Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

EmoBERTa: Speaker-Aware Emotion Recognition in Conversation with RoBERTa

About

We present EmoBERTa: Speaker-Aware Emotion Recognition in Conversation with RoBERTa, a simple yet expressive scheme of solving the ERC (emotion recognition in conversation) task. By simply prepending speaker names to utterances and inserting separation tokens between the utterances in a dialogue, EmoBERTa can learn intra- and inter- speaker states and context to predict the emotion of a current speaker, in an end-to-end manner. Our experiments show that we reach a new state of the art on the two popular ERC datasets using a basic and straight-forward approach. We've open sourced our code and models at https://github.com/tae898/erc.

Taewoon Kim, Piek Vossen• 2021

Related benchmarks

TaskDatasetResultRank
Emotion Recognition in ConversationIEMOCAP (test)
Weighted Average F1 Score67.42
154
Emotion Recognition in ConversationMELD
Weighted Avg F165.61
137
Conversational Emotion RecognitionIEMOCAP
Weighted Average F1 Score67.3
129
Emotion RecognitionIEMOCAP--
71
Emotion DetectionMELD (test)
Weighted-F10.652
32
Emotion Recognition in ConversationMELD 1.0 (test)
Weighted F165.61
17
Emotion Recognition in ConversationIEMOCAP 1.0 (test)
Weighted F1 Score68.57
17
Multimodal Emotion Recognition in ConversationMELD
Neutral Accuracy78.9
12
Coarse Sentiment ClassificationHotel Review dataset
Coarse Acc81.18
12
Fine Sentiment ClassificationHotel Review dataset
F-Score Accuracy66.7
12
Showing 10 of 11 rows

Other info

Code

Follow for update