Entity Relation Classification Using Position-Aware BiGRU Attention
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing deep learning models for classifying entity semantic relations in electronic medical records have low accuracy, hindering efficient collection of past clinical information for medical decision-making.
Innovation Solution
A BiGRU+Attention model is used to train an entity semantic relation classification model, incorporating position embedding vectors to enhance the classification accuracy by considering the positional contribution of words relative to entities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing deep learning models are used for classifying entity semantic relations, then the model can be trained with sample data, but the classification accuracy remains low
Solution Approach 1:
The patent applies local quality by introducing position embedding vectors that assign different weights to words based on their positions relative to entities. Words at different positions (e.g., closer to entities vs. farther away) receive different embedding representations, allowing the model to locally adapt to the positional importance of each word in determining entity semantic relations.
Solution Approach 2:
The patent adds a positional dimension to the traditional word embedding by incorporating position embedding vectors. This transforms the feature representation from a single-dimensional word embedding to a multi-dimensional representation that includes both word semantics and positional information, enabling the model to capture the spatial relationships between words and entities.
2Device complexity
If traditional deep learning models are used without position embedding, then the model structure is simpler, but the model cannot capture positional contribution of words effectively
Solution Approach 1:
The patent applies preliminary action by pre-calculating position embedding vectors for different positional relationships between words and entities before the main classification process. These position embeddings are computed in advance based on the relative positions and are then integrated with word embeddings during training, allowing the model to efficiently utilize positional information without adding significant computational complexity to the core architecture.
Data Source
AI summary
The examples of the present disclosure provide a method and apparatus for classifying an entity semantic relation, a model training method and apparatus, an electronic device, and relate to the field of text identification technology. A first entity and a second entity in a corpus is determined and obtained. According to a first position distance between each word in the corpus and the first entity and a second position distance between each word in the corpus and the second entity, a feature vector corresponding to each word is obtained. Then, feature vectors corresponding to all words in the corpus are combined to obtain a model input vector corresponding to the corpus. Then an entity semantic relation type corresponding to the corpus is obtained by using the model input vector corresponding to the corpus an input of the entity semantic relation classification model.