Using typestyles to prioritize and rank search results

• US 9,984,127 B2
• Filed: 01/09/2014
• Issued: 05/29/2018
• Est. Priority Date: 01/09/2014
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Bastide, Paul R, Broomhall, Matthew E., Loredo, Robert E., Lu, Fang
• Primary Examiner(s): Willis, Amanda L
• Application Number:
  US14/151,462
• Publication Number:
  US 20150193506A1
• Time in Patent Office: 1,601 Days
• CPC Class Codes: G06F 16/24573 using data annotations, e.g. user-defined metadata
• CPC Class Codes: G06F 16/901 Indexing; Data structures therefor; Storage structures for retrieval from the web G06F16/951
• CPC Class Codes: G06F 16/90332 Natural language query formulation or dialogue systems

Abstract

Computer-based search results are improved by taking in consideration emphasized content by extracting content of a data corpus items indicated by typestyle emphasis; indexing the extracted emphasized content in the searched corpus; in response to a natural language query from a requester, performing a search such as a deep question and answer search of the corpus including the indexed emphasized content; and producing search results to the requester from the corpus with preference in the order or presentation of the results according to the emphasized content.

Claims

1. An automated method for improving search results in consideration of emphasized content comprising:

prior to delivery to a user, intercepting natural language results from a search performed using a natural language query;

detecting, by one or more processors, a natural language in which the results are expressed;

retrieving according to the detected natural language, by the one or more processors, from a database, a cultural rule indicating how emphasis of words and sub-phrases is made using a shift from a default text typestyle to an emphasized text typestyle, wherein the shift occurs for one or more words in a phrase and the phrase is otherwise encoded in the default text typestyle, and wherein the emphasized text typestyle is selected from the group consisting of bolding, underlining, strikethrough, color and italicization;

finding, by the one or more processors, using the cultural rule, one or more emphasized words in the results;

assigning, the one or more processors, confidence scores to each result according to occurrences of found emphasized words relevant to the query;

re-ranking, by the one or more processors, the results according to an initial relevance and according to the confidence scores; and

producing, by the one or more processors, to the user, the re-ranked results.

2. The method as set forth in claim 1 wherein the producing of the results further comprises annotation of the results to reflect the detected emphasized one or more words.

3. The method as set forth in claim 1 further comprising:

revising the natural language query to expound on the one or more emphasized words in the results; and

performing a deep question and answer search of the corpus using the expounded natural language query.

4. The method as set forth in claim 1 further comprising:

subsequent to the presentation of results, receiving by a computer at least one user satisfaction indicator regarding the results; and

employing by a computer the satisfaction indicator in a subsequent search to improve search accuracy relative to preferred and non-preferred past results.

5. A computer program product for improving search results in consideration of emphasized content comprising:

a tangible, computer-readable storage memory device excluding a propagating signal; and

one or more program instructions embodied by the memory device for causing a processor to perform operations comprising;

prior to delivery to a user, intercepting natural language results from a search performed using a natural language query;

detecting a natural language in which the results are expressed;

retrieving, according to the detected natural language, from a database, a cultural rule for indicating emphasis of words and sub-phrases using a shift from a default text typestyle to an emphasized text typestyle for the detected natural language, wherein the shift occurs for one or more words in a phrase and the phrase is otherwise encoded in the default text typestyle, and wherein the emphasized text typestyle is selected from the group consisting of bolding, underlining, strikethrough and italicization;

finding, using the cultural rule, one or more emphasized words in the results;

assigning confidence scores to each result according to occurrences of found emphasized words relevant to the query;

re-ranking the results according to an initial relevance and according to the confidence scores; and

producing, to the user, the re-ranked results.

6. The computer program product as set forth in claim 5 wherein the producing of the results further comprises annotation of the results to reflect the detected emphasized one or more words.

7. The computer program product as set forth in claim 5 wherein the program instructions are further for causing a processor to perform operations comprising:

revising the natural language query to expound on the one or more emphasized words in the results; and

performing a deep question and answer search of the corpus using the expounded natural language query.

8. The computer program product as set forth in claim 5 wherein the program instructions are further for causing a processor to perform operations comprising:

subsequent to the presentation of results, receiving at least one user satisfaction indicator regarding the results; and

employing the satisfaction indicator in a subsequent search to improve search accuracy relative to preferred and non-preferred past results.

9. The computer program product as set forth in claim 5 wherein the computer program product is in the form of a computer system, and further comprising a computer processor which executes the program instructions embodied by the memory device.

10. A method for improving search results in consideration of emphasized content comprising:

receiving, by a machine logic based question-and-answer (QA) system, a query in a first natural language;

responsive to the receipt of the query, accessing a set of text data and associated font characteristic metadata, with;

(i) the set of text data corresponding to natural language text, and

(ii) the associated font characteristic metadata including information indicative of different fonts associated with different portions of the natural language text, wherein the associated font characteristic metadata indicates emphasized font characteristics specific to the first natural language;

responsive to the access of the set of text data and associated font characteristic metadata, performing, by the QA system, a natural language processing (NLP) operation on the set of text data and associated font characteristic metadata to obtain a plurality of query responses, wherein the query responses satisfy the query;

determining, by machine logic, a relevance ranking for each of the query responses based, at least in part, upon the emphasized font characteristics; and

producing, by a the QA system, to a data consumer, the query responses according to the relevance rankings.

11. The method of 10 wherein the query responses include a first responsive text portion having associated first font characteristic metadata, and further comprising;

receiving first origin data including information indicative of an origin of the first responsive text portion; and

determining, by machine logic, a meaning of one or more fonts indicated by the first font characteristic metadata.

12. The method of 10 wherein the producing comprises communicating to a human user in human understandable form and format.

13. The method of claim 12 further comprising:

responsive to the communicating,

receiving user input indicative of a level of the human users satisfaction with the relevance ranked responsive text portions;

responsive to the user input, adjusting use of font characteristics.