dejanseo/bulgarian-search-query-intent

Bulgarian Intent Classification Model

This is a Bulgarian language search query intent classification model trained by Dan Petrovic from DEJAN AI.

Model Overview

This model is fine-tuned from mDeBERTa V3 for search query intent classification in Bulgarian. It predicts one of five intent categories:

• COMMERCIAL_INVESTIGATION – Queries with a purchase intent but requiring additional research.
• INFORMATIONAL – Queries seeking knowledge or facts.
• LOCAL – Queries related to local services or locations.
• NAVIGATIONAL – Queries aiming to reach a specific website or service.
• TRANSACTIONAL – Queries with a direct intent to complete an action (e.g., purchase, reservation).

Usage

To use this model with the Hugging Face Transformers library:

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

# Load model and tokenizer
model_name = "dejanseo/bulgarian-search-query-intent"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

# Example query
query = "Купи нов телефон онлайн"

# Tokenize and predict
inputs = tokenizer(query, return_tensors="pt")
with torch.no_grad():
    outputs = model(**inputs)

logits = outputs.logits
predicted_class_id = logits.argmax().item()
predicted_label = model.config.id2label[str(predicted_class_id)]

print(f"Predicted Intent: {predicted_label}")

Training Details

• Base Model: mDeBERTa V3
• Dataset: Proprietary Bulgarian search query dataset
• Fine-tuning: 6 epochs, batch size 6, learning rate 2e-5
• Evaluation Metric: Accuracy

Limitations

• Designed for Bulgarian language only.
• May not generalize well to very niche queries outside typical search behavior.
• Performance depends on query clarity and domain-specific phrasing.

Citation

If you use this model, please cite:

@misc{petrovic2025bulgarian,
  title={Bulgarian Intent Classification Model},
  author={Dan Petrovic, DEJAN AI},
  year={2025},
  url={https://dejan.ai}
}

Acknowledgements

• Hugging Face
• DEJAN AI
• Microsoft DeBERTa