Skip to content

Latest commit

 

History

History
47 lines (36 loc) · 2.76 KB

index.md

File metadata and controls

47 lines (36 loc) · 2.76 KB
title
SemEval 2024 Task: Emotion Discovery and Reasoning its Flip in Conversation (EDiReF)

The EDiReF shared task at SemEval 2024 is an amalgamation of three subtasks tasks- (i) Emotion Recognition in Conversation (ERC) in Hindi-English code-mixed conversations, (ii) Emotion Flip Reasoning (EFR) in Hindi-English code-mixed conversations, and (iii) EFR in English conversations. The definition for the tasks of ERC and EFR are as illustrated below:

  • ERC: Given a dialogue, ERC aims to assign an emotion to each utterance from a predefined set of possible emotions.
  • EFR: Given a dialogue, EFR aims to identify the trigger utterance(s) for an emotion-flip in a multi-party conversation dialogue.

For example, given a sample dialogue below:

Speaker Utterance
Sp1 I had an awful day today!
Sp2 Oh no! What happened?
Sp1 Somebody ate my sandwich!
Sp2 I can make you a new one right now!
Sp1 That would be great! Thanks!

ERC aims to assign emotions to each utterance

Speaker Utterance Emotion
Sp1 I had an awful day today! Sad
Sp2 Oh no! What happened? Sad
Sp1 Somebody ate my sandwich! Sad
Sp2 I can make you a new one right now! Joy
Sp1 That would be great! Thanks! Joy

The example dialogue has two emotion flips. Sp1's emotion changed from Sad to Joy while Sp2's emotion also shifted from Sad to joy. EFR aims to justify such emotion flips using triggers.

Speaker Utterance Emotion Trigger
Sp1 I had an awful day today! Sad 0
Sp2 Oh no! What happened? Sad 0
Sp1 Somebody ate my sandwich! Sad 0
Sp2 I can make you a new one right now! Joy 1
Sp1 That would be great! Thanks! Joy 0

Organization

The organizers are: