Chinese Medical Question Answering Model Based on Lattice CNNs

Abstract

The application demand of intelligent medicine is growing, and medical question answering (QA) is an important part of it. Chinese medical question answering has difficulty in word segmentation, which leads to insufficient feature extraction. To solve this problem, this paper constructs a user-defined dictionary and uses LCN s to build a Chinese medical question answering model. First, a user-defined dictionary for Chinese medical QA is constructed based on ICD-10 professional medical thesaur-us, and the electronic medical record is processed by using Chine-se word segmentation tool and user-defined dictionary. Secondly, use the EMR (electronic medical record) after word segmentation as the data of word vector training, and use the Skip-gram model to train the medical word vector. Finally, taking the medical word vector as the input of the model, LCNs are used to extract sentence features, calculate the similarity between question feat-ures and answer features, and then train the Chinese medical question answering model. Experiments show that the accuracy of LCNs model can reach 88%, and its performance is better than other deep learning methods such as LSTM.