A Comparative Study of Deep Learning Models for Bias Detection in Bangla Job Advertisements
Bibliographic record
Abstract
Job advertisements often contain explicit or implicit discrimination that reinforces inequality in recruitment activities. While prior research has focused mainly on English texts, no systematic attempt has been made to study bias in Bangla job postings, which play a critical role in shaping labor-market access in Bangladesh. This study proposes a deep learning model to identify and categorize prejudices in the advertisements of jobs in Bangla language. A manually annotated corpus of 1536 sentences based on 200 job circulars was prepared and it is divided into five bias categories: masculine, feminine, age, experience and religious. Six architectures were considered, namely recurrent neural networks (LSTM, BiLSTM, GRU) and transformer-based ones (BanglaBERT, mBERT, XLM-RoBERTa). It is found that the transformer-based models significantly beat recurrent models, with MBERT showing the best results as far as test accuracy of 93.51% and F 1 -macro of 93.71% and minimum Hamming loss of 0.0138. The strength of the models in all categories of bias was also further validated by class-wise analysis. The results indicate that multilingual and language-specific transformer models are effective at detecting discriminatory patterns in Bengali advertisements and offering practical applications for fair recruitment monitoring and policy enforcement.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".