Process for identifying completion of domain adaptation dictionary activities

• US 10,303,763 B2
• Filed: 01/06/2017
• Issued: 05/28/2019
• Est. Priority Date: 01/06/2017
• Status: Active Grant

• Current Assignee: Callaway Cars, Inc., International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Beller, Charles E., Chase, Paul J., Darden, Richard, Drzewucki, Michael, Katz, Edward G., Phipps, Christopher
• Primary Examiner(s): Vo, Huyen X
• Application Number:
  US15/400,300
• Publication Number:
  US 20180196795A1
• Time in Patent Office: 872 Days
• Field of Search: 704 1- 10, 704230-257, 704270-277
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: G06F 16/3344 using natural language analysis
• CPC Class Codes: G06F 16/93 Document management systems
• CPC Class Codes: G06F 40/242 Dictionaries
• CPC Class Codes: G06F 40/284 Lexical analysis, e.g. tokenisation or collocates
• CPC Class Codes: G06F 40/30 Semantic analysis

Abstract

An apparatus comprising a memory and a processor configured for semi-autonomous natural language processing domain adaptation related activities. The processor coupled to the memory and configured to identify a corpus of documents of an evaluation domain and generate a first lexicon based on the corpus of documents of the evaluation domain, and determine a threshold that indicates a sufficiency of domain adaptation of the evaluation domain based at least in part on the first lexicon. The processor is further configured to identify a corpus of documents of a client domain, generate a second lexicon based on the corpus of documents of the client domain, determine a metric associated with the corpus of documents of the client domain and the second lexicon, and determine that domain adaptation of the client domain is complete when the metric exceeds the threshold.

Claims

1. An apparatus comprising:

a memory; and

a processor coupled to the memory and configured to;

identify a corpus of documents of an evaluation domain;

generate a first lexicon based on the corpus of documents of the evaluation domain;

determine a threshold that indicates a sufficiency of domain adaptation of the evaluation domain based at least in part on the first lexicon;

identify a corpus of documents of a client domain;

generate a second lexicon based on the corpus of documents of the client domain;

determine a metric associated with the corpus of documents of the client domain and the second lexicon by determining a ratio of newly extracted and unique domain terms extracted from the corpus of documents of the client domain for inclusion in the second lexicon to a total number of domain terms extracted from the corpus of documents of the client domain;

determine that domain adaptation of the client domain is complete when the metric exceeds the threshold;

receive a first question for processing according to natural language processing; and

perform first natural language processing to determine a first answer to the first question based at least in part on the second lexicon.

2. The apparatus of claim 1, wherein the processor is further configured to extract a first plurality of domain terms from the corpus of documents of the evaluation domain to generate the first lexicon.

3. The apparatus of claim 2, wherein the processor is further configured to determine the threshold that indicates the sufficiency of domain adaptation of the evaluation domain according to a comparison of a quantity of domain terms in a third lexicon to a quantity of the domain terms in the first lexicon.

4. The apparatus of claim 3, wherein the processor is further configured to determine the threshold that indicates the sufficiency of domain adaptation according to

5. The apparatus of claim 1, wherein the processor is further configured to determine the metric associated with the corpus of documents of the client domain and the second lexicon according to

6. The apparatus of claim 1, wherein the processor is further configured to:

divide the corpus of documents of the client domain into a plurality of ordered sub-corpora; and

generate a plurality of lexicon each corresponding uniquely to one of the plurality of ordered sub-corpora.

7. The apparatus of claim 1, wherein before identifying the corpus of documents of the evaluation domain, the processor is further configured to:

receive a second question for processing according to natural language processing; and

perform second natural language processing to determine a second answer to the second question,

wherein the first answer has a greater degree of accuracy with respect to the first question than a degree of accuracy of the second answer with respect to the second question.

8. A computer program product for performing domain adaptation of a domain, the computer program product comprising a non-transitory computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to:

identify a corpus of documents from within a client domain;

divide the corpus of documents into a plurality of sub-corpora;

extract at least one domain term from each of the plurality of sub-corpora, wherein domain terms extracted from one of the plurality of sub-corpora form a lexicon for that respective sub-corpora of the plurality of sub-corpora;

determine a metric having a relationship to the lexicon for that respective sub-corpora of the plurality of sub-corpora by determining a ratio of newly extracted and unique domain terms extracted from that respective sub-corpora of the plurality of sub-corpora for inclusion in the lexicon for that respective sub-corpora of the plurality of sub-corpora to a total number of domain terms extracted from that respective sub-corpora of the plurality of sub-corpora;

determine, based at least in part on the metric, that sufficient domain adaptation of the client domain has been performed;

receive a question for processing according to natural language processing; and

perform the natural language processing to determine a first answer to the first question based at least in part on the lexicon.

9. The computer program product of claim 8, wherein the metric is determined according to

10. The computer program product of claim 9, wherein sufficient domain adaptation of the client domain has been performed when the metric is less than a predetermined threshold indicative of sufficiency of domain adaptation.

11. The computer program product of claim 10, wherein the predetermined threshold is determined at least in part according to a test system evaluating an evaluation domain against at least one evaluation question to determine the threshold indicative of sufficiency of domain adaptation.

12. The computer program product of claim 11, wherein the corpus of documents is expanded when the metric is greater than the predetermined threshold indicative of sufficiency of domain adaptation.

13. The computer program product of claim 8 implemented as a service in a cloud computing environment.

14. A computer-implemented method, comprising:

identifying a corpus of documents of an evaluation domain;

generating a first lexicon based on the corpus of documents of the evaluation domain;

determining a threshold that indicates a sufficiency of domain adaptation of the evaluation domain based at least in part on the first lexicon;

identifying a corpus of documents of a client domain;

generating a second lexicon based on the corpus of documents of the client domain;

determining a metric associated with the corpus of documents of the client domain and the second lexicon by determining a ratio of newly extracted and unique domain terms extracted from the corpus of documents of the client domain for inclusion in the second lexicon to a total number of domain terms extracted from the corpus of documents of the client domain;

determining that domain adaptation of the client domain is complete when the metric exceeds the threshold;

receiving a first question for processing according to natural language processing; and

performing first natural language processing to determine a first answer to the first question based at least in part on the second lexicon.

15. The computer-implemented method of claim 14, further comprising extracting a first plurality of domain terms from the corpus of documents of the evaluation domain to generate the first lexicon.

16. The computer-implemented method of claim 14, further comprising determining the threshold that indicates the sufficiency of domain adaptation of the evaluation domain according to a comparison of a quantity of domain terms in a third lexicon to a quantity of the domain terms in the first lexicon.

17. The computer-implemented method of claim 16, further comprising determining the threshold that indicates the sufficiency of domain adaptation according to

18. The computer-implemented method of claim 14, further comprising determining the metric associated with the corpus of documents of the client domain and the second lexicon according to

19. The computer-implemented method of claim 14, further comprising:

dividing the corpus of documents of the client domain into a plurality of ordered sub-corpora; and

generating a plurality of lexicon each corresponding uniquely to one of the plurality of ordered sub-corpora.

20. The computer-implemented method of claim 14, wherein before identifying the corpus of documents of the evaluation domain, the method further comprises:

receiving a second question for processing according to natural language processing; and

performing second natural language processing to determine a second answer to the second question, wherein the first answer has a greater degree of accuracy with respect to the first question than a degree of accuracy of the second answer with respect to the second question.