Transformer-Based Knowledge Graphs for Biomedical Relation Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The rapid growth of biomedical research literature makes it impractical for researchers to manually keep up with the latest findings and access relevant information efficiently, leading to a need for automated methods to identify and synthesize data from large volumes of documents.
Innovation Solution
A knowledge graph system using a transformer-based deep neural network for relation classification models to analyze biomedical texts, identifying entities and relationships between them, enabling efficient querying and retrieval of concise answers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If researchers manually review biomedical research articles to identify relevant findings, then they can access complete information, but the time consumption and labor intensity increase significantly
Solution Approach 1:
The patent replaces the mechanical manual review process with an automated NLP-based system that uses transformer models to extract and classify relationships between biomedical entities. The system automatically processes full-text articles, identifies entities (genes, chemicals, diseases), and determines relationships without human intervention, thereby eliminating time consumption while preserving information completeness.
Solution Approach 2:
The patent introduces an intermediary NLP processing layer between the raw biomedical articles and the researcher. This intermediary system pre-processes articles by extracting key entities and relationships, generating structured representations that answer research questions without requiring researchers to read full texts, thus saving time while maintaining information accuracy.
2Loss of information
If researchers download and store full-text research articles on user devices, then they have access to complete information, but network bandwidth and local storage consumption increase significantly
Solution Approach 1:
The patent extracts only the essential information (entities and relationships) from full-text articles using NLP processing, rather than transferring or storing complete documents. The system processes articles on remote servers, extracts structured knowledge representations, and transmits only these condensed results to user devices, dramatically reducing network bandwidth and storage requirements while preserving access to relevant information.
Solution Approach 2:
The patent creates simplified copies of article content in the form of structured entity-relation representations rather than full-text copies. These abstracted representations capture the essential knowledge without requiring storage or transmission of original document formats, reducing resource consumption while maintaining informational value.
3Productivity
If a knowledge graph system processes and stores relationships from large volumes of biomedical documents, then information retrieval efficiency improves, but system complexity and computational resources increase
Solution Approach 1:
The patent segments the complex task of biomedical information processing into distinct modular components: entity recognition module, relationship extraction module, knowledge graph construction module, and query processing module. Each module handles a specific aspect of the workflow, making the overall system more manageable and maintainable while achieving high retrieval efficiency through specialized processing at each stage.
Solution Approach 2:
The patent employs universal transformer-based models that can handle multiple NLP tasks (entity recognition, relationship extraction, classification) within a single framework. This multi-functional approach reduces system complexity by eliminating the need for separate specialized systems for each task, while still achieving high productivity through the models' ability to process various biomedical information types uniformly.
Data Source
AI summary
This disclosure introduces a novel method and system for making a knowledge graph. A relation classification model is used to classify relationships between entities found in natural-language text. These entities become the nodes and the relationships between them become the edges in the knowledge graph. The entities are specific items that are relevant to the subject matter of the natural-language text. If the natural language documents are biomedical texts, the entities may be things such as chemicals, diseases, and genes. The relation classification model uses a transformer-based deep neural network architecture to understand the meaning of the text that contains the entities. The relation classification model also includes a classification layer that classifies the type of relationship between the entities found in the texts. With the knowledge graph, a user can efficiently receive answers to questions based on the aggregate knowledge found in many different documents.