Granted Patent B2

US 12,282,513 · App. 17/134,274 · Granted Apr 22, 2025

Inventors: Michael Robert Glass (Bayonne, NJ); Md Faisal Mahbub Chowdhury (Woodside, NY); Alfio Massimiliano Gliozzo (Brooklyn, NY)

Assignee: International Business Machines Corporation

G06F16/9035G06F16/901G06F16/90344G06F16/9038

Abstract

Claims (48)

1. A computer-implemented method comprising:

determining, using at least one hardware processor, an initial rank and a probability of relevance of each of a retrieved plurality of electronic documents relevant to a query;

for each of a plurality of candidate facets, determining, using the at least one hardware processor, a revised rank for each of said retrieved plurality of electronic documents relevant to said query;

selecting, using the at least one hardware processor, for each of said retrieved plurality of electronic documents relevant to said query, a minimum rank from among said initial rank and said revised rank for each of said plurality of candidate facets;

determining, using the at least one hardware processor, an expected discounted cumulative gain based on said probability of relevance and said minimum rank for each of said retrieved plurality of electronic documents relevant to said query;

selecting, using the at least one hardware processor, a set of optimistic facets based on maximizing said expected discounted cumulative gain;

for each of those of said plurality of candidate facets not included in said set of optimistic facets:

swapping, using the at least one hardware processor, a given one of those of said plurality of candidate facets not included in said set of optimistic facets with each of the optimistic facets; and

determining, using the at least one hardware processor, a corresponding improvement in discounted cumulative gain;

continuing said swapping and said determining of said corresponding improvement until further improvement in said discounted cumulative gain is not observed;

updating, using the at least one hardware processor, said set of optimistic facets based on said swapping; and

with a computerized search engine implemented on the at least one hardware processor, searching for an updated set of electronic documents relevant to an updated query, the updated query being based on said query and said updated set of optimistic facets, wherein said set of optimistic facets is formulated by performing an aggregate reduction, wherein the aggregate reduction comprises reducing redundant facets that commonly appear with another facet in the retrieved plurality of electronic documents, reducing facets that commonly appear in a majority of the retrieved plurality of electronic documents and reducing facets that are non-discriminative.

2. The computer-implemented method of claim 1 , further comprising selecting said candidate facets via computerized term embedding.

3. The computer-implemented method of claim 2 , wherein selecting said candidate facets via said computerized term embedding comprises applying cosine similarity.

4. The computer-implemented method of claim 1 , wherein said probability of relevance is determined as proportional to an inverse of a sum of said initial rank and a square root of said initial rank.

5. An apparatus comprising:

a memory;

a non-transitory computer readable medium including computer executable instructions; and

at least one processor, coupled to the memory and the non-transitory computer readable medium, and operative to execute the instructions to:

determine an initial rank and a probability of relevance of each of a retrieved plurality of electronic documents relevant to a query;

for each of a plurality of candidate facets, determine a revised rank for each of said retrieved plurality of electronic documents relevant to said query;

select, for each of said retrieved plurality of electronic documents relevant to said query, a minimum rank from among said initial rank and said revised rank for each of said plurality of candidate facets;

determine an expected discounted cumulative gain based on said probability of relevance and said minimum rank for each of said retrieved plurality of electronic documents relevant to said query;

select a set of optimistic facets based on maximizing said expected discounted cumulative gain; and

for each of those of said plurality of candidate facets not included in said set of optimistic facets:

swap a given one of those of said plurality of candidate facets not included in said set of optimistic facets with each of the optimistic facets; and

determine a corresponding improvement in discounted cumulative gain;

continue said swapping and said determining of said corresponding improvement until further improvement in said discounted cumulative gain is not observed;

update said set of optimistic facets based on said swapping; and

cause a computerized search engine to search for an updated set of electronic documents relevant to an updated query, the updated query being based on said query and said updated set of optimistic facets, wherein said set of optimistic facets is formulated by performing an aggregate reduction, wherein the aggregate reduction comprises reducing redundant facets that commonly appear with another facet in the retrieved plurality of electronic documents, reducing facets that commonly appear in a majority of the retrieved plurality of electronic documents and reducing facets that are non-discriminative.

6. The apparatus of claim 5 , wherein said at least one processor is further operative to select said candidate facets via computerized term embedding.

7. The apparatus of claim 6 , wherein selecting said candidate facets via said computerized term embedding comprises applying cosine similarity.

8. The apparatus of claim 7 , wherein said probability of relevance is determined as proportional to an inverse of a sum of said initial rank and a square root of said initial rank.

9. A computer program product comprising one or more computer readable storage media having stored thereon:

first program instructions executable by a computer system to cause the computer system to determine an initial rank and a probability of relevance of each of a retrieved plurality of electronic documents relevant to a query;

second program instructions executable by the computer system to cause the computer system to, for each of a plurality of candidate facets, determine a revised rank for each of said retrieved plurality of electronic documents relevant to said query;

third program instructions executable by the computer system to cause the computer system to select, for each of said retrieved plurality of electronic documents relevant to said query, a minimum rank from among said initial rank and said revised rank for each of said plurality of candidate facets;

fourth program instructions executable by the computer system to cause the computer system to determine an expected discounted cumulative gain based on said probability of relevance and said minimum rank for each of said retrieved plurality of electronic documents relevant to said query;

fifth program instructions executable by the computer system to cause the computer system to select a set of optimistic facets based on maximizing said expected discounted cumulative gain;

sixth program instructions executable by the computer system to cause the computer system to, for each of those of said plurality of candidate facets not included in said set of optimistic facets:

swap a given one of those of said plurality of candidate facets not included in said set of optimistic facets with each of the optimistic facets; and

determine a corresponding improvement in discounted cumulative gain;

seventh program instructions executable by the computer system to cause the computer system to continue said swapping and said determining of said corresponding improvement until further improvement in said discounted cumulative gain is not observed;

eighth program instructions executable by the computer; and

ninth program instructions executable by the computer system to cause the computer system to cause a computerized search engine to search for an updated set of electronic documents relevant to an updated query, the updated query being based on said query and said updated set of optimistic facets, wherein said set of optimistic facets is formulated by performing an aggregate reduction, wherein the aggregate reduction comprises reducing redundant facets that commonly appear with another facet in the retrieved plurality of electronic documents, reducing facets that commonly appear in a majority of the retrieved plurality of electronic documents and reducing facets that are non-discriminative.

10. The computer program product of claim 9 , wherein said one or more computer readable storage media having further stored thereon:

tenth program instructions executable by the computer system to cause the computer system to select said candidate facets via computerized term embedding.

11. The computer program product of claim 10 , wherein said probability of relevance is determined as proportional to an inverse of a sum of said initial rank and a square root of said initial rank.

Assignments (1)

Continuity (1)

References Cited (106)