Unsupervised adversarial training-based Chinese-Korean cross-lingual text classification method

This paper proposes a cross-lingual text classification model combining unsupervised word embedding mapping and adversarial training. Firstly, pre-trained monolingual word vectors were mapped into the same semantic space by a linear mapping method, and then adversarial training was performed using the source and target lang uages, thus guiding the classification model to be able to learn language-independent features for the purpose of improving cross-lingual text classification performance. The experimental results show that the method in this paper significantly improves the accuracy of cross-lingual text classification compared with the direct use of cross-lingua l word vectors.