ANA at SemEval-2019 Task 3: Contextual Emotion detection in Conversations through hierarchical LSTMs and BERT

TL;DR

This paper introduces a hierarchical LSTM model for contextual emotion detection in conversations, outperforming BERT and achieving top rankings in a competitive SemEval task.

Contribution

The novel Hierarchical LSTM for Contextual Emotion Detection (HRLCE) effectively captures conversational context, improving emotion classification accuracy over existing models like BERT.

Findings

HRLCE outperforms BERT in emotion detection accuracy.

Combining HRLCE and BERT yields the best overall performance.

Achieved 5th place among 165 teams in SemEval-2019 Task 3.

Abstract

This paper describes the system submitted by ANA Team for the SemEval-2019 Task 3: EmoContext. We propose a novel Hierarchical LSTMs for Contextual Emotion Detection (HRLCE) model. It classifies the emotion of an utterance given its conversational context. The results show that, in this task, our HRCLE outperforms the most recent state-of-the-art text classification framework: BERT. We combine the results generated by BERT and HRCLE to achieve an overall score of 0.7709 which ranked 5th on the final leader board of the competition among 165 Teams.