Emotion Cause Extraction (RECCON)
Extract emotion causes from conversational text based on RECCON (Poria et al., EMNLP 2020). Identify which utterances and specific spans caused an emotion expressed in dialogue.
Konfigurationsdateiconfig.yaml
# Emotion Cause Extraction (RECCON)
# Based on Poria et al., EMNLP 2020
# Paper: https://aclanthology.org/2020.emnlp-main.431/
# Dataset: https://github.com/declare-lab/RECCON
#
# RECCON focuses on recognizing emotion causes in conversations.
# For each utterance expressing an emotion, annotators identify:
# 1. The emotion being expressed
# 2. Which prior utterance(s) contain the cause
# 3. The specific span within those utterances that is causal
#
# Cause Types:
# - Self-contagion: Cause is in the same speaker's prior utterance
# - Inter-personal: Cause is in another speaker's utterance
# - No explicit cause: Emotion has no identifiable textual cause
#
# Annotation Guidelines:
# 1. First identify the target emotion in the highlighted utterance
# 2. Look backwards in the conversation for potential causes
# 3. Mark specific spans that directly trigger the emotion
# 4. A cause must be NECESSARY for the emotion (counterfactual test)
# 5. Multiple causes can exist for a single emotion
annotation_task_name: "Emotion Cause Extraction"
task_dir: "."
data_files:
- sample-data.json
item_properties:
id_key: "id"
text_key: "dialogue"
output_annotation_dir: "annotation_output/"
output_annotation_format: "json"
annotation_schemes:
# Step 1: Identify the emotion in the target utterance
- annotation_type: radio
name: emotion
description: "What emotion is expressed in the highlighted (last) utterance?"
labels:
- "Happiness"
- "Sadness"
- "Anger"
- "Fear"
- "Surprise"
- "Disgust"
- "Neutral"
tooltips:
"Happiness": "Joy, excitement, contentment, satisfaction, amusement"
"Sadness": "Sorrow, disappointment, grief, melancholy"
"Anger": "Frustration, irritation, annoyance, rage"
"Fear": "Anxiety, worry, nervousness, terror"
"Surprise": "Astonishment, amazement, shock (positive or negative)"
"Disgust": "Revulsion, disapproval, contempt"
"Neutral": "No clear emotion expressed"
# Step 2: Identify cause location
- annotation_type: radio
name: cause_type
description: "Where is the cause of this emotion located?"
labels:
- "In speaker's own prior utterance"
- "In other speaker's utterance"
- "In multiple utterances"
- "No explicit cause in text"
tooltips:
"In speaker's own prior utterance": "Self-contagion: the speaker's earlier words triggered their current emotion"
"In other speaker's utterance": "Inter-personal: another speaker's words triggered this emotion"
"In multiple utterances": "The cause spans multiple prior utterances"
"No explicit cause in text": "The emotion has no identifiable textual cause in the conversation"
# Step 3: Mark causal spans
- annotation_type: span
name: causal_spans
description: "Highlight the specific text spans that CAUSED the emotion"
labels:
- "Cause"
label_colors:
"Cause": "#ef4444"
tooltips:
"Cause": "The specific words/phrases that directly triggered the emotion (necessary condition)"
allow_overlapping: false
allow_all_users: true
instances_per_annotator: 50
annotation_per_instance: 2
allow_skip: true
skip_reason_required: false
Beispieldatensample-data.json
[
{
"id": "ece_001",
"dialogue": "Speaker A: I heard you got the promotion! Congratulations!\nSpeaker B: Thank you so much! I've been working towards this for three years.\n[TARGET] Speaker B: I'm absolutely thrilled - this is a dream come true!"
},
{
"id": "ece_002",
"dialogue": "Speaker A: Did you hear? They cancelled the concert we had tickets for.\nSpeaker B: What? We've been planning this for months!\n[TARGET] Speaker B: I can't believe it. I'm so disappointed."
}
]
// ... and 6 more itemsDieses Design herunterladen
Clone or download from the repository
Schnellstart:
git clone https://github.com/davidjurgens/potato-showcase.git cd potato-showcase/text/emotion-sentiment/emotion-cause-extraction potato start config.yaml
Details
Annotationstypen
Bereich
Anwendungsfälle
Schlagwörter
Problem gefunden oder möchten Sie dieses Design verbessern?
Issue öffnenVerwandte Designs
Character Identification on Multiparty Dialogues
Identification and linking of character mentions in TV show dialogue, combining span annotation with entity resolution for the main cast of Friends. Based on SemEval-2018 Task 4.
Dialogue Relation Extraction (DialogRE)
Extract relations between entities in dialogue. Based on Yu et al., ACL 2020. Identify 36 relation types between speakers and entities mentioned in conversations.
Emotion Discovery and Reasoning its Flip
Emotion discovery and flip reasoning task requiring annotators to identify the current emotion in a dialogue turn, and locate trigger spans and flip points where emotions change. Based on SemEval-2024 Task 10.