Word Embedding Region Definition for Custom Entity Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for processing raw and unstructured medical text data require significant human intervention for entity recognition and annotation, limiting efficiency and scalability, and struggle to identify custom entity types beyond predefined categories.
Innovation Solution
A computing system employs unsupervised learning techniques using word embeddings to automatically identify entities in medical text data, expands the recognition of entities through dynamically expanding windows of words, utilizes character embeddings for unknown words, and defines custom tags by creating regions in the word embedding space for custom entity types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If human intervention is used for entity recognition and annotation, then accuracy can be maintained, but processing efficiency and scalability deteriorate
Solution Approach 1:
The system enables unsupervised learning where the model automatically performs entity recognition and annotation without human intervention. The word embedding model processes medical text data independently, identifying entities and assigning custom tags autonomously, thereby eliminating the need for manual annotation while maintaining scalability.
Solution Approach 2:
The patent replaces the mechanical human annotation process with an automated computational system. The word embedding model, trained on medical text data, substitutes human experts in performing entity recognition and classification tasks, achieving both high efficiency and acceptable accuracy through algorithmic processing.
2Ease of manufacture
If predefined entity categories are used, then processing simplicity is maintained, but adaptability to custom entity types deteriorates
Solution Approach 1:
The system dynamically adapts to custom entity types by allowing flexible definition of custom tags specific to different medical domains. The word embedding model can be retrained or fine-tuned to recognize new entity types without requiring complete system redesign, enabling both simplicity for standard cases and adaptability for custom requirements.
Solution Approach 2:
The patent enables customization of entity types by changing the parameter definitions in the knowledge base. Users can define custom tags and modify entity category parameters to suit specific medical domains, allowing the same base system to handle both predefined and custom entity types through parameter adjustment rather than structural changes.
3Extent of automation
If word embeddings are used for entity recognition, then automation is improved, but precision in identifying custom entities deteriorates
Solution Approach 1:
The system performs preliminary training of the word embedding model on domain-specific medical text data before deployment. By pre-training on relevant corpora and incorporating domain knowledge during the training phase, the model achieves better precision in identifying custom entities when automation is applied, rather than using generic pre-trained embeddings.
Solution Approach 2:
The system incorporates feedback mechanisms where the model's performance on custom entity recognition is evaluated and used to refine the word embeddings. Through iterative training and validation, the model adjusts its parameters to improve precision in identifying custom entities while maintaining the benefits of automation.
Data Source
AI summary
Some embodiments provide a non-transitory machine-readable medium that stores a program. The program receives a plurality of sets of words. Each set of words in the plurality of sets of words includes a word annotated as being an entity having a same custom entity type. The program further determines a plurality of word embeddings in a word embedding space for the plurality of annotated words. The program also defines a region in the word embedding space based on the received plurality of word embeddings. The program further receives a set of words. The program also determines a word embedding for a subset of the set of words. The program further determines whether the word embedding falls within the defined region in the word embedding space. Upon determining that the word embedding falls within the defined region in the word embedding space, the program also determines that the subset of the set of words represents an entity having the custom entity type.