MELD: A Multimodal Multi-Party Dataset for Emotion Recognition in Conversations

Soujanya Poria, Devamanyu Hazarika, Navonil Majumder, Gautam Naik, Erik Cambria, Rada F. Mihalcea · 2019

Read MELD: A Multimodal Multi-Party Dataset for Emotion Recognition in Conversations by Soujanya Poria, Devamanyu Hazarika, Navonil Majumder, Gautam Naik, Erik Cambria, Rada F. Mihalcea (2019) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/openalex-w2895615590.txt.

Description

Soujanya Poria, Devamanyu Hazarika, and Navonil Majumder, with colleagues, introduced the MELD dataset at the 2019 ACL meeting. Emotion recognition in conversations is a prized, hard problem for artificial intelligence. Existing emotional dialogue databases held at most two speakers per conversation. Real conversations, though, are multi-party, and feelings unfold across turns and speakers.

MELD extends the EmotionLines dataset using dialogues from the television series Friends. It holds about thirteen thousand utterances from over fourteen hundred dialogues. Every utterance carries emotion and sentiment labels. Text, audio, and video are all included, so models can read faces and voices too. The authors supply strong multimodal baseline systems. Their results show context and multiple modalities both matter for accuracy. The dataset is public, giving the field a shared benchmark for conversational emotion.

Catalog record

CREATORSoujanya Poria, Devamanyu Hazarika, Navonil Majumder, Gautam Naik, Erik Cambria, Rada F. Mihalcea
YEAR2019
COLLECTIONPAPERS
FORMATTXT
FILE SIZE41 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEOpenAlex / Annual Meeting of the Association for Computational Linguistics (ACL)

Source: https://doi.org/10.18653/v1/p19-1050

Browse the full library · About the Tower of Babel