Team Phoenix at WASSA 2021: Emotion Analysis on News Stories with Pre-Trained Language Models.

摘要: Emotion is fundamental to humanity. The ability perceive, understand and respond social interactions in a human-like manner one of the most desired capabilities artificial agents, particularly social-media bots. Over past few years, computational understanding detection emotional aspects language have been vital advancing human-computer interaction. WASSA Shared Task 2021 released dataset news-stories across two tracks, Track-1 for Empathy Distress Prediction Track-2 Multi-Dimension prediction at essay-level. We describe our system entry (for both Track-2), where we leveraged information from Pre-trained models Track-specific Tasks. Our proposed achieved an Average Pearson Score 0.417, Macro-F1 0.502 Track 1 2, respectively. In leaderboard, secured fourth rank second 2.