System and method for supplementing a question answering system with mixed-language source documents

• US 10,303,767 B2
• Filed: 06/08/2018
• Issued: 05/28/2019
• Est. Priority Date: 10/19/2016
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Ackermann, Christopher F., Beller, Charles E., Boxwell, Stephen A., Vernier, Stanley J.
• Primary Examiner(s): Jackson, Jakieda R
• Application Number:
  US16/003,794
• Publication Number:
  US 20180293223A1
• Time in Patent Office: 354 Days
• Field of Search: 704 8, 704 9, 706 45, 706 46, 706 47
• CPC Class Codes: G06F 16/24522 Translation of natural language queries to structured queries
• CPC Class Codes: G06F 40/263 Language identification
• CPC Class Codes: G06F 40/268 Morphological analysis
• CPC Class Codes: G06F 40/295 Named entity recognition
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/006 based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
• CPC Class Codes: G06N 5/025 Extracting rules from data
• CPC Class Codes: G06N 5/04 Inference or reasoning models

Abstract

Embodiments can provide a computer implemented method, in a data processing system comprising a processor and a memory comprising instructions which are executed by the processor to cause the processor to implement a mixed-language question answering supplement system, the method comprising receiving a question in a target language; determining the question cannot be answered using a target-language only corpus; applying natural language processing to parse the question into at least one focus; for each focus, determining if one or more target language verbs share direct syntactic dependency with the focus; for each of the one or more verbs sharing direct syntactic dependency, determining if one or more target language entities share direct syntactic dependency with the verb; determining one or more Abstract Universal Verbal Types associated with each verb; for each of the one or more Abstract Universal Verbal Types, determining whether a dependency between a source language entity and a source language verb is of the same type as the dependency between the target language verb and the target language entity; if the dependency is similar, returning the source language entity as a member of a set; populating the set of returned source language entities for each focus in the target language question; identifying one or more parallel passages wherein all core arguments are matched; for each parallel passage: identifying the presence or absence of oblique nominal arguments; and measuring the precision of the oblique nominal arguments in the parallel passages against those present in the target language question; and returning an answer to the target question in the target language based on a scoring of the parallel passages based on the accuracy of their respective oblique nominal arguments.

Claims

1. A computer implemented method in a data processing system comprising a processor and a memory comprising instructions, which are executed by the processor to cause the processor to determine language independent candidate answers, the method comprising:

ingesting a corpus, wherein the corpus may comprise information in one or more languages;

using a cognitive system to convert the corpus into one or more acyclic graphs, wherein nodes in the one or more acyclic graphs represent facts and connectors in the one or more acyclic graphs represent connections between two or more facts;

generating an element data structure using the one or more acyclic graphs;

identifying clusters of elements in the element data structure based on characteristics of the elements;

storing the clusters in a cluster data structure;

in response to receiving a question,

searching the cluster data structure to generate a listing of element clusters having elements involving the question;

generating a filtered listing of candidate element clusters, wherein the filtered listing comprises elements and their clusters from the listing of element clusters that are present in the question;

training the cognitive system with association rules, wherein the association rules specify compatibility of elements with knowledge domains;

applying the trained cognitive system to analyze the filtered listing of candidate element clusters to identify the candidate element clusters that are compatible with the knowledge domain of the question;

ranking the candidate element clusters; and

outputting the candidate element cluster corresponding to the highest ranked element cluster.

2. The method as recited in claim 1, further comprising:

wherein the element data structure defines knowledge domains in the ingested corpus.

3. The method as recited in claim 1, further comprising:

wherein the element data structure includes elements from other corpora.

4. The method as recited in claim 1, further comprising:

wherein the element data structure includes elements from subject matter experts.

5. The method as recited in claim 1, further comprising:

wherein the clusters of elements in the element data structure are identified based on similar characteristics.

6. The method as recited in claim 1, further comprising:

wherein one or more of the elements in the element data structure comprise sub-elements.

7. The method as recited in claim 6, further comprising:

wherein the sub-elements are clustered.

8. The method as recited in claim 1, further comprising:

wherein the training comprises determining combinations or patterns of one or more elements found in questions asked by subject matter experts working in the same knowledge domain.

9. A computer program product for determining language independent candidate answers, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to:

ingest a corpus, wherein the corpus may comprise information in one or more languages;

use a cognitive system to convert the corpus into one or more acyclic graphs, wherein nodes in the one or more acyclic graphs represent facts and connectors in the one or more acyclic graphs represent connections between two or more facts;

generate an element data structure using the one or more acyclic graphs;

identify clusters of elements in the element data structure based on characteristics of the elements;

store the clusters in a cluster data structure;

in response to receiving a question;

search the cluster data structure to generate a listing of element clusters having elements involving the question;

generate a filtered listing of candidate element clusters, wherein the filtered listing comprises elements and their clusters from the listing of element clusters that are present in the question;

train the cognitive system with association rules, wherein the association rules specify compatibility of elements with knowledge domains;

apply the trained cognitive system to analyze the filtered listing of candidate element clusters to identify the candidate element clusters that are compatible with the knowledge domain of the question;

rank the candidate element clusters; and

output the candidate element cluster corresponding to the highest ranked element cluster.

10. The computer program product as recited in claim 9, further comprising:

wherein the element data structure defines knowledge domains in the ingested corpus.

11. The computer program product as recited in claim 9, further comprising:

wherein the element data structure includes elements from other corpora.

12. The computer program product as recited in claim 9, further comprising:

wherein the element data structure includes elements from subject matter experts.

13. The computer program product as recited in claim 9, further comprising:

wherein the clusters of elements in the element data structure are identified based on similar characteristics.

14. The computer program product as recited in claim 9, further comprising:

wherein one or more of the elements in the element data structure comprise sub-elements.

15. The computer program product as recited in claim 14, further comprising:

wherein the sub-elements are clustered.

16. The computer program product as recited in claim 9, further comprising:

wherein the training comprises determining combinations or patterns of one or more elements found in questions asked by subject matter experts working in the same knowledge domain.

17. A system for determining language independent candidate answers, comprising:

a language independent candidate answer determination processor configured to;

ingest a corpus, wherein the corpus may comprise information in one or more languages;

use a cognitive system to convert the corpus into one or more acyclic graphs, wherein nodes in the one or more acyclic graphs represent facts and connectors in the one or more acyclic graphs represent connections between two or more facts;

generate an element data structure using the one or more acyclic graphs;

identify clusters of elements in the element data structure based on characteristics of the elements;

store the clusters in a cluster data structure;

in response to receiving a question;

search the cluster data structure to generate a listing of element clusters having elements involving the question;

generate a filtered listing of candidate element clusters, wherein the filtered listing comprises elements and their clusters from the listing of element clusters that are present in the question;

train the cognitive system with association rules, wherein the association rules specify compatibility of elements with knowledge domains;

apply the trained cognitive system to analyze the filtered listing of candidate element clusters to identify the candidate element clusters that are compatible with the knowledge domain of the question;

rank the candidate element clusters; and

output the candidate element cluster corresponding to the highest ranked element cluster.

18. The system as recited in claim 17, further configured to:

train the cognitive system to learn the rules, wherein the training comprises determining combinations or patterns of one or more elements found in questions asked by subject matter experts working in the same knowledge domain.