Predicting Intent of a User from Anomalous Profile Data

• US 20200133966A1
• Filed: 12/19/2019
• Published: 04/30/2020
• Est. Priority Date: 01/31/2018
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Canim, Mustafa, Farrell, Robert G., Gunnels, John A., Iyengar, Arun K.
• Granted Patent: US 10,909,152 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: G10L 15/16 using artificial neural networks
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: H04L 67/306 User profiles

Abstract

Mechanisms are provided for conducting a natural language dialogue between the automatic dialogue system and a user of a client computing device. An automatic dialogue system receives natural language text corresponding to a user input from the user via the client computing device, the natural language text having an ambiguous portion of natural language text. The automatic dialogue system analyzes user profile information corresponding to the user to identify an anomaly in the user profile information and predicts a user intent associated with the anomaly. The automatic dialogue system disambiguates the ambiguous portion of the natural language text based on the predicted user intent and generates a response to the user input based on the disambiguated natural language text which is output to the client computing device to thereby conduct the natural language dialogue.

Claims

1. A method, in a data processing system comprising at least one processor and at least one memory, the at least one memory comprising instructions executed by the at least one processor to cause the at least one processor to implement an automatic dialogue system, for conducting a natural language dialogue between the automatic dialogue system and a user of a client computing device, wherein the method comprises:

receiving, by the automatic dialogue system, natural language text corresponding to a user input from the user via the client computing device, the natural language text having an ambiguous portion of natural language text;

analyzing, by the automatic dialogue system, user profile information corresponding to the user to identify at least one anomaly in the user profile information;

predicting, by the automatic dialogue system, at least one user intent associated with the at least one anomaly, wherein the user intent indicates a potential reason for the user input from the user;

disambiguating, by the automatic dialogue system, the ambiguous portion of the natural language text based on the predicted at least one user intent to generate a disambiguated natural language text corresponding to the user input;

generating, by the automatic dialogue system, a response to the user input based on the disambiguated natural language text; and

outputting, by the automatic dialogue system, the response to the client computing device to thereby conduct the natural language dialogue.

2. The method of claim 1, wherein disambiguating the ambiguous portion of the natural language text comprises at least one of word disambiguation, reference disambiguation, topic disambiguation, or parse disambiguation, based on the prediction of the at least one user intent.

3. The method of claim 1, wherein analyzing user profile information corresponding to the user to identify at least one anomaly in the user profile information comprises identifying statistics, patterns, or trends in the user profile information over a predetermined period of time indicating an anomalous change in at least one variable of the user profile information, and determining associated factors indicating reasons for the anomalous change in the at least one variable.

4. The method of claim 3, wherein the at least one anomaly is identified by applying predefined rules to the statistics, patterns, or trends in the user profile information over the predetermined period of time, and wherein each of the predefined rules specify a corresponding threshold amount of change indicative of an anomaly for a corresponding variable in the at least one variable of the user profile information.

5. The method of claim 4, wherein the threshold amount of change specified in a predefined rule is automatically learned through a machine learning process based on other user responses to dialogue generated by the automatic dialogue system for the corresponding variable in other dialogue sessions.

7. The method of claim 1, wherein analyzing the user profile information further comprises mapping an intensity of the at least one anomaly to an intensity classification at least by generating a probability value corresponding to the intensity classification, and wherein predicting the at least one user intent associated with the at least one anomaly comprises selecting a user intent having a highest probability value for use in disambiguating the ambiguous portion of the natural language text.

8. The method of claim 1, wherein disambiguating the ambiguous portion of the natural language text based on the predicted at least one user intent comprises:

generating, for each at least one user intent, a corresponding disambiguated version of the natural language text in which the ambiguous portion is disambiguated, wherein for different user intents in the at least one user intent, different disambiguated versions are generated with different disambiguations of the ambiguous portion;

weighting a confidence score associated with each of the disambiguated versions of the natural language text based on a weight value associated with a corresponding at least one user intent; and

selecting a disambiguated version of the natural language text based on the weighted confidence scores of each of the disambiguated versions.

9. The method of claim 8, wherein weighting the confidence score further comprises weighting the confidence score based on an intensity of a change corresponding to the at least one intent and a confidence of a parse of the natural language text.

