Natural Language Analysis TRIZ Cases
- Transformer Model for Automatic Document Structure Identification
A transformer-over-transformer model generates entity embeddings to identify structural classes in plain text.
- Commodity Short Text Core Word Extraction via Vector Clustering
A commodity short text core word extraction method uses document vector clustering and weight determination to identify key terms.
- Entity Disambiguation via Knowledge Base Matching
A method modifies named entities using predefined rules to generate candidate matches against a knowledge base.
- Visio-Linguistic Reasoner Using Knowledge Graph Embeddings
Contextual language model reasoners process image captions and confidence scores to generate compatible embeddings for downstream tasks.
- Intent Identification Using Historical Context Vectors
Historical vectors merge current and past statement vectors to resolve accuracy issues from ignoring context, reducing inconsistent responses.
- Business Categorization via Invoice Clustering
Automated clustering of invoice vectors generates standardized business categories, resolving inconsistent data entry and improving search accuracy.
- App-Based Reminder Generation via Text Analysis
Machine learning models extract task details from app communications to automate trigger detection, reducing manual entry complexity.
- Predictive Aspect Formatting for Communication Contexts
Analyzes communication context data to automatically apply visual formatting changes, reducing manual user burden and improving message conveyance efficiency.
- Abstractive Summarization Model Factual Consistency via Virtual Samples
Virtual sample pairs contradict summary model knowledge to improve factual consistency across varying original text content.