Question classification method and application thereof

Abstract

The invention discloses a question classification method which comprises the steps: carrying out the training under a large-scale statement through a pre-trained BERT model, obtaining the word vectorrepresentation of a question, and enabling each word to have different vector representations under different questions; processing a text sequence by utilizing a two-layer bidirectional GRU network structure, wherein an original word vector is combined with the first-layer output of the GRU network to serve as the second-layer input of the GRU by combining the thought of a residual network, so that the convergence rate of the model can be increased, important characteristics of questions are concerned by utilizing an attention mechanism, and the classification precision of the questions is improved. According to the method, the dependency problem of the question text can be further captured by utilizing the GRU model, and meanwhile, the attention mechanism is utilized to endow the features with greater influence on question classification with higher weights, and therefore the question classification precision in the medical question and answer system is further improved.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a question classification method and its application, in particular to a Chinese question classification method, in particular to a Chinese medical question answering system question classification method. Background technique [0002] With the development and popularization of network technology, the application prospect of medical question answering system is very broad. Users can learn about relevant medical knowledge or consult treatment methods for certain diseases through the medical question-and-answer system. At present, the mainstream construction method of the medical question answering system based on the knowledge base is based on the method of text matching, that is, firstly identify the relevant entities in the question and the corresponding category of the question, and then find the answer from the knowledge base according to the entity and the category of the q...

AI Technical Summary

Problems solved by technology

[0005] In view of the above defects, the present invention provides a method for classifying questions, which can be used in Chinese medical question answering systems. The method uses a multi-layer attention mechanism to obtain questions using a BERT (Bidirectional Encoder Representations from Transformers) model trained on a large-scale corpus. The word vector representation solves the ambiguity problem of existing word vectors, and proposes to combine the idea of ​​residual network to solve the defect of difficult training of cyclic neural network, speed up the training process, make the model converge faster, and improve the classification accuracy of questions

Images