T5 Small Intent-Slot Model
This is a fine-tuned T5 model designed for Intent Detection and Slot Filling — a core task in natural language understanding (NLU) for chatbots, virtual assistants, and conversational AI.
What does this model do?
Imagine you’re teaching a smart assistant to understand user requests like:
- “Book a hotel in London for 3 nights.”
- “Find me an Italian restaurant nearby.”
- “What’s the weather tomorrow in Paris?”
This model reads the input sentence and simultaneously figures out:
- The user's intent (e.g., booking, searching)
- The slots (key details like location, date, type)
It outputs a structured sequence that identifies these elements, so your app can respond intelligently.
Model Details
- Based on T5 small architecture (6 layers, 512 hidden size)
- Trained for conditional generation of intents and slots from text
- Uses SentencePiece tokenizer with custom added tokens
- Model weights stored as safetensors for efficiency and safety
Files in this repository
| File | Description |
|---|
config.json | Model architecture and params |
generation_config.json | Text generation settings |
model.safetensors | Model weights |
tokenizer_config.json | Tokenizer settings |
spiece.model | SentencePiece tokenizer model |
added_tokens.json | Custom tokens added during training |
special_tokens_map.json | Mapping of special tokens |
.gitattributes | Git LFS config for large files |
README.md | This documentation |
