Mobile based lexicon and forecasting

• US 9,430,461 B2
• Filed: 04/11/2014
• Issued: 08/30/2016
• Est. Priority Date: 04/11/2014
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Baughman, Aaron K., Dolph, Blaine H., Khan, Kamran R., Paez, Carlos A. Jr., Sakthi, Palani
• Primary Examiner(s): JACKSON, JAKIEDA R
• Application Number:
  US14/250,748
• Publication Number:
  US 20150293899A1
• Time in Patent Office: 872 Days
• Field of Search: 704/9
• US Class Current: 1/1
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: G06F 16/334 Query execution G06F16/335 takes precedence
• CPC Class Codes: G06F 16/3344 using natural language analysis
• CPC Class Codes: G06F 16/335 Filtering based on additional data, e.g. user or group profiles filtering in web context G06F16/9535, G06F16/9536
• CPC Class Codes: G06F 40/284 Lexical analysis, e.g. tokenisation or collocates
• CPC Class Codes: G06F 40/30 Semantic analysis

Abstract

An approach is provided for ranking candidate answers to a natural language question. A natural language question is received from a user of a mobile device. Candidate answers to the received natural language question are generated. A lexicon is generated based on contextual information of the user determined by the mobile device. Contextual information of the user is forecasted based on usage of the mobile device. Based in part on the lexicon and the forecasted contextual information, the candidate answers to the natural language question are ranked.

Claims

1. A method of ranking candidate answers to a natural language question, the method comprising the steps of:

a computer receiving the natural language question from a mobile device via a computer network by which the computer and the mobile device are coupled, the natural language question being asked by a user of the mobile device;

the computer generating candidate answers to the received natural language question;

subsequent to the step of generating the candidate answers, the computer identifying first contextual information about the user, the first contextual information including (1) interests of the user based on historical usage of the mobile device by the user to run applications on the mobile device, visit websites via a browser provided by the mobile device, display textual articles on the mobile devices, enter terms into a search engine via the mobile device, and capture multimedia content, (2) a geographic location of the user provided by a global positioning system (GPS) coupled to the mobile device, (3) data specifying characteristics of an environment in proximity to the user, the characteristics of the environment being detected and measured by a first sensor coupled to the mobile device, and (4) data specifying a bodily function of the user, the bodily function being detected and measured by a second sensor coupled to the mobile device;

based on the first contextual information, the computer determining a prioritization of definitions of terms;

based on the first contextual information and the prioritization of the definitions of the terms, the computer generating a lexicon of the terms;

using a probability density function and based on (1) the historical usage of the mobile device by the user to run the applications on the mobile device, visit the websites via the browser provided by the mobile device, display the textual articles on the mobile devices, enter the terms into the search engine via the mobile device, and capture the multimedia content, (2) the geographic location of the user provided by the GPS coupled to the mobile device, (3) the data specifying characteristics of the environment in proximity to the user, and (4) the data specifying the bodily function of the user, the computer forecasting second contextual information that indicates future behavior of the user;

based on the forecasted second contextual information, the computer performing word sense disambiguation of the terms in the lexicon and adjusting the prioritization of the definitions of the terms in the lexicon;

based on the word sense disambiguation of the terms in the lexicon and the adjusted prioritization of the definitions of the terms, the computer modifying the candidate answers;

based in part on the adjusted prioritization of the definitions of the terms in the lexicon, the modified candidate answers, and the forecasted second contextual information, the computer ranking the modified candidate answers to the natural language question, the highest ranked modified candidate answer being more likely to be a correct answer to the natural language question than the other candidate answers.

2. The method of claim 1, further comprising the computer adjusting weights of a machine learning ensemble based on the forecasted second contextual information, wherein the machine learning ensemble includes (1) logistic regression phases and routes, (2) dimensions of evidence refinement, (3) probability density functions, classification techniques, and regression techniques, (4) natural language processing systems, and (5) a data retrieval system.

3. The method of claim 1, wherein the step of generating the lexicon includes generating the lexicon based on textual information displayed by the mobile device to the user.

4. The method of claim 1, further comprising providing at least one support service for at least one of creating, integrating, hosting, maintaining, and deploying computer-readable program code in the computer, the program code being executed by a processor of the computer to implement the steps of receiving, generating the candidate answers, identifying the first contextual information, determining the prioritization of the definitions of the terms, generating the lexicon, forecasting, performing the word sense disambiguation of the terms in the lexicon, modifying the candidate answers, and ranking.

