Cross-dialect named entity recognition improves bangla language models
Cross-Dialect NER for Bangla Regional Dialects Using Leave-One-Dialect-Out Cross-Validation and Explainable AI
Computation and Language
Summary
Different regions of the Bangla language have unique ways of speaking and writing, which makes it hard for computer programs to recognize names and places in all dialects. The authors tested various advanced language models to see how well they work across these dialects by training on some and testing on others. They found that one model, Multilingual-E5 Large, worked best but still struggled with certain dialects. They also showed the model mostly relies on the exact appearance of names rather than understanding the context around them. This work sets a baseline for future improvements in handling Bangla dialects.
What this means in practice
- •For language technology developers: Develop NER tools that perform well on multiple Bangla regional dialects without needing separate models for each dialect.
- •For local content creators: Improve content tagging and automated information extraction for Bangla texts from different regions to support better search and organization.
Authors
Shamim Rahim Refat, Faika Fairuj Preotee, Shuvashis Sarker, Shifat Islam, Bidyarthi Paul, Mohammad Ashraful Hoque
Abstract
Bangla, the seventh most spoken language in the world, exhibits significant regional dialectal diversity, with dialects such as Barishal, Chattogram, Sylhet, Noakhali, and Mymensingh differing in lexical, morphological, and syntactic characteristics. These variations pose substantial challenges for Named Entity Recognition (NER), limiting the generalization of models trained on Standard Bangla or a single regional dialect. This paper presents a cross-dialect Bangla NER framework using the publicly available ANCHOLIK-NER dataset, comprising 17,405 annotated sentences and 101,817 tokens across five major Bangla regional dialects. A Leave-One-Dialect-Out Cross-Validation (LODOCV) strategy is adopted, training models on four dialects and evaluating on the remaining unseen dialect. Eight pretrained transformer-based models, including BanglaBERT, MuRIL, XLM-RoBERTa, and Multilingual-E5, are evaluated under identical experimental settings. Multilingual-E5 Large achieves the highest F1-score in every fold, peaking at 97.26% on Mymensingh and reaching its lowest, 82.38%, on Chattogram, the most challenging target dialect. To improve interpretability, Local Interpretable Model-agnostic Explanations (LIME) are applied to word-level predictions, revealing that the model's decisions are driven primarily by the surface form of the target entity word itself rather than by surrounding sentence context. These findings establish a benchmark for cross-dialect Bangla NER and demonstrate the effectiveness of transformer-based transfer learning for low-resource regional dialects, while highlighting the surface-form dependence of current models as a direction for future work.