Medical Knowledge Graph Generation via Open Extraction and Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing knowledge graph construction methods are inefficient due to the need for predefined graph schema, which can lead to excessive prompt information and limitations in extracting comprehensive entity and relation information.
Innovation Solution
A two-stage process is employed, first extracting entity words and types in an open manner using a large model, followed by alignment with predefined schemas to generate a knowledge graph, allowing for more comprehensive mining and reducing the reliance on lengthy prompt information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a predefined graph schema is provided in prompt information for knowledge graph construction, then the extraction process can be guided with structured entity and relation types, but the prompt information becomes excessively long and extraction efficiency deteriorates
Solution Approach 1:
The patent divides the knowledge graph construction process into two separate stages: first performing entity extraction without predefined schema constraints to obtain candidate entities, then performing relation extraction based on these candidates. This segmentation allows each stage to operate independently with optimized prompt lengths, resolving the contradiction between comprehensive entity coverage and extraction efficiency.
Solution Approach 2:
The patent extracts and removes the predefined graph schema from the prompt information used during entity extraction. By taking out the schema constraint from the first stage, the prompt length is reduced significantly while still achieving comprehensive entity extraction. The schema is then applied in the second relation extraction stage where it is more efficiently utilized.
2Ease of manufacture
If entity words and relation information are extracted through a large model with predefined schema, then the extraction process follows a structured approach, but the extracted information is limited by the predefined schema and comprehensiveness deteriorates
Solution Approach 1:
The patent inverts the conventional approach by first extracting entities without predefined schema constraints (open extraction), then subsequently applying the schema for relation extraction. This inversion allows the system to capture all possible entities first, preventing information loss, and then structure the relations in a second pass, thereby maintaining both comprehensiveness and structural organization.
3Reliability
If a large amount of graph schema content is included in prompt information, then the extraction process has comprehensive guidance, but the processing time increases and efficiency deteriorates
Solution Approach 1:
The patent segments the schema application into two distinct phases: entity extraction phase without schema (reducing prompt size and processing time) and relation extraction phase with schema (ensuring reliability through structured guidance). This temporal segmentation of schema application maintains extraction reliability while significantly reducing overall processing time.
Data Source
AI summary
Embodiments of this specification provide a method and an apparatus for generating a medical knowledge graph based on a text corpus. When a knowledge graph is constructed based on a text corpus, a data obtaining process of the knowledge graph can be divided into two stages: open extraction and alignment. Specifically, entity words and corresponding entity types are first extracted from a raw text corpus in an open manner through a large model, and a corresponding connection relation is further extracted based on the extracted entity words and entity types. Then, entity and relation alignment is performed based on a predefined entity schema and connection schema, and the knowledge graph is constructed based on an alignment result.