Tibetan-Chinese Cross-Lingual Sentiment Classification Based on Adversarial Network
Tingting Zhang, Tao Jiang, Ruikang Shan · 2021
The high-quality sentiment classification data between different languages is quite uneven. As a low-resource language, the lack of annotated corpus seriously hinders the research of text sentiment classification in Tibetan. This paper uses the Chinese-Tibetan bilingual word vector to map the two languages to the same shared space, uses the language adversarial network to learn the joint features of Chinese and Tibetan, shares the emotional knowledge of Chinese and Tibetan, and builds a Tibetan-Chinese cross-lingual emotional classification based on the adversarial network. The model (Ti-Ch CLSC) achieves a better sentiment classification effect in a small number of Tibetan sentiment annotation sets. In order to verify the effectiveness of the model in this paper, the model is compared with several mainstream Tibetan sentiment classification methods. Experiments show that Ti-Ch CLSC significantly improves the effect of Tibetan text sentiment classification.