Pre-Conference Talk by YU Jianfei | Leveraging Auxiliary Tasks for Document-Level Cross-Domain Sentiment Classification

Leveraging Auxiliary Tasks for Document-Level Cross-Domain

Sentiment Classification

With the recent trend of deep learning, a large amount of neural network models have been proposed for sentiment classification. However, as with any supervised learning method, the neural network models also suffer from the domain adaptation problem, where training data and test data come from different domains. In this paper, we study domain adaptation with a state-of-the-art hierarchical neural network for document-level sentiment classification. We first design a new auxiliary task based on sentiment scores of domain-independent words. We then propose two neural network architectures to respectively induce document embeddings and sentence embeddings that work well for different domains. When these document and sentence embeddings are used for sentiment classification, we find that with both pseudo and external sentiment lexicons, our proposed methods can perform similarly to or better than several highly competitive domain adaptation methods on a benchmark dataset of product reviews.

This a pre-conference talk for 8th International Joint Conference on Natural Language Processing (IJCNLP 2017).