Multi-Modality Email Classification Using Social Network Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Document classification in large corporate litigation cases is time-consuming and prone to errors due to the reliance on human annotators, with existing systems often focusing solely on textual content and lacking consistency, especially when only a small corpus of labeled documents is available.
Innovation Solution
A multi-modality classification method that extracts both text-based and social network features from electronic mail messages, using a semi-supervised approach to train classifiers, which combines representations from different modalities to improve classification accuracy and efficiency, even with limited labeled data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If human annotators review documents manually, then classification can be performed with full contextual understanding, but the process becomes time-consuming and costly
Solution Approach 1:
The patent introduces an automated classification system as an intermediary between the documents and human reviewers. The system extracts features from both textual content and social network structures, trains classifiers on labeled data, and generates preliminary classifications that guide human review, thereby reducing the time burden while maintaining accuracy through the collaborative human-machine approach
Solution Approach 2:
The system performs preliminary classification actions before human review by automatically analyzing document features and social network patterns. This preliminary sorting and grouping of documents allows human annotators to focus only on borderline cases or high-priority items, significantly reducing overall review time while maintaining high classification accuracy
2Device complexity
If only textual content is used for classification, then the system remains simple to implement, but classification accuracy deteriorates when labeled data is limited
Solution Approach 1:
The patent merges multiple data modalities including textual content features and social network structure features into a unified classification framework. By combining these complementary information sources, the system achieves higher classification accuracy with limited labeled data, as the social network features provide additional contextual signals that text alone cannot capture
Solution Approach 2:
The classification system is designed to handle multiple types of features and data sources universally. The framework can process both traditional textual features and social network-derived features through a common machine learning pipeline, making the system multi-functional and adaptable to different data types without requiring separate specialized systems
3Measurement precision
If multi-modality features are extracted and combined, then classification accuracy improves significantly, but the computational complexity and processing time increase
Solution Approach 1:
The patent segments the classification process into distinct modular components: text feature extraction, social network feature extraction, data integration, and classification. Each module handles a specific aspect of the multi-modality processing independently, which manages computational complexity by breaking down the overall task into manageable segments that can be optimized separately
4Measurement precision
If a large corpus of labeled documents is used for training, then classifier accuracy improves, but the cost and time for annotation increase significantly
Solution Approach 1:
The social network structure acts as an intermediary that provides additional training signals. By leveraging the relational information between documents (co-authorship, citation networks, organizational hierarchies), the system can achieve better generalization with fewer labeled examples, as the social context provides implicit supervision that reduces the need for extensive manual labeling
Data Source
AI summary
A classification apparatus, method, and computer program product for multi-modality classification are disclosed. For each of a plurality of modalities, the method includes extracting features from objects in a set of objects. The objects include electronic mail messages. A representation of each object for that modality is generated, based on its extracted features. At least one of the plurality of modalities is a social network modality in which social network features are extracted from a social network implicit in the set of electronic mail messages. A classifier system is trained based on class labels of a subset of the set of objects and on the representations generated for each of the modalities. With the trained classifier system, labels are predicted for unlabeled objects in the set of objects.