Error Correction of Tibetan Verbs Based on Deep Learning

Abstract

Verbs are the core of the semantic structure represented by Tibetan sentences, and automatic error correction of Tibetan verbs is one of the important research topics in Tibetan language processing. In this paper, we propose a Bi-LSTM neural network model for Tibetan verb error correction by analyzing the usage rules of verbs in Tibetan, summarizing the grammatical, semantic and spelling features of verbs, and proposes a Bi-LSTM neural network model for Tibetan verb error correction based on these features, which not only extracts various features of verbs, but also capture the contextual information of verbs through Bi-LSTM neural network. The method proposed in this work solves the drawbacks of the traditional methods of low generalization and inability to obtain long-distance contextual implicit information. The experimental results show that the accuracy, recall and F1 values of the proposed method on the test set reach 97.3%,95.7% and 96.9%, respectively, indicating the effectiveness of the proposed method on the task of automatic error correction for Tibetan verbs.