Efficient dynamic feature adaptation for cross language sentiment analysis with biased adversarial training

Rui Li, Cheng Liu, Dazhi Jiang

Abstract: Highlights•A dynamic feature adaptation module is proposed for efficient sentiment adaptation.•A biased adversarial training strategy is proposed to improve the adaptation result.•Our model achieves SOTA results on multiple cross-lingual sentiment benchmarks.