Emotion Prediction Oriented Method With Multiple Supervisions for Emotion-Cause Pair Extraction

Emotion-cause pair extraction (ECPE) task aims to extract all the pairs of emotions and their causes from an unannotated emotion text. The previous works usually emotion-cause two perspectives cause. However, is more crucial ECPE than cause extraction. Motivated by this analysis, we propose end-to-end approach oriented toward prediction (EPO-ECPE), aiming fully exploit potential enhance Considering strong dependence between extraction, a synchronization mechanism share improvement in training process. That is, can facilitate then results also be used improve accuracy simultaneously. For divide it into genuine supervision fake supervision, where learns with possibility pairs. In contrast, other way, extracted directly pair, thereby reducing difficulty Experimental show that our outperforms 13 compared systems achieves new state-of-the-art performance.