsequent question answering sessions to refine the characterization information and the context data.

2. The method of claim 1, wherein the context data is configured to include a location, a physical condition, an interest score, and a focus area, and the sensory data is configured to include visual data, textual data, and audio data.

3. The method of claim 2, wherein determining the user state includes:

collecting, by the user monitoring system using a set of sensors, environment data and biometric data for the user; and

establishing, by analyzing the environment data and biometric data, the user state.

4. The method of claim 3, further comprising:

determining, by the user monitoring system, that the user state includes a gaze having a gaze duration greater than a gaze threshold;

collecting, by a mobile device, characterization information including visual data, textual data, and audio data from a first object, wherein the first object is a target of the gaze;

comparing the characterization information of the first object with the subset of the context data;

extracting a correlation between the characterization information of the first object and the focus area of the user state; and

generating, based on the correlation between the characterization information of the first object and the focus area of the user state, a set of inferred questions for a question answering system.

5. The method of claim 4, wherein the mobile device is one or more selected from the group consisting of smart phones, tablets, smart watches, and wearable computing devices.

6. The method of claim 3, wherein the environment data includes at least one member of a group consisting of weather data, crowd density, noise level, and temperature;

and the biometric data includes at least one member of a group consisting of heart rate data, eye movement data, gait data, posture data, and vocal data.

7. The method of claim 1, wherein the attention synthesis algorithm involves at least a portion of user profile data for the user.

8. The method of claim 1, wherein determining the relationship between the object and the user state includes:

comparing the characterization information of the object with the subset of the context data; and

extracting a correlation between at least a portion of the characterization information and the subset of the context data.

9. The method of claim 1, wherein generating the set of inferred questions includes:

identifying, by a natural language processing technique configured to analyze metadata for both the sensory data and the context data, a query subject based on the characterization information of the object;

deriving, using an ontology database of structured affiliations, a set of subject modifiers for the query subject; and

formulating the set of inferred questions using the context data, the query subject, and the set of subject modifiers for the query subject.

10. The method of claim 1, further comprising identifying the first object satisfying the attention criterion using a digital image processing algorithm configured to use one or more object recognition techniques including Scale-Invariant Feature Transform (SIFT).

11. A system for generating a set of inferred questions for a question answering system, the system comprising:

a set of sensors;

one or more processors; and

a memory communicatively coupled to the one or more processors,

wherein the memory comprises instructions which, when executed by the one or more processors, cause the one or more processors to perform a method comprising the steps of;

collecting, using the set of sensors, context data including environment data and biometric data for a user;

establishing, by analyzing the context data, a user state;

identifying, an object satisfying an attention criterion, wherein the attention criterion is defined by an attention threshold value calculated by an attention synthesis algorithm;

extracting, using a data extraction algorithm, characterization information for the object satisfying the attention criterion, wherein the characterization information is configured to include sensory data of the object, wherein the characterization information includes data collected to describe or characterize the object;

determining, a relationship between the object and the user state based on the characterization information for the object and a subset of the context data;

generating, based on the relationship between the object and the user state, a set of inferred questions for the question answering system; and

generating, using the set of inferred questions and by the question answering system, a set of candidate answers from which one or more answers are selected; and

using feedback from the user on the one or more answers selected for the set of inferred questions in one or more subsequent question answering sessions to refine the characterization information and the context data.

12. The system of claim 11, wherein the context data is configured to include a location, a physical condition, an interest score, and a focus area, and the sensory data is configured to include visual data, textual data, and audio data.

13. The system of claim 11, wherein the attention synthesis algorithm involves at least a portion of user profile data for the user.

14. The system of claim 11, wherein the one or more processors are further configured such that determining the relationship between the object and the user state further comprises:

comparing the characterization information of the object with the subset of the context data; and

extracting a correlation between at least a portion of the characterization information and the subset of the context data.

15. The system of claim 11, wherein the one or more processors are further configured to perform the method further comprising:

identifying, using a natural language processing technique configured to analyze metadata for both the sensory data and the context data, a query subject based on the characterization information of the object;

deriving, using an ontology database of structured affiliations, a set of subject modifiers for the query subject; and

formulating the set of inferred questions using the context data, the query subject, and the set of subject modifiers for the query subject.

16. The system of claim 11, wherein the environment data includes weather data, crowd density, noise level, and temperature, and the biometric data includes heart rate, eye movement, gait, posture, and vocal data.

17. A computer program product comprising a computer readable storage medium having a computer readable program stored therein, wherein the computer readable program, when executed on a first computing device, causes the first computing device to:

collect, using a set of sensors, context data including environment data and biometric data for a user;

establish, by analyzing the context data, a user state;

identify, an object satisfying an attention criterion, wherein the attention criterion is defined by an attention threshold value calculated by an attention synthesis algorithm;

extract, using a data extraction algorithm, characterization information for the object satisfying the attention criterion, wherein the characterization information is configured to include sensory data of the object, wherein the characterization information includes data collected to describe or characterize the object;

determine a relationship between the object and the user state based on the characterization information for the object and a subset of the context data;

identify, using a natural language processing technique configured to analyze metadata for both the sensory data and the context data, a query subject based on the characterization information of the object;

derive, using an ontology database of structured affiliations, a set of subject modifiers for the query subject; and

formulate a set of inferred questions using the context data, the query subject and the set of subject modifiers for the query subject; and

generate one or more answers for the set of inferred questions; and

use feedback from the user on the one or more answers generated for the set of inferred questions in one or more subsequent question answering sessions to refine the characterization information and the context data.

18. The computer program product of claim 17, wherein the context data is configured to include a location, a physical condition, an interest score, and a focus area, and the sensory data is configured to include visual data, textual data, and audio data.

19. The computer program product of claim 18, wherein the computer readable program is further configured to cause the first computing device to:

determine, by the user monitoring system, that the user state includes a gaze having a gaze duration greater than a gaze threshold;

collect, by a mobile device, characterization information including visual data, textual data, and audio data from a first physical object, wherein the first physical object is a target of the gaze;

compare the characterization information of the first physical object with a subset of the context data;

extract a correlation between the characterization information of the first physical object and the focus area of the user state; and

generate, based on the correlation between the characterization information of the first physical object and the focus of the user state, a set of inferred questions for a question answering system.

20. The computer program product of claim 17, wherein the environment data includes weather data, crowd density, noise level, and temperature, and the biometric data includes heart rate, eye movement, gait, posture, and vocal data.