BlessedRemzy/nostalgic-persuasive-models

Nostalgic Persuasive Models

This repository contains trained models for the Nostalgic Persuasive Model research project.

Models Included

1. Emotion Model (emotion_model/)

• Architecture: RoBERTa-based text classification
• Task: Multi-class emotion classification from text
• Format: Hugging Face Transformers (safetensors)

2. Stress Detection Model (stress_detection_mental_roberta/)

• Architecture: Mental-RoBERTa fine-tuned
• Task: Binary stress detection from text
• Format: Hugging Face Transformers (safetensors)

3. Movie Recommender (movie_recommender/)

• Architecture: LightFM hybrid collaborative filtering
• Task: Movie recommendation based on user preferences
• Format: Pickle files (.pkl)
• Files:
  • lightfm_model.pkl - Trained LightFM model
  • lightfm_dataset.pkl - Dataset object for mappings
  • item_features.pkl - Item feature matrix

4. Song Recommender (song_recommender/)

• Architecture: Content-based filtering with TF-IDF
• Task: Music recommendation based on audio features and lyrics
• Format: Joblib files
• Files:
  • audio_scaler.joblib - StandardScaler for audio features
  • genre_encoder.joblib - Label encoder for genres
  • tfidf_vectorizer.joblib - TF-IDF vectorizer for lyrics

5. Contextual Bandit (bandit/)

• Architecture: Thompson Sampling contextual bandit
• Task: Personalized content selection optimization
• Format: Joblib files

Usage

# For Transformers models (emotion, stress)
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model = AutoModelForSequenceClassification.from_pretrained("BlessedRemzy/nostalgic-persuasive-models/emotion_model")
tokenizer = AutoTokenizer.from_pretrained("BlessedRemzy/nostalgic-persuasive-models/emotion_model")

# For pickle/joblib models
import joblib
from huggingface_hub import hf_hub_download

model_path = hf_hub_download(repo_id="BlessedRemzy/nostalgic-persuasive-models", filename="movie_recommender/lightfm_model.pkl")
model = joblib.load(model_path)

License

MIT License