distilbert-metacritic-sentiment-classifier

DistilBERT Metacritic Sentiment Classifier

Description

This model performs sentiment analysis on video game reviews from Metacritic.

It has been fine-tuned from distilbert-base-uncased using the dataset: Wada1/Metacritic_Games_Reviews_Dataset

The model classifies reviews into:

• Positive
• Negative

Neutral reviews were removed before training.

Dataset

Features used:

• Review → review text
• Sentiment → target label

Label mapping:

• Positive → 1
• Negative → 0

Neutral reviews were excluded.

Evaluation Results

Performance on evaluation split:

• Accuracy: 0.8953
• F1-score: 0.9164
• Precision: 0.8886
• Recall: 0.9459

Usage

Example:

from transformers import pipeline

classifier = pipeline( "text-classification", model="angelhm/distilbert-metacritic-sentiment-classifier" )

result = classifier("This game is amazing. Great story and gameplay.") print(result)

Labels

• Positive → positive review
• Negative → negative review

Limitations

• Only works with English text
• Trained only on Metacritic video game reviews
• May not generalize well to other domains
• Neutral sentiment is not predicted

Demo

You can test the model interactively in the Hugging Face Space:

https://huggingface.co/spaces/angelhm/distilbert-metacritic-sentiment-classifier

Write a video game review in English and the model will predict whether the sentiment is Positive or Negative.