Sequence Labeling Matrix for Entity Text Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional sequence labeling methods for entity recognition, such as those based on conditional random fields (CRF), suffer from insufficient accuracy in recognizing entity labels.
Innovation Solution
A method and apparatus for sequence labeling that determine the start position of an entity text within a target text, generate a first matrix indicating focusable weights of words, and use this matrix to train a named entity recognition model for improved entity label recognition, incorporating position and range information to enhance recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional CRF-based sequence labeling methods are used, then the processing flow is simple, but the entity label recognition accuracy is insufficient
Solution Approach 1:
The patent introduces a new dimension of position and range information into the sequence labeling task. By constructing a position matrix that captures both positional relationships and span ranges of entities, the model moves beyond conventional one-dimensional sequence processing to a two-dimensional representation that simultaneously encodes where entities occur and how long they are, thereby improving recognition accuracy without fundamentally changing the CRF framework
Solution Approach 2:
The patent performs preliminary processing by pre-calculating the position matrix before feeding data to the CRF model. This position matrix, which encodes start positions, end positions, and spans of entities, is prepared in advance as additional feature input. This preliminary action enriches the feature space with positional and range information that the CRF model can then utilize to improve labeling accuracy
2Measurement precision
If position and range information are introduced into the model, then entity label recognition accuracy improves, but the complexity of feature extraction increases
Solution Approach 1:
The patent segments the feature extraction process into distinct components: first extracting basic sequence features (words, characters), then separately computing position-based features (start position, end position, span length), and finally combining these segmented features into the CRF model. This segmentation makes the complex feature extraction process more manageable and modular, with each component serving a specific function
Data Source
AI summary
A method and an apparatus for sequence labeling on an entity text, and a non-transitory computer-readable recording medium are provided. In the method, a start position of an entity text within a target text is determined. Then, a first matrix is generated based on the start position of the entity text. Elements in the first matrix indicates focusable weights of each word with respect to other words in the target text. Then, a named entity recognition model is generated using the first matrix. The named entity recognition model is obtained by training using first training data, the first training data includes word embeddings corresponding to respective texts in a training text set, and the texts are texts whose entity label has been labeled. Then, the target text is input to the named entity recognition model, and probability distribution of the entity label is output.