metadata
base_model:
- google-bert/bert-base-uncased
Model Card: Emotion Detection Model
Model Overview
This model is trained to detect nuanced emotions in text data, specifically focusing on dialogues from the TV show Friends and additional curated online content. By leveraging advanced deep learning architectures such as BERT and GPT, the model performs multi-label classification to identify up to 2–3 emotions per input dialogue. The targeted emotions include:
- Happiness
- Sadness
- Anger
- Fear
- Surprise
- Disgust
- Love
- Excitement
- Anticipation
- Contentment
- Confusion
- Frustration
- Nostalgia
Intended Use
- Primary Use Case: Emotion detection in textual data, particularly for analyzing media content like TV shows, movies, or social media.
- Scope of Application: Suitable for analyzing dialogue or conversational data to identify emotional trends, character development, or audience engagement.
Dataset
Source
- Dialogues from the TV show Friends.
Labeling
- Dialogues were labeled with 2–3 emotions per text using OpenAI's ChatGPT API.
- Labels represent a rich variety of emotional states.
Methodology
Labeling Process
- ChatGPT API was used to classify text into multiple emotions.
- Outputs were reviewed for accuracy and adjusted to minimize misclassification.
Model Architecture
- Pre-trained Models Used: BERT
- Fine-tuned on labeled dialogue data for emotion detection.
- Supports multi-label classification to capture multiple emotions simultaneously.