Entity Relation Extraction Using Structured Topic Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing natural language processing technologies face challenges in extracting entity relations from documents with precision, particularly when relations span beyond a single sentence, leading to a loss of important contextual information.
Innovation Solution
A method that extracts feature vectors from unstructured text, weights them based on similarity to topic vectors derived from structured information, and uses a machine learning model to classify these vectors and determine entity relations within the document, enabling the identification of relations across the entire document.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing natural language processing approaches are used to extract entity relations, then the processing speed is maintained, but the precision and completeness of relation extraction deteriorates due to inability to recognize relations beyond single sentence boundaries
Solution Approach 1:
The patent segments the document processing into distinct modules: a feature module that extracts feature vectors from unstructured text, a weighting module that applies weights based on structured information, and a machine learning model that performs classification. This segmentation allows each module to specialize in a specific task, improving overall precision while managing complexity through modular design.
Solution Approach 2:
The patent introduces feature vectors as an intermediary representation between the raw unstructured text and the machine learning model. These feature vectors capture semantic information and are weighted using structured data before classification, enabling the system to recognize relations across sentences with high precision without overwhelming computational complexity.
2Loss of information
If existing single-sentence relation recognition approaches are used, then the computational resources are conserved, but important contextual information from other portions of the document is lost
Solution Approach 1:
The patent performs preliminary weighting of feature vectors using structured information from the document before the main classification process. This preliminary action incorporates contextual information from throughout the document into the feature representation, ensuring that no important contextual information is lost while maintaining processing efficiency through optimized computation.
Solution Approach 2:
The patent transitions from single-sentence analysis to document-level analysis by incorporating structured information as an additional dimension. The weighting module uses structured data to adjust feature vector weights, effectively adding a new dimension of contextual information that enables recognition of relations spanning multiple sentences without sacrificing processing efficiency.
3Reliability
If comprehensive document-level relation extraction is implemented, then the contextual understanding is improved, but the computational complexity and processing time increases
Solution Approach 1:
The patent changes the parameters of feature vectors by applying weights derived from structured information. This parameter transformation allows the system to prioritize relevant features and reduce noise, achieving reliable contextual understanding across the entire document while controlling processing time through efficient weighted classification in the machine learning model.
Data Source
AI summary
Methods and systems for processing records include extracting feature vectors from words in an unstructured portion of a record. The feature vectors are weighted based similarity to a topic vector from a structured portion of the record associated with the unstructured portion. The weighted feature vectors are classified using a machine learning model to determine respective probability vectors that assign a probability to each of a set of possible relations for each feature vector. Relations between entities are determined within the record based on the probability vectors. An action is performed responsive to the determined relations.