How Effectively Can BERT Models Interpret Context and Detect Bengali Communal Violent Text?

Abdullah Khondoker; Enam Ahmed Taufik; Farig Sadeque; Md. Iftekhar Islam Tashik; S M Ishtiak Mahmud

REVIEW 2 major objections 7 minor 36 references

How Effectively Can BERT Models Interpret Context and Detect Bengali Communal Violent Text?

T0 review · 2 major / 7 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper claims that an ensemble of fine-tuned BanglaBERT variants detects Bengali communal violent text at macro F1 0.63, up from a 0.36 baseline.

desk verdict A useful, honest low-resource hate-speech paper whose headline F1 gains are probably inflated by model-based test-set selection; worth reviewing, but the authors need to re-evaluate on a random split. read the letter →

arxiv 2506.19831 v1 pith:MJF64AF2 submitted 2025-06-24 cs.CL

Abdullah Khondoker , Enam Ahmed Taufik , Md. Iftekhar Islam Tashik , S M Ishtiak Mahmud , Farig Sadeque This is my paper · ORCID

classification cs.CL

keywords communalviolencedetectionBengaliNLPBanglaBERThatespeechclassificationdataaugmentationensemblelearningLIMEinterpretabilityexplainableAI

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This study claims that a Bengali-specific BERT model, and more strongly an ensemble of five such models, can classify social-media comments that incite communal violence, raising macro F1 from 0.36 in the baseline to 0.60 for the best single model and 0.63 for the ensemble. The improvement depends on adding 1,794 manually annotated comments to the original 12,791-comment dataset, mostly targeting the Ethno and Nondenominational communal classes that had almost no training examples. The paper also establishes that the remaining errors are systematic: neutral words such as 'human' or 'Bangladesh' lie close to religious trigger words in BanglaBERT's embedding space, and LIME analysis shows that a small set of religious terms drives most Religio-communal misclassifications. If these findings hold, a fine-tuned Bengali BERT ensemble is a practical first-line detector for communal hate, with the caveat that annotation noise and context-dependent ambiguity still cap its reliability.

What carries the argument

The machinery is a fine-tuned BanglaBERT classifier, a transformer language model pre-trained specifically for Bengali, combined at the end into an ensemble of five fine-tuned models whose predictions are merged by mean averaging, majority voting, or an MLP. The data-side mechanism is manual augmentation: after SMOTE, zero-shot, few-shot, and paraphrasing all failed to produce usable minority-class examples, the paper used a fine-tuned BanglaBERT to pre-screen roughly 300,000 collected comments, then manually annotated the model-positive ones, adding 1,794 new instances and rebalancing the four classes. The diagnostic mechanism is the pairing of cosine similarity between class-word embeddings and LIME explanations, which localises the persistent Religio-communal versus Noncommunal confusion to lexical overlap in the embedding space.

What would settle it

Collect a random sample from the same 300,000-comment pool, annotate it with humans only, with no classifier pre-screening, train the identical BanglaBERT and ensemble on that data, and evaluate on a fixed held-out test set; if macro F1 does not reproduce the 0.60 to 0.63 range, the reported gain was an artifact of model-biased data selection.

Watch

Extended reading notes

Core claim

The central claim is that four-way classification of Bengali communal violent text can be taken from a baseline macro F1 of 0.36 to 0.60 by fine-tuning BanglaBERT on an augmented dataset, and to 0.63 by ensembling four fine-tuned BanglaBERT variants with a fifth BanglaBERT model and combining their predictions by majority voting. The gains are concentrated in the minority classes: Nondenominational communal F1 rises from 0.16 to 0.69 and Ethno communal from 0.32 to 0.66 after adding the 1,794 manually annotated instances, while BanglaBERT Large and mBERT both underperform the standard model. The paper further claims that the main error pattern is not random: cosine similarity between the embeddings of Noncommunal words like 'human' or 'Bangladesh' and Religio-communal words like 'infidel', 'atheist', 'Allah', or 'Muslim' is very high in BanglaBERT and even higher in BanglaBERT Large, so the models blur the boundary between neutral and religiously charged text. LIME explanations identify a small set of religious trigger words as the dominant cause of Religio versus Noncommunal mistakes, and a manual review of 300 errors attributes 18% of them to annotation problems in the dataset itself.

Load-bearing premise

The manual augmentation step used a fine-tuned BanglaBERT to pre-screen which comments were worth annotating, so the added training data may carry the model's own biases and inflate the reported F1 improvement.

