Model Overview

The model used is t5-small, a lightweight transformer model from the T5 (Text-To-Text Transfer Transformer) family.
It has been fine-tuned specifically for intent recognition on the labeled dataset provided in the data directory.
The fine-tuning process allows the model to extract structured information such as action, amount, currency, and recipient from user commands.

Model Access

The trained model is hosted on Hugging Face and can be accessed here: RayBe/t5-finetuned-final.

To use the model, you can load it using the transformers library:

from transformers import T5Tokenizer, T5ForConditionalGeneration

model_name = "RayBe/t5-finetuned-final"
tokenizer = T5Tokenizer.from_pretrained(model_name)
model = T5ForConditionalGeneration.from_pretrained(model_name)

Usage

The model is intended for real-time extraction of transaction details from natural language inputs.

Example input:

"send 5102.47 GBP to my brother."

Expected output:

{
    "action": "send",
    "amount": 5102.47,
    "currency": "GBP",
    "recipient": "my brother"
}