Knowledge Graph Surrogate Entities for Search Precision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Information retrieval systems, particularly those for domain-specific media collections, face challenges with low recall and precision due to incomplete or rich metadata annotations, leading to insufficient or irrelevant search results.
Innovation Solution
The implementation of a retrieval network that leverages external knowledge by adding query mentions as surrogate entities to a knowledge graph, generating an updated embedding space, and retrieving search results based on neighboring entities, thereby enhancing search query suggestions and result relevance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If neural networks and sophisticated embedding systems are used to generate complex high dimensional embeddings, then the capability to compare and differentiate text or image objects is improved, but the retrieval effectiveness deteriorates due to incomplete or insufficient metadata annotations
Solution Approach 1:
The patent introduces an intermediary knowledge graph that bridges the gap between query embeddings and document embeddings. The knowledge graph contains entities and relationships that serve as a mediator, allowing the system to retrieve relevant documents even when direct metadata annotations are incomplete. The knowledge graph embedding technique transforms the knowledge graph into vector representations that can be compared with query embeddings in the same embedding space.
2Measurement precision
If domain-specific media collections rely on rich metadata annotations, then search precision is improved, but recall deteriorates due to low result set size when annotations are incomplete
Solution Approach 1:
The patent adds another dimension to the search process by introducing knowledge graph entities and relationships as an additional layer of information. Instead of relying solely on traditional metadata annotations, the system projects knowledge graph data into the embedding space, creating a new dimension for comparison. This allows the system to find relevant documents through knowledge graph entity matches even when traditional metadata is insufficient.
3Speed
If traditional information retrieval systems compare embedded representations directly, then processing speed is improved, but retrieval accuracy deteriorates when metadata annotations are insufficient
Solution Approach 1:
The patent performs preliminary action by pre-computing knowledge graph embeddings and storing them in the embedding space before actual search queries are executed. When a query arrives, the system can immediately compare the query embedding with pre-computed document embeddings and knowledge graph entity embeddings without needing to process metadata annotations at query time. This preliminary preparation maintains processing speed while improving retrieval accuracy through knowledge graph enhancements.
Data Source
AI summary
The present disclosure describes systems and methods for information retrieval. Embodiments of the disclosure provide a retrieval network that leverages external knowledge to provide reformulated search query suggestions, enabling more efficient network searching and information retrieval. For example, a search query from a user (e.g., a query mention of a knowledge graph entity that is included in a search query from a user) may be added to a knowledge graph as a surrogate entity via entity linking. Embedding techniques are then invoked on the updated knowledge graph (e.g., the knowledge graph that includes additional edges between surrogate entities and other entities of the original knowledge graph), and entities neighboring the surrogate entity are retrieved based on the embedding (e.g., based on a computed distance between the surrogate entity and candidate entities in the embedding space). Search results can then be ranked and displayed based on relevance to the neighboring entity.