Nested Entity Recognition via Self-Consistent Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional entity recognition approaches face challenges in accurately identifying sub-entities within nested entities due to poor mobility and loss of context information, often screening out sub-entities that are inconsistent with contextual information, leading to ineffective information output in real business scenarios.
Innovation Solution
A computer-implemented method using multi-label classification in a neural network for nested named entity recognition (NER) that employs multi-task learning and self-consistent verification, involving joint context verbs or predicates, replacement entities, and a translation model for confidence verification to constrain generated entities and improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional entity recognition approaches are used to identify sub-entities in nested entities, then the processing is simpler, but the accuracy deteriorates due to loss of context information and poor mobility
Solution Approach 1:
The patent applies nested doll by implementing nested entity recognition where sub-entities are identified within parent entities. The system processes nested structures by recursively identifying entities at different levels, allowing sub-entities to be recognized while preserving their contextual relationships with parent entities, thus preventing context information loss while maintaining recognition accuracy.
Solution Approach 2:
The patent introduces an additional dimension by incorporating confidence scores and verification mechanisms beyond simple entity classification. The system evaluates multiple candidate sub-entities with confidence scoring and performs self-consistent verification, adding a verification dimension that improves accuracy without losing context information.
2Productivity
If traditional entity recognition approaches screen out sub-entities inconsistent with context, then processing efficiency is improved, but the usefulness of information output deteriorates in real business scenarios
Solution Approach 1:
The patent implements feedback through self-consistent verification where the system checks whether identified sub-entities are consistent with their contextual environment. The verification mechanism provides feedback by comparing candidate sub-entities against contextual constraints and confidence thresholds, allowing the system to maintain high processing efficiency while ensuring information usefulness through automated consistency checking.
Solution Approach 2:
The patent changes parameters by introducing confidence score thresholds and verification criteria that can be adjusted based on business scenario requirements. The system allows flexible parameter tuning to balance processing efficiency and information usefulness, enabling adaptation to different real-world applications without sacrificing either productivity or reliability.
3Adaptability or versatility
If multi-label classification with multiple candidate types is used, then the adaptability improves, but the device complexity increases due to retraining requirements
Solution Approach 1:
The patent applies preliminary action by pre-training the neural network with multi-label classification capabilities to handle multiple candidate types. The system prepares the model in advance with versatile classification abilities, allowing it to adapt to different entity types without requiring complex retraining for each new scenario, thus reducing operational complexity while maintaining adaptability.
Solution Approach 2:
The patent implements universality by designing a single neural network model that can handle multiple entity types and classification labels simultaneously. The multi-label classification framework enables the model to perform multiple functions across different entity categories without requiring separate specialized models, reducing overall system complexity while maintaining high adaptability to various entity types.
Data Source
AI summary
An approach is provided for improving a named entity recognition. Using a multi-label classification in a neural network, a sub-entity is identified in an original sentence. First and second labels are determined indicating first and second candidate types of the sub-entity. First and second replacement sentences are generated. The first replacement sentence replaces the sub-entity in the original sentence with a first sub-entity of the first candidate type. The second replacement sentence replaces the sub-entity in the original sentence with a second sub-entity of the second candidate type. First and second confidence scores for the first and second replacement sentences are determined. Based on the first confidence score exceeding the second confidence score by more than a threshold amount, the neural network is retrained by selecting the first instead of the second candidate type as the sub-entity type.