Dual-Word Embedding Model Considering Syntactic Information for Cross-Domain Sentiment Classification

The purpose of cross-domain sentiment classification (CDSC) is to fully utilize the rich labeled data in source domain help target perform even when are insufficient. Most existing methods focus on obtaining transferable semantic information but ignore syntactic information. performance BERT may decrease because transfer, and traditional word embeddings, such as word2vec, cannot obtain contextualized vectors. Therefore, achieving best results CDSC difficult only or word2vec used. In this paper, we propose a Dual-word Embedding Model Considering Syntactic Information for Cross-domain Sentiment Classification. Specifically, dual-word embeddings using word2vec. After performing embedding, pay closer attention information, mainly self-attention TextCNN. embedding obtained, graph network used extract document, mechanism important aspects. Experiments two real-world datasets show that our model outperforms other strong baselines.