Entity Attribute Relation Identification via Distributional Embeddings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for identifying entity-attribute relationships in text corpora, such as search-based applications, face limitations in accurately discerning relations beyond the context in which entities and attributes appear, leading to missed connections between similar entities sharing attributes.
Innovation Solution
The approach generates embeddings for words in sentences containing entities and attributes, uses known entity-attribute pairs to create distributional embeddings for entities and attributes, and employs a feedforward network to determine actual attribute relationships by concatenating and processing these embeddings, thereby representing entities and attributes with attributes shared by similar entities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing methods for identifying entity-attribute relationships are used, then the process is simple and fast, but the accuracy is limited and missed connections between similar entities occur
Solution Approach 1:
The patent introduces distributional embeddings as an intermediary mechanism that bridges entities and attributes through their contextual relationships. The embedding model acts as a mediator that captures semantic similarities and relationships, enabling the system to identify entity-attribute connections that are not explicitly stated but are implied through contextual patterns in the text corpus.
Solution Approach 2:
The patent transforms the entity-attribute identification problem from a direct textual analysis approach to a multi-dimensional vector space approach. By representing entities, attributes, and their relationships as vectors in a high-dimensional space, the system can capture complex semantic relationships and similarities that are not apparent in the original text structure, thereby improving identification accuracy.
2Measurement precision
If distributional embeddings are generated using known entity-attribute pairs, then accuracy improves by considering attributes of similar entities, but computational complexity increases
Solution Approach 1:
The patent performs preliminary action by pre-training the embedding model on a large corpus of text and known entity-attribute pairs before actual entity-attribute identification. This pre-training phase creates the distributional embedding space that captures general relationships between entities and attributes, which can then be efficiently applied to new queries without requiring intensive computation during the actual identification process.
Solution Approach 2:
The patent uses distributional embeddings to create vector representations (copies) of entities and attributes that preserve their semantic relationships. These vector copies can be efficiently manipulated and compared in the embedding space, reducing the computational burden compared to analyzing the full text corpus each time an entity-attribute relationship needs to be identified.
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, that facilitate identifying entity-attribute relationships in text corpora. Methods include determining whether an attribute in a candidate entity-attribute pair is an actual attribute of the entity in the entity-attribute candidate pair. This includes generating embeddings for words in the set of sentences that include the entity and the attribute and generating, using known entity-attribute pairs. This also includes generating an attribute distributional embedding for the entity based on other attributes associated with the entity from the known entity-attribute pairs, and generating an attribute distributional embedding for the attribute based on known attributes associated with known entities of the attribute in the known entity-attribute pairs. Based on these embeddings, a feedforward network determines whether the attribute in the entity-attribute candidate pair is an actual attribute of the entity in the entity-attribute candidate pair.