Research on Tibetan Verb Character Recognition Based on Bert Model

The traditional recognition methods of Tibetan verbs are mainly based on rule matching and statistical features,however,these methods are often limited by the complex structure of Tibetan language, rules are not fixed and other factors,the accuracy rate is low and the generalization ability is limited. The two-way information of the input text is taken into account in the coding process of Bert model, which avoids the loss of one-way information in the traditional model. By the Bert model,we have realized the accurate recognition of Tibetan verbs. The experimental results show that our method has a significant improvement on the recognition of verb error character, the accuracy and generalization ability are significantly improved compared with the traditional methods. The results of this study provide new ideas and methods for the research and application of Tibetan text processing and contribute to the development of Tibetan natural language processing technology.