Cognitive search query validation

• US 10,606,894 B2
• Filed: 05/12/2017
• Issued: 03/31/2020
• Est. Priority Date: 05/12/2017
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: DeLuca, Lisa Seacat, Gert van der stockt, Stefan A.
• Primary Examiner(s): To, Baoquoc N
• Application Number:
  US15/593,925
• Publication Number:
  US 20180330011A1
• Time in Patent Office: 1,054 Days
• Field of Search: 707760, 707761
• CPC Class Codes: G06F 16/9032 Query formulation
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques

Abstract

Methods, computer program products, and systems are presented. The methods include, for instance: generating domain language detection models respective to a plurality of domains. A query is mapped to a domain language detection model and intent of the query is determined by use of classification labels of the domain language detection model. Based on the classification labels that may be identified as the intent of the query, alternative queries are formed to be meaningful valid and are produced to a user.

Claims

1. A computer implemented method for cognitively validating semantics of a query, comprising:

generating, by one or more processor, a plurality of domain language detection models respectively corresponding to each domain from a plurality of domains, wherein the plurality of domain language detection models include respective hierarchies of classification labels applicable for respective content of each domain;

obtaining, by the one or more processor, a query as submitted by a user;

selecting, by the one or more processor, a domain language detection model applicable to the query from the plurality of domain language detection models;

determining, by the one or more processor, intent of the query by use of the domain language detection model from selecting;

formulating, by the one or more processor, a preconfigured number of alternative queries based on one or more classification labels corresponding to the intent of the query from the determining, wherein each of the queries is semantically valid; and

producing, by the one or more processor, the alternative queries from the formulating to the user.

2. The computer implemented method of claim 1, the generating comprising:

associating, respectively, words from content of a domain into classification labels, by use of natural language classification tools; and

organizing the classifications labels in a hierarchy suitable for the domain.

3. The computer implemented method of claim 1, the obtaining comprising:

receiving the query from a search engine, responsive to sending the query by the search engine as being unsuccessful to generate a search result corresponding to the query.

4. The computer implemented method of claim 1, the selecting comprising:

applying preconfigured domain selection rules selected from;

a category of a present page from which the query was submitted, a department of the present page, one or more user-selected tags for narrowing a search result responding to the query, a topic of the query as returned by a high level topic detector, and combinations thereof; and

determining one of the plurality of domain language detection model amongst respective domain language detection models corresponding to domains that fit best to the preconfigured domain selection rules as the domain language detection models applicable to the query.

5. The computer implemented method of claim 1, the determining comprising:

generating a plurality of tuples, wherein each tuple includes a classification label from the domain language detection model and a confidence value, wherein the classification label indicates one of intents that may be meant by the query, and wherein the confidence value indicates how probable the query meant the classification label corresponding to the confidence value.

6. The computer implemented method of claim 1, the formulating comprising:

generating a first alternative query from the preconfigured number of alternative queries by use of a first classification label corresponding to the greatest confidence value, and one or more other classification label relevant to the first classification label such that the first alternative query is semantically valid and most likely to mean what the user might have meant by the query.

7. The computer implemented method of claim 1, the producing further comprising:

providing a search result corresponding to one of the preconfigured number of alternative queries in association with the preconfigured number of alternative queries.

8. The computer implemented method of claim 1, wherein the generating a plurality of domain language detection models respectively corresponding to each domain from a plurality of domains includes generating a plurality of domain language detection models respectively corresponding to each domain from a plurality of domains, wherein different ones of the plurality of domains are differentiated by subject matter.

9. The computer implemented method of claim 1, wherein the method includes receiving feedback from the user as to the accuracy of the alternative queries, and adjusting a domain language detection model for a certain domain in dependence on the feedback.

10. The computer implemented method of claim 1, wherein the method includes receiving input from the user in response to the alternative queries, and submitting a query of the alternative queries based on the input.

11. The computer implemented method of claim 1, wherein the determining the intent of the query by use of the domain language detection model includes selecting hierarchical classification labels of the domain language detection model based on terms of the query.

12. The computer implemented method of claim 1, wherein a hierarchy of classification labels of the domain language detection model includes (a) a top level having a classification label specifying a general product category, (b) a second level below the top level having classification labels that specify product types within the general product category, and (c) a third level below the second level having classification labels, for a certain product type referenced in the second level, different brands of the certain product type.