5. The method of claim 1, wherein the step of forecasting the second contextual information includes utilizing a software-based forecaster that employs autoregressive integrated moving average (ARIMA) based modeling.

6. The method of claim 1, wherein the step of forecasting the second contextual information includes utilizing a software-based forecaster that employs lag-based modeling.

7. The method of claim 1, wherein the step of forecasting the second contextual information includes utilizing a software-based forecaster that employs residual calculations.

8. The method of claim 1, wherein the step of receiving the natural language question includes receiving a question about a treatment for a medical condition of the user, the question not including a country in which the user is located, wherein the step of identifying the first contextual information about the user includes identifying the current geographic location of the user provided by the GPS coupled to the mobile device, and wherein the step of ranking the modified candidate answers is based in part on the current geographic location of the user.

9. The method of claim 1, wherein the steps of determining the prioritization of the definitions of the terms based on the first contextual information, generating the lexicon of terms based on the first contextual information and the prioritization of the definitions, forecasting the second contextual information, performing word sense disambiguation, adjusting the prioritization of the definitions, modifying the candidate answers, and ranking the modified candidate answers are an enhancement to a question answering system that, without the enhancement, delivers answers to questions by receiving and parsing input text, looking for multiple candidate answers in a data set, and assigning confidence scores to respective candidate answers without taking into account a context of users who ask the questions.

10. A computer program product, comprising:

a computer-readable storage device; and

a computer-readable program code stored in the computer-readable storage device, the computer-readable program code containing instructions that are carried out by a central processing unit (CPU) of a computer system to implement a method of ranking candidate answers to a natural language question, the method comprising the steps of;

the computer system receiving the natural language question from a mobile device via a computer network by which the computer system and the mobile device are coupled, the natural language question being asked by a user of the mobile device;

the computer system generating candidate answers to the received natural language question;

subsequent to the step of generating the candidate answers, the computer system identifying first contextual information about the user, the first contextual information including (1) interests of the user based on historical usage of the mobile device by the user to run applications on the mobile device, visit websites via a browser provided by the mobile device, display textual articles on the mobile devices, enter terms into a search engine via the mobile device, and capture multimedia content, (2) a geographic location of the user provided by a global positioning system (GPS) coupled to the mobile device, (3) data specifying characteristics of an environment in proximity to the user, the characteristics of the environment being detected and measured by a first sensor coupled to the mobile device, and (4) data specifying a bodily function of the user, the bodily function being detected and measured by a second sensor coupled to the mobile device;

based on the first contextual information, the computer system determining a prioritization of definitions of terms;

based on the first contextual information and the prioritization of the definitions of the terms, the computer system generating a lexicon of the terms;

using a probability density function and based on (1) the historical usage of the mobile device by the user to run the applications on the mobile device, visit the websites via the browser provided by the mobile device, display the textual articles on the mobile devices, enter the terms into the search engine via the mobile device, and capture the multimedia content, (2) the geographic location of the user provided by the GPS coupled to the mobile device, (3) the data specifying characteristics of the environment in proximity to the user, and (4) the data specifying the bodily function of the user, the computer system forecasting second contextual information that indicates future behavior of the user;

based on the forecasted second contextual information, the computer system performing word sense disambiguation of the terms in the lexicon and adjusting the prioritization of the definitions of the terms in the lexicon;

based on the word sense disambiguation of the terms in the lexicon and the adjusted prioritization of the definitions of the terms, the computer system modifying the candidate answers;

based in part on the adjusted prioritization of the definitions of the terms in the lexicon, the modified candidate answers, and the forecasted second contextual information, the computer system ranking the modified candidate answers to the natural language question, the highest ranked modified candidate answer being more likely to be a correct answer to the natural language question than the other candidate answers.

11. The program product of claim 10, wherein the method further comprises the computer system adjusting weights of a machine learning ensemble based on the forecasted second contextual information, wherein the machine learning ensemble includes (1) logistic regression phases and routes, (2) dimensions of evidence refinement, (3) probability density functions, classification techniques, and regression techniques, (4) natural language processing systems, and (5) a data retrieval system.

12. The program product of claim 10, wherein the step of generating the lexicon includes generating the lexicon based on textual information displayed by the mobile device to the user.

