resume-summerization-classification/bangla_resume_summarizer_banglat5

Bangla Resume Summarizer — BanglaT5

Fine-tuned BanglaT5 for abstractive summarization of Bangla-language resumes, section by section. Developed as part of the Onneshon thesis project at Bangladesh University of Professionals (BUP).

Given the raw sentences of a resume section (Objective, Experience, Education, or Skill), the model generates a fluent 2–3 sentence Bangla summary.

---

Model Details

Property	Value
Base model	csebuetnlp/banglat5
Architecture	T5ForConditionalGeneration (encoder-decoder)
Parameters	12 encoder layers, 12 decoder layers, 768 hidden dim
Language	Bengali (bn)
Task	Abstractive summarization (per section)
Training data	317 resume-summary pairs (80% of Onneshon dataset)
Validation data	79 resume-summary pairs (20% of Onneshon dataset)
Epochs	5 (with early stopping, patience=2)
Learning rate	5e-4
Batch size	4
Max input tokens	512
Max output tokens	128

---

Dataset

Trained on Onneshon — an original Bangla resume dataset of 100 annotated resumes spanning 20+ professions.
Published on Mendeley Data: DOI: 10.17632/4md7bx6fd7.1

Reference summaries were generated using GPT-OSS-120B (via OpenRouter) and human-verified, producing 396 section-level summary pairs across 4 categories: Objective, Experience, Education, Skill.

---

Evaluation Results

Evaluated against human-verified abstractive reference summaries using a Bengali-aware ROUGE tokenizer and semantic similarity:

Metric	Score
ROUGE-1	0.6198
ROUGE-2	0.4216
ROUGE-L	0.5189
Semantic Similarity (paraphrase-multilingual-MiniLM-L12-v2)	0.8079

Note on ROUGE scores: ROUGE measures exact word overlap. Since references are abstractive paraphrases and Bangla has rich morphology, ROUGE can still underestimate quality relative to semantic similarity. Semantic similarity of 0.808 confirms strong meaning alignment with human-verified references.

Per-section ROUGE-1:

Section	ROUGE-1	Semantic Sim
Objective	0.6460	0.8483
Experience	0.5040	0.7305
Education	0.7859	0.8716
Skill	0.5380	0.7788

---

Usage

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

tokenizer = AutoTokenizer.from_pretrained("your-username/bangla-resume-summarizer-banglat5", use_fast=False)
model     = AutoModelForSeq2SeqLM.from_pretrained("your-username/bangla-resume-summarizer-banglat5")
model.eval()

def summarize(text):
    inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=512)
    outputs = model.generate(
        inputs["input_ids"],
        max_length=128,
        min_length=20,
        num_beams=4,
        length_penalty=1.2,
        repetition_penalty=2.0,
        no_repeat_ngram_size=3,
        early_stopping=True
    )
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

# Example — Experience section sentences joined by ' । '
text = "সিনিয়র সফ্টওয়্যার ইঞ্জিনিয়ার । টেকনোলজি সলিউশনস লিমিটেড । জুলাই ২০২০ - বর্তমান । মাইক্রোসার্ভিস আর্কিটেকচার ব্যবহার করে স্কেলেবল ব্যাকএন্ড সিস্টেম ডিজাইন ও উন্নয়ন"
print(summarize(text))

---

Input Format

Feed one section at a time. Join multiple sentences with । (Bengali danda):

sentence_1 । sentence_2 । sentence_3

The model works best when given 2–6 sentences per section. For Skill sections, list skills separated by ।.

---

Limitations

• Trained on only 317 pairs — a small dataset by deep learning standards. Outputs may be generic for unusual professions.
• Optimized for resume text. Performance on other Bangla document types is untested.
• Not suitable for very long inputs (>512 tokens); truncation will occur.
• Compared to mT5, this model has a marginally higher overall ROUGE-1 (0.620 vs 0.617) but slightly lower semantic similarity (0.808 vs 0.818) — its edge is clearest on the Experience section (0.504 vs 0.489); on Skill, mT5 is actually marginally ahead (0.548 vs 0.538).

---

Citation

If you use this model, please cite the Onneshon dataset:

@misc{onneshon2026,
  title   = {Onneshon: A Bangla Resume NLP Dataset},
  author  = {Tanvir and Shruti Khisa and Shaira Akther Diba and Fazli Rabbi Noor},
  year    = {2026},
  doi     = {10.17632/4md7bx6fd7.1},
  publisher = {Mendeley Data}
}

---

Project

Part of the Onneshon thesis project — a Bangla NLP pipeline for resume processing including section classification, extractive summarization (TF-IDF & TextRank), abstractive summarization (BanglaT5 & mT5), and AI vs human writing detection (BanglaBERT).

• Dataset: Mendeley Data DOI: 10.17632/4md7bx6fd7.1
• Institution: Bangladesh University of Professionals (BUP), Dhaka, Bangladesh
• Supervisor: Rumana Yasmin, Lecturer, CSE Department, BUP