Domain-Specific Knowledge Graph Generation from Unstructured Text
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data delivery systems struggle to efficiently generate domain-specific knowledge graphs from unstructured computer text, leading to inaccurate and time-consuming personalized content recommendations due to the limitations of general knowledge graphs and the high cost and maintenance requirements of existing systems like Google™ Knowledge Graph and Wikidata.
Innovation Solution
A system and method that leverages domain-independent knowledge graphs and unstructured computer text to train AI-based classification models, identifying entity relationships within domain-specific documents, enabling rapid creation of a multidimensional domain-specific knowledge graph.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If general knowledge graphs like Google Knowledge Graph or Wikidata are used, then broad information coverage is achieved, but domain-specific accuracy and personalization capability deteriorate
Solution Approach 1:
The patent segments the knowledge graph into domain-independent and domain-specific components. The domain-independent knowledge graph provides broad information coverage, while domain-specific knowledge graphs are constructed for particular fields (e.g., finance, healthcare) to ensure domain-specific accuracy. This segmentation allows the system to leverage both general and specialized information appropriately.
Solution Approach 2:
The patent applies local quality by tailoring the knowledge graph structure and content to specific domains. Different domains receive customized knowledge graphs with domain-relevant entities, relationships, and reasoning rules, rather than applying a uniform general knowledge graph structure to all domains. This enables accurate domain-specific inference while maintaining broad coverage through the domain-independent component.
2Measurement precision
If specialized computing techniques like machine learning classification modeling and neural networks are used to build knowledge graphs, then domain-specific accuracy is improved, but time consumption and computational cost worsen
Solution Approach 1:
The patent performs preliminary action by pre-processing and structuring domain-specific documents before building the knowledge graph. Documents are parsed, entities are extracted and standardized, and relationships are pre-identified using domain-specific schemas. This preliminary structuring reduces the computational burden during query processing and enables faster knowledge graph construction compared to building from raw unstructured text each time.
Solution Approach 2:
The patent uses copying by leveraging the structure and methodology from the domain-independent knowledge graph construction process and applying it to domain-specific contexts. Rather than developing entirely new complex machine learning models for each domain, the system adapts and copies the proven framework, reducing development time and computational overhead while maintaining domain-specific accuracy through customized entity types and relationships.
3Quantity of substance
If knowledge graphs are grown in scope and coverage, then information completeness is improved, but effort in building and maintaining the graph worsens
Solution Approach 1:
The patent introduces a hierarchical dimension to the knowledge graph system, organizing knowledge graphs across multiple levels: domain-independent (broad coverage) and domain-specific (specialized). This dimensional organization allows the system to scale information completeness by adding new domain-specific graphs without proportionally increasing maintenance complexity, as each domain-specific graph can be independently managed and updated.
Solution Approach 2:
The patent applies universality by designing a unified framework that handles both domain-independent and domain-specific knowledge graphs using the same basic architecture and processing mechanisms. The system uses universal entity extraction, relationship identification, and query processing methods that work across all domains, reducing maintenance effort compared to having separate specialized systems for each domain.
4Measurement precision
If domain-specific knowledge graphs are created from unstructured text, then personalized content recommendation accuracy is improved, but processing speed deteriorates
Solution Approach 1:
The patent performs preliminary action by pre-processing domain-specific documents to extract entities, relationships, and semantic structures before the knowledge graph is queried. Documents are parsed, keyed, and structured according to domain-specific schemas in advance, creating a ready-to-query knowledge graph that enables fast retrieval and accurate personalized recommendations without performing heavy processing at query time.
Data Source
AI summary
Methods and apparatuses are described for generating a domain-specific knowledge graph from unstructured computer text. A computing device extracts unstructured computer text associated with pairs of entities from domain-independent documents, and trains an entity relationship classification model using a domain-independent knowledge graph and the extracted text. The computing device extracts other unstructured text associated with a first domain from domain-specific documents. The computing device identifies pairs of entities contained within the text for the first domain, and executes the trained model to determine a relationship between the entities in each pair of entities identified from the text for the first domain. The computing device generates a domain-specific knowledge graph using (i) the pairs of entities identified from the text for the first domain and (ii) the relationships between the entities in each pair of entities identified from the text for the first domain.