Entity Resolution for Search Result Consolidation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current search engines face challenges in retrieving entities based on approximate descriptions and combining features and ranks across multiple sources, leading to diluted diversity in results due to unresolved indexed pages and failure in surfacing rich content.
Innovation Solution
Integrate the advantages of vertical search and Web search by determining entity types for search queries, performing source-specific searches, filtering results, consolidating equivalent entities, and ordering them based on aggregate rank values to provide a rich search experience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If Web search is used to retrieve entities, then a large index of related terms and powerful intent analysis is available, but rich content from vertical domains is not surfaced and results are diluted by unresolved indexed pages
Solution Approach 1:
The search system is segmented into multiple specialized vertical search engines, each dedicated to a specific entity type (movies, TV shows, music, books, games, celebrities). Each vertical engine maintains its own indexed collection of entities with type-specific attributes, allowing deep coverage within each domain while avoiding the dilution problem of a single general-purpose index.
Solution Approach 2:
An entity resolution component acts as an intermediary that receives results from multiple vertical search engines, resolves equivalent entities across different sources using entity identification, and consolidates results into a unified ranking. This mediator ensures that rich content from various vertical domains is properly integrated without duplication.
2Reliability
If vertical search is used for a single entity type, then rich content is available, but queries with ambiguous descriptions or semantically relevant text not in the index fail to return results
Solution Approach 1:
The system provides universal entity search functionality across multiple entity types through a single interface. The entity resolution component can resolve entities from any vertical domain, and the faceted search interface allows users to search across all entity types while filtering by specific types, making the system adaptable to various query intents.
Solution Approach 2:
The system uses feedback from entity resolution to improve search results. When entities are resolved from vertical search results, their metadata and attributes are fed back into the ranking process, allowing the system to retrieve entities based on approximate descriptions and semantically relevant text even when exact matches are not in the index.
3Quantity of substance
If multiple sources are searched to retrieve entities, then comprehensive results are obtained, but equivalent entities from different sources dilute result diversity
Solution Approach 1:
The entity resolution component merges equivalent entities identified by different vertical search engines into single consolidated results. By using entity identification to detect equivalent entities across sources and combining their attributes and rankings, the system maintains comprehensive search coverage while eliminating duplication and preserving result diversity.
4Ease of operation
If indexed pages are not resolved with entities, then Web search can be performed, but rich content is not surfaced and numerous results link to instances of the same basic entity
Solution Approach 1:
The system performs preliminary entity resolution on indexed pages before presenting search results. By resolving entities in advance and organizing them by type, the system ensures that when users perform searches, they receive high-quality results with rich content from vertical domains, while the underlying simplicity of Web search operations is maintained.
Data Source
AI summary
Systems, methods, and computer-readable storage media for providing a rich search experience utilizing entity-type characterization are provided. A search query is received and an entity type is determined for the query. A Web search is then performed for the query against a predefined list of sources determined to be authoritative for the determined entity type. The results of the source-specific searches are filtered based upon the entity type providing a filtered list of results that each pertains to an entity of the entity type determined to be associated with the query. The filtered list are then compared to a list of resolved entities to determine equivalent entities identified by different searched sources, the equivalent entities being combined into a single potential search result. The consolidated search results are then ordered based upon rank values. At least a portion of the results may then be presented to the user.