Graph Neural Entity Linking for Ambiguous Name Assignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing entity linking methods, such as string similarity matching techniques, struggle with ambiguity and inconsistency in entity names, often failing to account for underlying relationships, leading to inaccurate assignments of unknown entity names to known entities.
Innovation Solution
A computer-implemented method using a graph neural network (GNN) for entity linking, involving an extraction module, tokenization, graph generation, and a recommendation module to generate a tripartite graph and apply it to a GNN model for accurate assignment of unknown names to known entities, leveraging positive and negative samples for training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If string similarity matching techniques are used for entity linking, then the process can be automated, but the accuracy of entity name assignment deteriorates due to sensitivity to data quality and failure to account for underlying relationships
Solution Approach 1:
The patent replaces string similarity matching techniques with a graph neural network-based system. The mechanical/string-based approach is substituted with a neural network that processes entity names through tokenization, graph generation, and embedding transformations to achieve more accurate entity linking while maintaining automation.
Solution Approach 2:
The patent changes the fundamental parameters of entity linking from string-based similarity scores to graph-based embeddings and representations. By transforming entity names into graph structures and using neural network embeddings, the system achieves higher accuracy in capturing underlying relationships between entities.
2Device complexity
If string similarity matching techniques are used, then processing can be performed with simple algorithms, but the system fails to account for underlying relationships between entity names
Solution Approach 1:
The patent transitions from one-dimensional string similarity comparison to multi-dimensional graph-based representation. By creating tripartite graphs with multiple layers representing different aspects of entity relationships, the system captures underlying relationships that simple string matching cannot detect.
Solution Approach 2:
The patent uses composite graph structures that combine multiple types of information about entities and their relationships. The tripartite graph integrates unknown entities, known entities, and their textual representations into a unified structure that captures complex relationships beyond simple string similarity.
3Productivity
If automated entity linking is implemented to handle vast information sources, then productivity increases, but the complexity of the system increases due to the need for accurate handling of ambiguous names
Solution Approach 1:
The patent segments the entity linking process into distinct modules: extraction module for retrieving names, tokenization module for processing text, graph generation module for creating graph structures, and recommendation module for final assignment. This segmentation allows each component to be optimized independently while maintaining overall system productivity.
Solution Approach 2:
The patent performs preliminary actions by pre-processing entity names through tokenization and generating graph representations before the final linking decision. This preliminary processing of training data and graph construction enables faster and more accurate entity linking during actual operation.
Data Source
AI summary
Entity linking using a graph neural network is disclosed. Entity linking can include tokenizing an unknown name, tokenizing a known name from a set of known names, identifying a candidate from the set of known names, and generating a tripartite graph. The tripartite graph can include a first layer node corresponding to the unknown name, second layer nodes corresponding to words of the known name and the candidate, and a third layer node corresponding to the candidate. The method can further include assigning the unknown name to one of the known names by applying the tripartite graph to a graph neural network model.