Meld

multimodal
+
+
+
+
About

MELD (Multimodal EmotionLines Dataset) is a comprehensive multimodal multi-party dataset for emotion recognition in conversation, containing over 1,400 dialogues and 13,000 utterances from Friends TV series. Each utterance is labeled with seven emotions (anger, disgust, sadness, joy, neutral, surprise, fear) and sentiment annotations. The dataset provides textual, audio, and visual modalities, enabling research in conversational emotion recognition, multimodal sentiment analysis, and affective dialogue systems.

+
+
+
+
Evaluation Stats
Total Models1
Organizations1
Verified Results0
Self-Reported1
+
+
+
+
Benchmark Details
Max Score1
Language
en
+
+
+
+
Performance Overview
Score distribution and top performers

Score Distribution

1 models
Top Score
57.0%
Average Score
57.0%
High Performers (80%+)
0

Top Organizations

#1Alibaba Cloud / Qwen Team
1 model
57.0%
+
+
+
+
Leaderboard
1 models ranked by performance on Meld
LicenseLinks
Mar 27, 2025
Apache 2.0
57.0%
+
+
+
+
Resources