boltuix / bert-trip-classification

🚗 BERT Trip Classification (RoBERTa-based)

A fine-tuned roberta-base model for trip-related multi-class text classification .
This model predicts the most likely location or service a user is referring to in a sentence, based on intent and context.

🧠 Model Details

• Architecture: RoBERTa (base)
• Task: Text Classification
• Labels: 89 categories (e.g., pharmacy, hotel, airport, etc.)
• Languages: English ( en )
• Base model: roberta-base
• Fine-tuned by: boltuix

📊 Evaluation Metrics

Note: Scores based on evaluation with a custom validation set.

📁 Dataset

Custom dataset containing user queries labeled with categories like:

• pharmacy
• restaurant
• bus station
• hotel
• gym
• ... and 80+ more classes

Each row in the training data contains:

• sentence : user query
• category : true label

🔧 How to Use

from transformers import pipeline  # 🤗 Load the Hugging Face Transformers pipeline

# 🧭 Load the custom RoBERTa model for trip-related location/service classification
classifier = pipeline("text-classification", model="boltuix/bert-trip-classification")

# 🗣️ User wants to work out — let's classify the intent!
result = classifier("i wanna to work out")  # 💪 Should predict something like "gym"

print(result)  # 🖨️ Output: [{'label': 'gym', 'score': 0.9999}]

🗂️ Categories with Emoji