Complementary Topic Model for Faster Document Information Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models struggle to efficiently extract relevant information from complex, lengthy documents due to limited training data and require significant effort to train, leading to suboptimal performance in identifying important data.
Innovation Solution
A complementary model is generated by combining a relevancy model and a knowledge graph model, which assigns relevance scores to nodes based on their connections, allowing for the extraction of the most relevant information from documents.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional machine learning models are used to extract information from documents, then the models can identify relevant information, but the training requires significant effort and time due to limited training data
Solution Approach 1:
The patent segments the information extraction task into two distinct models: a relevancy model that identifies important words and phrases, and a knowledge graph model that structures relationships between entities. This segmentation allows each model to specialize and be trained more efficiently on specific aspects of the task, reducing overall training time while maintaining extraction accuracy.
Solution Approach 2:
The patent introduces a complementary model that acts as an intermediary between the relevancy model and knowledge graph model. This complementary model integrates outputs from both models and uses heuristics to resolve conflicts, enabling the system to achieve high information extraction accuracy without requiring extensive training data for a single monolithic model.
2Ease of manufacture
If traditional machine learning models are trained with limited documents, then the training effort is reduced, but the model performance deteriorates
Solution Approach 1:
The patent merges multiple specialized models (relevancy model, knowledge graph model, and complementary model) into an integrated system. Each model is trained on limited documents for its specific function, but their combined output achieves high reliability in information extraction, overcoming the limitations of any single model trained on small datasets.
Solution Approach 2:
The complementary model functions as a composite structure that combines outputs from the relevancy model and knowledge graph model. By integrating multiple sources of information and using heuristics to weigh and reconcile their outputs, the system achieves high performance reliability even when individual components are trained on limited data.
3Loss of information
If complex, lengthy documents are analyzed to identify important information, then comprehensive information can be extracted, but the time required increases significantly
Solution Approach 1:
The relevancy model performs preliminary action by identifying and ranking important words and phrases before the full analysis occurs. This preliminary filtering reduces the complexity of the documents that need to be processed by subsequent models, enabling comprehensive information extraction from lengthy documents without proportionally increasing analysis time.
Solution Approach 2:
The patent transforms the one-dimensional text analysis into a multi-dimensional knowledge graph structure. By organizing information into entities, attributes, and relationships across multiple dimensions, the system can extract comprehensive information from complex documents more efficiently, as the knowledge graph structure enables parallel processing and faster query resolution.
Data Source
AI summary
Aspects of the present disclosure describe techniques for generating a machine learning model for extracting information from textual content. The method generally includes receiving a training data set including a plurality of documents having related textual strings. A relevancy model is generated from the training data set. The relevancy model is generally configured to generate relevance scores for a plurality of words extracted from the plurality of documents. A knowledge graph model illustrating relationships between the plurality of words extracted from the plurality of documents is generated from the training data set. The relevancy model and the knowledge graph model are aggregated into a complimentary model including a plurality of nodes from the knowledge graph model and weights associated with edges between connected nodes, wherein the weights comprise relevance scores generated from the relevancy model, and the complimentary model is deployed for use in analyzing documents.