13. The computer implemented method of claim 1, wherein a hierarchy of classification labels of the domain language detection model includes (a) a top level having a classification label specifying a general product category, (b) a second level below the top level having classification labels that specify product types within the general product category, and (c) a third level below the second level having classification labels, for a certain product type referenced in the second level, different brands of the certain product type, wherein the determining the intent of the query by use of the domain language detection model includes selecting hierarchical classification labels of the domain language detection model based on terms of the query, wherein the method includes receiving feedback from the user as to the accuracy of the alternative queries, and adjusting a domain language detection model for a certain domain in dependence on the feedback, wherein the method includes receiving input from the user in response to the alternative queries, and submitting a query of the alternative queries based on the input, wherein the formatting includes generating a first alternative query from the preconfigured number of alternative queries by use of a first classification label corresponding to the greatest confidence value, and one or more other classification label relevant to the first classification label such that the first alternative query is semantically valid and most likely to mean what the user might have meant by the query, wherein the determining comprises generating a plurality of tuples, wherein each tuple includes a classification label from the domain language detection model and a confidence value, wherein the classification label indicates one of intents that may be meant by the query, and wherein the confidence value indicates how probable the query meant the classification label corresponding to the confidence value.

14. A computer program product comprising:

a computer readable storage medium readable by one or more processor and storing instructions for execution by the one or more processor for performing a method for cognitively validating semantics of a query, comprising;

generating a plurality of domain language detection models respectively corresponding to each domain from a plurality of domains, wherein the plurality of domain language detection models include respective hierarchies of classification labels applicable for respective content of each domain;

obtaining a query as submitted by a user;

selecting a domain language detection model applicable to the query from the plurality of domain language detection models;

determining intent of the query by use of the domain language detection model from selecting;

formulating a preconfigured number of alternative queries based on one or more classification labels corresponding to the intent of the query from the determining, wherein each of the queries is semantically valid; and

producing the alternative queries from the formulating to the user.

15. The computer program product of claim 14, the generating comprising:

associating, respectively, words from content of a domain into classification labels, by use of natural language classification tools; and

organizing the classifications labels in a hierarchy suitable for the domain.

16. The computer program product of claim 14, the obtaining comprising:

receiving the query from a search engine, responsive to sending the query by the search engine as being unsuccessful to generate a search result corresponding to the query.

17. The computer program product of claim 14, the selecting comprising:

applying preconfigured domain selection rules selected from;

a category of a present page from which the query was submitted, a department of the present page, one or more user-selected tags for narrowing a search result responding to the query, a topic of the query as returned by a high level topic detector, and combinations thereof; and

determining one of the plurality of domain language detection model amongst respective domain language detection models corresponding to domains that fit best to the preconfigured domain selection rules as the domain language detection models applicable to the query.

18. The computer program product of claim 14, the determining comprising:

generating a plurality of tuples, wherein each tuple includes a classification label from the domain language detection model and a confidence value, wherein the classification label indicates one of intents that may be meant by the query, and wherein the confidence value indicates how probable the query meant the classification label corresponding to the confidence value.

19. A system comprising:

a memory;

one or more processor in communication with the memory; and

program instructions executable by the one or more processor via the memory to perform a method for cognitively validating semantics of a query, comprising;

generating a plurality of domain language detection models respectively corresponding to each domain from a plurality of domains, wherein the plurality of domain language detection models include respective hierarchies of classification labels applicable for respective content of each domain;

obtaining a query as submitted by a user;

selecting a domain language detection model applicable to the query from the plurality of domain language detection models;

determining intent of the query by use of the domain language detection model from selecting;

formulating a preconfigured number of alternative queries based on one or more classification labels corresponding to the intent of the query from the determining, wherein each of the queries is semantically valid; and

producing the alternative queries from the formulating to the user.

20. The system of claim 19, the formulating comprising:

generating a first alternative query from the preconfigured number of alternative queries by use of a first classification label corresponding to the greatest confidence value, and one or more other classification label relevant to the first classification label such that the first alternative query is semantically valid and most likely to mean what the user might have meant by the query.