No abstract

No abstract

In current years, there are many open corpus and research results for Chinese clinical natural language processing (for short NLP) of Biomedical and Traditional Chinese Medicine. However, the research in this area for Traditional Tibetan Medicine lags behind. We know that Traditional Tibetan Medicine has its own unique set of theory system in treating and saving people. Therefore, it is imperative to speed up the research on the Tibetan Clinical Natural Language Processing. Medical named entity recognition is an important subtask of Clinical Natural Language Processing. So Tibetan medical named entity recognition is an urgent and basic research work for Traditional Tibetan Medicine. Due to the scarcity of labeled datasets, the Medical Named Entity Recognition task of Traditional Tibetan Medicine clinical text is still an unvisited researching area. In this work, we firstly manually construct a labeled dataset for this task and then explore this area with deep learning approaches by designing a Tibetan Lattice-LSTM-CRF neural network architecture. To further improve the model performance, we also incorporate both syllable and word level pre-trained representation. The final empirical results show that the proposed models could produce accuracy rate, recall rate and F1 values of 91.89%, 93.15% and 92.52%, respectively on our test set, which shows the validity of the model in the paper.

No abstract