13. The computer program product of claim 10, wherein the step of forecasting the second contextual information includes utilizing a software-based forecaster that employs autoregressive integrated moving average (ARIMA) based modeling.

14. The computer program product of claim 10, wherein the step of forecasting the second contextual information includes utilizing a software-based forecaster that employs lag-based modeling.

15. The computer program product of claim 10, wherein the step of forecasting the second contextual information includes utilizing a software-based forecaster that employs residual calculations.

16. The computer program product of claim 10, wherein the step of receiving the natural language question includes receiving a question about a treatment for a medical condition of the user, the question not including a country in which the user is located, wherein the step of identifying the first contextual information about the user includes identifying the current geographic location of the user provided by the GPS coupled to the mobile device, and wherein the step of ranking the modified candidate answers is based in part on the current geographic location of the user.

17. The computer program product of claim 10, wherein the steps of determining the prioritization of the definitions of the terms based on the first contextual information, generating the lexicon of terms based on the first contextual information and the prioritization of the definitions, forecasting the second contextual information, performing word sense disambiguation, adjusting the prioritization of the definitions, modifying the candidate answers, and ranking the modified candidate answers are an enhancement to a question answering system that, without the enhancement, delivers answers to questions by receiving and parsing input text, looking for multiple candidate answers in a data set, and assigning confidence scores to respective candidate answers without taking into account a context of users who ask the questions.

18. A computer system comprising:

a central processing unit (CPU);

a memory coupled to the CPU; and

a computer readable storage device coupled to the CPU, the storage device containing instructions that are executed by the CPU via the memory to implement a method of ranking candidate answers to a natural language question, the method comprising the steps of;

the computer system receiving the natural language question from a mobile device via a computer network by which the computer system and the mobile device are coupled, the natural language question being asked by a user of the mobile device;

the computer system generating candidate answers to the received natural language question;

subsequent to the step of generating the candidate answers, the computer system identifying first contextual information about the user, the first contextual information including (1) interests of the user based on historical usage of the mobile device by the user to run applications on the mobile device, visit websites via a browser provided by the mobile device, display textual articles on the mobile devices, enter terms into a search engine via the mobile device, and capture multimedia content, (2) a geographic location of the user provided by a global positioning system (GPS) coupled to the mobile device, (3) data specifying characteristics of an environment in proximity to the user, the characteristics of the environment being detected and measured by a first sensor coupled to the mobile device, and (4) data specifying a bodily function of the user, the bodily function being detected and measured by a second sensor coupled to the mobile device;

based on the first contextual information, the computer system determining a prioritization of definitions of terms;

based on the first contextual information and the prioritization of the definitions of the terms, the computer system generating a lexicon of the terms;

using a probability density function and based on (1) the historical usage of the mobile device by the user to run the applications on the mobile device, visit the websites via the browser provided by the mobile device, display the textual articles on the mobile devices, enter the terms into the search engine via the mobile device, and capture the multimedia content, (2) the geographic location of the user provided by the GPS coupled to the mobile device, (3) the data specifying characteristics of the environment in proximity to the user, and (4) the data specifying the bodily function of the user, the computer system forecasting second contextual information that indicates future behavior of the user;

based on the forecasted second contextual information, the computer system performing word sense disambiguation of the terms in the lexicon and adjusting the prioritization of the definitions of the terms in the lexicon;

based on the word sense disambiguation of the terms in the lexicon and the adjusted prioritization of the definitions of the terms, the computer system modifying the candidate answers;

based in part on the adjusted prioritization of the definitions of the terms in the lexicon, the modified candidate answers, and the forecasted second contextual information, the computer system ranking the modified candidate answers to the natural language question, the highest ranked modified candidate answer being more likely to be a correct answer to the natural language question than the other candidate answers.

19. The computer system of claim 18, wherein the method further comprises the computer system adjusting weights of a machine learning ensemble based on the forecasted second contextual information, wherein the machine learning ensemble includes (1) logistic regression phases and routes, (2) dimensions of evidence refinement, (3) probability density functions, classification techniques, and regression techniques, (4) natural language processing systems, and (5) a data retrieval system.

20. The computer system of claim 18, wherein the step of generating the lexicon includes generating the lexicon based on textual information displayed by the mobile device to the user.