Model Card: Sentiment Classifier (DistilBERT - SST-2)

Overview

This model is a fine-tuned version of distilbert-base-uncased on the SST-2 dataset, designed for binary sentiment classification: labeling text as either positive or negative.

It’s fast, compact, and suitable for real-time inference tasks such as social media monitoring, customer feedback triage, and lightweight embedded NLP.

Use Cases

Detecting sentiment in tweets, reviews, or comments
Routing customer support tickets by tone
Analyzing product sentiment in e-commerce or app stores
Monitoring brand perception over time

Example

Input: "This new update is amazing — so much faster!"
Output: Positive

Input: "This feature is broken and support isn't helping."
Output: Negative

---

## Strengths

- Extremely lightweight: good for mobile and low-latency use
- Fine-tuned on a benchmark sentiment dataset (SST-2)
- Strong out-of-the-box performance for informal English

---

## Limitations

- Binary only (positive/negative) — no neutral or nuanced emotion
- Trained on English movie reviews — may misinterpret sarcasm, cultural tone, or domain-specific feedback
- Not ideal for clinical, legal, or safety-critical sentiment tasks

---

## Model Details

- Architecture: DistilBERT
- Base model: `distilbert-base-uncased`
- Fine-tuning dataset: SST-2 (Stanford Sentiment Treebank)
- Max input: 512 tokens
- Classes: `Positive`, `Negative`

---

## License

MIT License — free to use, adapt, and deploy commercially.

---

## Authorship Note

This model card was written by [Sarah Mancinho](https://huggingface.co/Sarah-h-h) as part of a public AI/LLM contribution series on Hugging Face.

Original model: [`distilbert-base-uncased-finetuned-sst-2-english`](https://huggingface.co/distilbert-base-uncased-finetuned-sst-2-english)

---

## Citation