Document Classification via Semantic Feature Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current document classification systems, particularly the bag-of-words approach, face limitations such as inability to handle multi-lingual documents, inability to reuse models between cases, poor performance with dynamic collections, and difficulty in explaining classifications, especially in eDiscovery and compliance where defensibility and clarity are crucial.
Innovation Solution
A system and method for automatic document classification that extracts structural, syntactical, and semantic information from documents using text-mining techniques, normalizes this information, and uses machine learning algorithms like TF-IDF, decision trees, and support vector machines to generate feature vectors for supervised and unsupervised clustering, enabling classification across multiple languages and dynamic collections.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If bag-of-words approach is used for document classification, then implementation is simple, but classification performance is poor and models cannot be reused between cases
Solution Approach 1:
The patent transforms the document representation from raw word counts to structured feature vectors containing semantic information (entities, concepts, relationships). This parameter transformation enables better classification performance while maintaining model reusability across different cases, directly resolving the contradiction between simple implementation and reliable classification.
2Productivity
If bag-of-words approach is used, then processing is fast, but ability to handle multi-lingual documents and dynamic collections is poor
Solution Approach 1:
The patent segments documents into structured components (entities, concepts, relationships) rather than treating them as unstructured word bags. This segmentation enables language-independent processing and adapts to dynamic collections while maintaining efficient processing through the structured feature vector representation.
3Measurement precision
If detailed word-level analysis is performed, then classification accuracy may improve, but defensibility and clarity decrease
Solution Approach 1:
The patent extracts and highlights key semantic elements (entities, concepts, relationships) from the full text, presenting only the most relevant features for classification. This extraction maintains high classification accuracy while improving defensibility by showing which specific semantic elements drove the classification decision, rather than presenting all raw words.
Data Source
AI summary
A system, method and computer program product for automatic document classification, including an extraction module configured to extract structural, syntactical and/or semantic information from a document and normalize the extracted information; a machine learning module configured to generate a model representation for automatic document classification based on feature vectors built from the normalized and extracted semantic information for supervised and/or unsupervised clustering or machine learning; and a classification module configured to select a non-classified document from a document collection, and via the extraction module extract normalized structural, syntactical and/or semantic information from the selected document, and generate via the machine learning module a model representation of the selected document based on feature vectors, and match the model representation of the selected document against the machine learning model representation to generate a document category, and/or classification for display to a user.