B-NER: A Novel Bangla Named Entity Recognition Dataset With Largest Entities and Its Baseline Evaluation
Bibliographic record
Abstract
Within the Natural Language Processing (NLP) framework, Named Entity Recognition (NER) is regarded as the basis for extracting key information to understand texts in any language. As Bangla is a highly inflectional, morphologically rich, and resource-scarce language, building a balanced NER corpus with large and diverse entities is a demanding task. However, previously developed Bangla NER systems are limited to recognizing only three familiar entities: person, location, and organization. To address this significant limitation, we introduce a novel Bangla NER dataset B-NER, which was created using 22,144 manually annotated Bangla sentences collected from Bangla newspapers and Bangla Wikipedia. This dataset includes a total of 9,895 unique words which were manually categorized into eight different entity types, such as a person, organization, event, artifact, time indicator, natural phenomenon, geopolitical entity, and geographical location. Inter-annotator agreement experiments were conducted to validate the quality of annotations performed by three annotators, resulting in a Kappa score of 0.82. In this paper, we provide an outline of the annotation guideline illustrated with examples, discuss the B-NER dataset properties, and present benchmark evaluations of the dataset. To establish that B-NER is more comprehensive and balanced in comparison to other publicly accessible datasets, we conducted cross-dataset modeling and validation, i.e. trained NER model on one dataset while tested on another, and found that the model trained on B-NER performed the best in that settings. Furthermore, we performed exhaustive benchmark evaluations based on Bidirectional LSTM with fastText embeddings and sentence transformer models. Among these models, fine-tuned IndicBERT achieved noticeable results with a Macro-F1 of 86%. This dataset and baseline results will be publicly available under a CC-BY 4.0 license in the CoNLL-2002 format to facilitate further research on Bangla NER.
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 | 0.003 | 0.007 |
| Meta-epidemiology (narrow) | 0.002 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.004 | 0.003 |
| Science and technology studies | 0.002 | 0.001 |
| Scholarly communication | 0.002 | 0.004 |
| Open science | 0.003 | 0.002 |
| Research integrity | 0.003 | 0.002 |
| Insufficient payload (model declined to judge) | 0.005 | 0.007 |
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".