You can deploy this model and start earning money today!

GuruPRaju/
app-review-sentiment-distilbert

Name: app-review-sentiment-distilbert
Availability: InStock

safetensors

distilbert

Views

No views yet

📖 README: App Review Sentiment Classifier

This repository hosts a fine-tuned DistilBERT model for classifying app review sentiment into 1 to 5 star ratings.

Dataset

The model was trained using the sealuzh/app_reviews dataset, available on Hugging Face Datasets. This dataset contains a large collection of app reviews with corresponding star ratings.

Model Architecture

The base model used is distilbert-base-uncased, a smaller, faster, and lighter version of BERT. It was fine-tuned for sequence classification with 5 labels (corresponding to 1-5 stars).

Training Details

Base Model: distilbert-base-uncased
Task: Sequence Classification (5 classes)
Epochs: 3
Learning Rate: 2e-5
Batch Size (Train): 16
Batch Size (Eval): 32
Max Sequence Length: 128
Metrics: Accuracy

Preprocessing

App reviews were tokenized using the distilbert-base-uncased tokenizer. Reviews were truncated to a maximum length of 128 tokens and padded to ensure uniform input size. Star ratings (1-5) were converted to 0-indexed labels (0-4) for training.

Usage

To use this model for inference, you can leverage the transformers library:

python

from transformers import pipeline

classifier = pipeline(
    "text-classification",
    model="GuruPRaju/app-review-sentiment-distilbert", # Using the actual repo_id here
    tokenizer="GuruPRaju/app-review-sentiment-distilbert"
)

review = "This app is absolutely amazing, highly recommend!"
result = classifier(review)
stars = int(result[0]["label"].split("_")[1]) + 1
print(f"Review: {review}
Predicted Stars: {stars}")

Acknowledgments

Hugging Face for the transformers library and Hub.
The creators of the sealuzh/app_reviews dataset.