Cross-Domain Named Entity Recognition via Ontology Graph
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning models for named entity recognition (NER) are domain-specific and require large volumes of annotated data, limiting their ability to perform accurately across different domains without extensive training, which is resource-intensive and time-consuming.
Innovation Solution
A system generates a machine learning model for cross-domain NER using a domain ontology graph to establish relationships between entities across domains, allowing the model to transfer knowledge from a source domain to a target domain without requiring annotated data from the target domain.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a machine learning model is trained using annotated data from a particular domain, then named entity recognition accuracy is improved for that domain, but the model cannot perform effectively for other domains without extensive retraining
Solution Approach 1:
The patent introduces an intermediary representation layer that maps domain-specific entities to a universal entity space. This intermediary layer enables the model to transfer knowledge across domains by translating domain-specific named entities into a common representation, allowing the model to maintain high accuracy in the source domain while gaining adaptability to target domains without extensive retraining.
Solution Approach 2:
The patent segments the named entity recognition task into two distinct components: domain-specific entity recognition (handled by the source domain training) and cross-domain entity mapping (handled by the intermediary representation). This segmentation allows the model to specialize in source domain accuracy while separately learning generalizable mapping patterns that enable cross-domain transfer.
2Adaptability or versatility
If large volumes of annotated data are collected for multiple domains, then cross-domain named entity recognition capability is improved, but computing resources and time required are substantially increased
Solution Approach 1:
The patent performs preliminary action by pre-training the model on source domain data to learn domain-specific entity patterns, then uses this pre-learned knowledge as a foundation for cross-domain transfer. This preliminary training eliminates the need to collect and process large volumes of annotated data for every target domain, significantly reducing the time and computational resources required for cross-domain adaptation.
Solution Approach 2:
The patent changes the parameter space by transforming domain-specific entity parameters into a universal parameter representation through the intermediary layer. This parameter transformation allows the model to generalize from source domain parameters to target domain parameters without requiring extensive retraining data, thereby reducing computational resources and training time while maintaining cross-domain capability.
3Measurement precision
If domain-specific machine learning models are created for each domain, then named entity recognition accuracy is improved, but device complexity and data construction effort are increased
Solution Approach 1:
The patent implements universality by designing a single machine learning model architecture that can perform named entity recognition across multiple domains. The intermediary representation layer serves as a universal interface that accommodates domain-specific variations while maintaining a consistent model structure, eliminating the need to create and maintain separate domain-specific models and reducing overall system complexity.
Data Source
AI summary
Methods and systems are presented for generating and using a machine learning model configured to perform cross-domain named entity recognition. The machine learning model is generated to accept a sentence associated with a target domain and to predict, for at least a word in the sentence, a corresponding entity associated with the target domain, without having been trained using training data associated with the target domain. In particular, the machine learning model is trained using only training data associated with a source domain. Based on derived relationships between entities associated with the source domain and entities associated with the target domain, the machine learning model is configured to transfer knowledge associated with the source domain to the target domain such that the machine learning model can map words within a sentence to entities associated with the target domain.