TiBERT: Tibetan Pre-trained Language Model

TL;DR

TiBERT is a Tibetan-specific pre-trained language model trained on large-scale Tibetan text data, achieving state-of-the-art results in downstream NLP tasks like text classification and question generation, addressing the resource scarcity in Tibetan NLP.

Contribution

This paper introduces TiBERT, the first monolingual Tibetan pre-trained language model trained on a large Tibetan corpus with a specialized vocabulary, improving NLP task performance for Tibetan.

Findings

TiBERT outperforms classic and multilingual models in downstream tasks.

Constructed a vocabulary covering 99.95% of Tibetan corpus words.

Demonstrated effectiveness of Tibetan monolingual pre-training.

Abstract

The pre-trained language model is trained on large-scale unlabeled text and can achieve state-of-the-art results in many different downstream tasks. However, the current pre-trained language model is mainly concentrated in the Chinese and English fields. For low resource language such as Tibetan, there is lack of a monolingual pre-trained model. To promote the development of Tibetan natural language processing tasks, this paper collects the large-scale training data from Tibetan websites and constructs a vocabulary that can cover 99.95$\%$ of the words in the corpus by using Sentencepiece. Then, we train the Tibetan monolingual pre-trained language model named TiBERT on the data and vocabulary. Finally, we apply TiBERT to the downstream tasks of text classification and question generation, and compare it with classic models and multilingual pre-trained models, the experimental results show that TiBERT can achieve the best performance. Our model is published in http://tibert.cmli-nlp.com/