Tibetan text classification based on RNN

Abstract In this paper, a deep learning RNN model is used to classify Tibetan texts. The core idea first preprocess the news corpus, and then use syllables construct syllable table based on lexical grammatical structure of Tibetan, embed in sentence, represent each as fixed Numerical vector. Secondly, cyclic neural network constructed. First, text different lengths filled or truncated into sequence length uniform length. For input text, vector representation time step train model. test samples were evaluate accuracy classification by introducing recall rate, precision rate F-test. Finally, compared with traditional machine Logistic algorithm, polynomial naive Bayes algorithm KNN results show that has better effect.