CoolFace
Modelpublic

salimalsazu/smart-category-detector-v1

sourceHugging Facemitupdated 7mo agoView on Hugging Face
0likes14downloads
Model Card

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

smart-category-detector-v1

Lightweight Transformer-based classifier for short text categorization (60 categories)

smart-category-detector-v1 is a multi-class text classification model designed to predict one of 60 categories from short text inputs such as:

  • —product titles
  • —marketplace listings
  • —event announcements
  • —restaurant menu items with prices

The model is optimized for fast inference and practical categorization tasks.

Model Details

Field Value ------------------ --------------------------------- Developer Salim Al Sazu Hugging Face salimalsazu Model Type DistilBERT Fine-tuned Base Model distilbert-base-uncased Task Multi-class Text Classification Categories 60 Training Samples \~600,000 Language English License MIT

Training Dataset

The model was trained on a curated dataset of approximately 600,000 samples containing short text entries mapped to one of 60 categories.

Dataset Structure

Column Description ---------- ----------------------------- text Short input text category Target classification label

Example:

Samsung Galaxy S24 Ultra 512GB Mobile -\> smartphones\ Dhaka International Book Fair 2026 -\> book_fair\ Jamboo Burger Tk 220 -\> burgers\ Lenovo ThinkPad X1 Carbon Laptop -\> laptops\ Chicken Biryani Tk 180 -\> biryani

Dataset Distribution

The dataset is divided into three main groups, each containing 20 categories.

Event Categories (20) --- \~200,000 samples

sports\ music\ techconference\ educationseminar\ businesssummit\ startuppitch\ jobfair\ artexhibition\ culturalfestival\ religiousevent\ politicalrally\ charityevent\ workshop\ webinar\ networkingevent\ bookfair\ foodfestival\ fashionshow\ award_ceremony\ hackathon

Product Categories (20) --- \~200,000 samples

electronics\ smartphones\ laptops\ fashionclothing\ shoes\ beautycosmetics\ groceryfood\ furniture\ homeappliances\ kitchenitems\ sportsequipment\ books\ toys\ babyproducts\ healthsupplements\ automotive\ gaming\ jewelry\ officesupplies\ petproducts

Restaurant / Menu Categories (20) --- \~200,000 samples

burgers\ pizza\ sandwichwraps\ friessides\ friedsnacks\ streetfood\ biryani\ ricedishes\ noodlespasta\ curries\ bbqgrill\ seafooddishes\ breakfastitems\ soupssalads\ cakespastries\ icecream\ traditionalsweets\ coffeetea\ softdrinks\ shakessmoothies

Example Predictions

Input Prediction --------------------------------------- ------------- Samsung Galaxy S24 Ultra 512GB Mobile smartphones Dhaka International Book Fair 2026 book_fair Jamboo Burger Tk 220 burgers HP Pavilion RTX 4060 Gaming Laptop laptops Beef Burger Combo Tk 350 burgers

Quick Start

python
from transformers import pipeline

clf = pipeline(
    "text-classification",
    model="salimalsazu/smart-category-detector-v1",
    top_k=5
)

print(clf("Samsung Galaxy S24 Ultra 512GB Mobile"))
print(clf("Dhaka International Book Fair 2026"))
print(clf("Jamboo Burger Tk 220"))

Limitations

  • —Works best with short text
  • —Designed primarily for English text
  • —Mixed-language inputs may reduce accuracy
  • —Limited to 60 predefined categories
  • —Unknown categories may be mapped to the closest label

Future Improvements

Potential improvements include:

  • —adding real-world marketplace data
  • —improving Bangla language support
  • —increasing spelling robustness
  • —adding an "unknown" label
  • —publishing benchmark evaluation metrics

Citation

Salim Al Sazu\ smart-category-detector-v1