10. The method of claim 1, wherein analyzing user profile information corresponding to the user to identify at least one anomaly in the user profile information and predicting at least one user intent associated with the at least one anomaly are performed prior to receiving the natural language text and the identification of the at least one anomaly and the corresponding predicted at least one user intent are stored in association with the user profile information.

6. (canceled)

11. 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 computing device, causes the computing device to implement an automatic dialogue system for conducting a natural language dialogue between the automatic dialogue system and a user of a client computing device, wherein the automatic dialogue system operates to:

receive natural language text corresponding to a user input from the user via the client computing device, the natural language text having an ambiguous portion of natural language text;

analyze user profile information corresponding to the user to identify at least one anomaly in the user profile information;

predict at least one user intent associated with the at least one anomaly, wherein the user intent indicates a potential reason for the user input from the user;

disambiguate the ambiguous portion of the natural language text based on the predicted at least one user intent to generate a disambiguated natural language text corresponding to the user input;

generate a response to the user input based on the disambiguated natural language text; and

output the response to the client computing device to thereby conduct the natural language dialogue.

12. The computer program product of claim 11, wherein disambiguating the ambiguous portion of the natural language text comprises at least one of word disambiguation, reference disambiguation, topic disambiguation, or parse disambiguation, based on the prediction of the at least one user intent.

13. The computer program product of claim 11, wherein analyzing user profile information corresponding to the user to identify at least one anomaly in the user profile information comprises identifying statistics, patterns, or trends in the user profile information over a predetermined period of time indicating an anomalous change in at least one variable of the user profile information, and determining associated factors indicating reasons for the anomalous change in the at least one variable.

14. The computer program product of claim 13, wherein the at least one anomaly is identified by applying predefined rules to the statistics, patterns, or trends in the user profile information over the predetermined period of time, and wherein each of the predefined rules specify a corresponding threshold amount of change indicative of an anomaly for a corresponding variable in the at least one variable of the user profile information.

15. The computer program product of claim 14, wherein the threshold amount of change specified in a predefined rule is automatically learned through a machine learning process based on other user responses to dialogue generated by the automatic dialogue system for the corresponding variable in other dialogue sessions.

17. The computer program product of claim 11, wherein analyzing the user profile information further comprises mapping an intensity of the at least one anomaly to an intensity classification at least by generating a probability value corresponding to the intensity classification, and wherein predicting the at least one user intent associated with the at least one anomaly comprises selecting a user intent having a highest probability value for use in disambiguating the ambiguous portion of the natural language text.

18. The computer program product of claim 11, wherein disambiguating the ambiguous portion of the natural language text based on the predicted at least one user intent comprises:

generating, for each at least one user intent, a corresponding disambiguated version of the natural language text in which the ambiguous portion is disambiguated, wherein for different user intents in the at least one user intent, different disambiguated versions are generated with different disambiguations of the ambiguous portion;

weighting a confidence score associated with each of the disambiguated versions of the natural language text based on a weight value associated with a corresponding at least one user intent; and

selecting a disambiguated version of the natural language text based on the weighted confidence scores of each of the disambiguated versions.

19. The computer program product of claim 18, wherein weighting the confidence score further comprises weighting the confidence score based on an intensity of a change corresponding to the at least one intent and a confidence of a parse of the natural language text.

16. (canceled)

20. An apparatus comprising:

a processor; and

a memory coupled to the processor, wherein the memory comprises instructions which, when executed by the processor, cause the processor to implement an automatic dialogue system for conducting a natural language dialogue between the automatic dialogue system and a user of a client computing device, wherein the automatic dialogue system operates to;

receive natural language text corresponding to a user input from the user via the client computing device, the natural language text having an ambiguous portion of natural language text;

analyze user profile information corresponding to the user to identify at least one anomaly in the user profile information;

predict at least one user intent associated with the at least one anomaly, wherein the user intent indicates a potential reason for the user input from the user;

disambiguate the ambiguous portion of the natural language text based on the predicted at least one user intent to generate a disambiguated natural language text corresponding to the user input;

generate a response to the user input based on the disambiguated natural language text; and

output the response to the client computing device to thereby conduct the natural language dialogue.