---
tags:
- text-classification
- intent-classification
- moviebot
- distilbert
- huggingface-pipeline
- gradio-app
datasets:
- custom
language: en
license: mit
widget:
  - text: "Can you recommend a good comedy?"
  - text: "Who directed Inception?"
  - text: "Do you like movies?"
---

# DistilBERT Intent Classifier for Movie & TV Assistant

This model is a fine-tuned DistilBERT-based intent classifier for a conversational movie recommendation assistant. It classifies user queries into one of three intents:

- **`generic`** – general small talk or casual movie mentions
- **`recommendation`** – requests for movie or TV suggestions
- **`factual`** – questions about movie facts, cast, streaming availability, etc.

## Model Details

- **Base model:** `distilbert-base-uncased`
- **Training data:** 6,000+ custom-labeled queries across all 3 intents
- **Special augmentations:**
  - Out-of-domain recommendation phrasing (e.g. restaurants, gadgets)
  - Thematic recommendation queries (e.g. “movies for Valentine’s Day”)
- **Use case:** Used inside a RAG-based chatbot for intent-routing and retrieval logic

## Example Predictions

| Input | Predicted Intent |
|-------|------------------|
| "Can you recommend a good horror movie?" | recommendation |
| "Who directed Parasite?" | factual |
| "Do you like sci-fi shows?" | generic |
| "Suggest restaurants in Rome?" | generic (OOD) |

## Usage

```python
from transformers import pipeline

classifier = pipeline("text-classification", model="your-username/intent-classifier-distilbert-moviebot")

query = "Can you suggest a good action movie?"
result = classifier(query, top_k=None)
print(result)