Semantic Encoding Neural Network for Unsupervised Knowledge Graph Construction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current natural language processing (NLP) tasks for unsupervised learning of domain-specific knowledge graphs from textual data and language generation from knowledge graphs require labeled data, hand-crafted rules, or domain ontology, which are costly and prone to errors.
Innovation Solution
The method employs a semantic encoding and language neural network that uses reinforcement learning to automatically parse unstructured data into knowledge graphs and generate text from these graphs, leveraging two machine learning models: a semantic encoder and a semantic decoder.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If labeled data, hand-crafted rules, or domain ontology are used for NLP tasks, then learning accuracy and language generation quality are improved, but development cost and time consumption increase significantly
Solution Approach 1:
The system performs self-service by automatically learning domain-specific knowledge from unstructured text data without requiring external labeled data or manual rule creation. The neural network autonomously parses text, extracts entities and relations, constructs knowledge graphs, and generates language rules, eliminating the need for human annotators and domain experts to manually prepare training materials.
Solution Approach 2:
The patent replaces the mechanical process of manual knowledge graph construction and language rule creation with an automated neural network system. The semantic encoder, knowledge graph constructor, and semantic decoder work together to automatically transform unstructured text into structured knowledge representations and generate language generation rules, substituting human manual work with intelligent automation.
2Measurement precision
If labeled data, hand-crafted rules, or domain ontology are used for NLP tasks, then learning accuracy and language generation quality are improved, but development cost increases
Solution Approach 1:
The system performs self-service by automatically learning domain-specific knowledge from unstructured text data without requiring external labeled data or manual rule creation. The neural network autonomously parses text, extracts entities and relations, constructs knowledge graphs, and generates language rules, eliminating the need for human annotators and domain experts to manually prepare training materials.
Solution Approach 2:
The system creates a copy of domain knowledge by automatically extracting and representing it in the form of knowledge graphs and language generation rules from unstructured text. This copied knowledge structure can be reused for multiple NLP tasks without requiring additional manual effort or cost.
3Measurement precision
If manual knowledge graph construction and language rule creation are used, then domain-specific accuracy is improved, but device complexity and operational difficulty increase
Solution Approach 1:
The patent merges multiple functions into a unified neural network system: the semantic encoder for text parsing, the knowledge graph constructor for structure building, and the semantic decoder for language generation. These components work together as an integrated system that automatically achieves domain-specific accuracy without requiring separate manual processes for each task.
Solution Approach 2:
The neural network system performs multiple functions simultaneously: it parses unstructured text, extracts entities and relations, constructs knowledge graphs, and generates language generation rules. This multi-functional system replaces multiple specialized manual processes, reducing operational complexity while maintaining domain-specific accuracy across different NLP tasks.
Data Source
AI summary
Embodiments are provided for unsupervised learning of domain specific knowledge graph from textual data and language generation from knowledge graph via reinforcement learning in a computing system by a processor. Unstructured data is automatically parsed into one or more knowledge graphs based on the unstructured data and a list of candidate relations using a first machine learning model. Text data is generated from the one or more knowledge graphs using a second machine learning model.