Named entity recognition method based on feature fusion

Abstract

A named entity recognition method based on feature fusion belongs to the field of computers, and text features, concept features and non-concept word features with different granularities are extracted and fused through two aspects, so that the named entity recognition accuracy is improved, and the calculation amount is reduced. The method comprises a data preprocessing module, a feature construction module, a named entity training network model module and a named entity classifier module, wherein the feature module comprises four sub-modules of semantic feature extraction, word feature extraction, character feature extraction and feature fusion. In this method, the temporal memory characteristics of the neural network model LSTM (Long Short-Term Memory) or GRU (Gated Recurrent Unit) are combined to consider the context information of the named entity task, and finally the softmax prediction entity category label is used.In the process of model construction, sparse data can be used asa training set, LSTM neural network models and GRU neural network models can be compared, and it is ensured that the method can achieve a satisfactory effect on entity recognition tasks.

Description

technical field [0001] The invention belongs to the field of computers and relates to a named entity recognition method based on feature fusion. Background technique [0002] In recent years, with the widespread application of artificial intelligence technology in the field of Natural Language Processing (NLP), people have explored more and more domain knowledge. Named entity recognition is the basis of domain knowledge, and it is also a crucial step. For example, named entity recognition is required in fields such as knowledge graph construction, text retrieval, text classification, and information extraction. [0003] Named Entity Recognition (NER) can be seen as a sequence labeling task, which uses the extracted information to find entities and classify them into a fixed set of categories. The two main approaches to the traditional NER problem are rule-based learning methods and supervised learning methods, where supervised learning methods dominate. Both rule-based and...

AI Technical Summary

Problems solved by technology

[0007] Although the above methods can complete the task of named entity recognition, the existing named entity recognition methods assume that there is no domain knowledge, and the features are only learned through the training set. However, in real life, most fields have partial domain knowledge. Although not perfect, these domain knowledge can help us better identify named entities in sparse data, and can also reduce the huge amount of calculations caused by inconsistent expressions to a certain extent.

Images