Editorial extensions

If this is right

With enough manually added examples, minority classes such as Nondenominational and Ethno communal become learnable, while larger or multilingual models do not automatically help: BanglaBERT Large and mBERT both scored lower than standard BanglaBERT.
Combining several fine-tuned BERT variants pushes macro F1 from 0.60 to 0.63 and specifically reduces the Religio-communal versus Noncommunal confusion, with majority voting slightly outperforming averaging on minority classes.
The embedding-similarity result implies that any model using BanglaBERT's representations will tend to see neutral text containing religious words as communally violent, so surface lexical cues alone are not sufficient for this task.
LIME's finding that 65% of Religio-communal misclassifications contain a small set of religious terms suggests that a simple lexicon-based baseline would reproduce the same confusion, and that context-sensitive disambiguation is the real bottleneck.
Annotator noise found in 18% of the reviewed errors means that part of the measured ceiling is label error rather than model failure, so cleaner annotations should yield further gains independent of architecture.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Because the 1,794 augmented comments were pre-selected by the same BanglaBERT flavour that is later evaluated, a portion of the measured F1 gain could reflect the model confirming its own preferences rather than genuinely learning to generalise; an independent human-only annotation of a random candidate sample would settle this.
The cosine-similarity result points to a specific cheap intervention worth testing: adding contrastive examples in which religious words co-occur with Noncommunal labels, or debiasing the embedding space, could reduce the dominant confusion more directly than adding more data.
The methodology should transfer to other low-resource languages with communal-violence dynamics, but the trigger-word sets and embedding geometry would have to be re-derived per language; the reported F1 numbers are unlikely to carry over unchanged.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Colleague,

Worth a look if you work on low-resource hate speech or on leakage in data augmentation. The paper fine-tunes BanglaBERT variants on a Bengali communal-violence dataset, adds 1,794 manually annotated examples, and reports macro F1 0.60 for a single model and 0.63 for an ensemble. The error analysis is the most valuable part: the authors quantify annotation problems (18% of 300 reviewed misclassifications are bad labels), use cosine similarity to show how BanglaBERT embeddings conflate religious and non-communal words, and use LIME to trace specific misclassification patterns. That kind of self-diagnosis is rare and useful. They also tried SMOTE, zero/few-shot, and paraphrase augmentation, and reported the failures rather than hiding them.

The soft spot is real and probably load-bearing. In Section 4.1, they used a fine-tuned BanglaBERT to screen 300k comments, then manually annotated only the flagged ones. The 80/20 train/test split was done after this augmentation, so the test set contains model-selected comments. The minority-class examples in the test set are exactly the kind a BanglaBERT-like model already thinks look communal. Since the ensemble is built from BanglaBERT variants, part of the reported gain may be agreement with the pre-screener rather than generalization. The labels themselves are human, so this is selection bias rather than label circularity, but it still shifts the test distribution. Their own numbers support caution: Nondenominational F1 jumps from 0.16 to 0.69 after this augmentation; some of that is real (more data), but some is likely due to test-set composition. There are also no error bars or significance tests; the 0.60-to-0.63 gap could easily be noise. And the data and code are not released, which limits reuse.

The paper is honest about its limitations and the writing is clear. The abstract says "improved performance," not "breakthrough," and the task is socially relevant. The general claim that fine-tuning Bengali BERT helps probably survives; the exact F1 numbers shouldn't be taken at face value.

My recommendation: send to peer review, but require a revision. Ask the authors to report results on a test split that excludes the model-screened additions, or at least evaluate on the original 12,791-instance split. If that is impossible, the headline F1 should be framed as a proof-of-concept under a shifted distribution, not a clean estimate. Confidence intervals or a significance test would also help. I would not cite the 0.63 number as-is, but I would cite the augmentation pitfall and the error analysis.

Referee Report

2 major / 7 minor

Summary. The paper addresses detection of Bengali communal violent text on social media. The authors fine-tune BanglaBERT, BanglaBERT Large, and mBERT on a four-class dataset derived from Tasnim et al. [33], expand the minority classes with 1,794 manually annotated comments (pre-screened by a fine-tuned BanglaBERT), and combine five fine-tuned models into ensembles using mean value, voting, and MLP fusion. The headline result is a macro F1 of 0.63 for the voting ensemble, up from a baseline of 0.35–0.36. The paper also presents error analysis, including annotation anomaly review, cosine similarity analysis of word embeddings, and LIME-based interpretability.

Significance. If the quantitative claims were reliable, the paper would be a useful contribution to an under-resourced and socially important task: it provides a new Bengali communal-violence dataset expansion, benchmarks several BERT variants, honestly reports failures of SMOTE/zero-shot/few-shot augmentation, and includes qualitative interpretability analysis. The error-analysis disclosure (18% problematic annotations) and the systematic comparison of augmentation strategies are commendable. However, the central performance claim rests on an evaluation protocol that is contaminated by the model-assisted data selection, so the significance of the reported F1 improvement is currently not established.

major comments (2)

[Section 4.1, final paragraph] The manual augmentation pipeline uses a fine-tuned BanglaBERT to pre-screen 300,000 comments and only comments predicted as communally violent proceed to manual annotation. The 1,794 added entries are then mixed into the dataset before the 80/20 train/test split, so the test set's minority-class examples are predominantly comments that a BanglaBERT-like model already judged to be violent. Because the final ensemble is composed of BanglaBERT variants, evaluation on this test set measures agreement with the pre-screener's decision boundary rather than generalization to the true distribution of Bengali social media comments. This is a covariate-shift problem, not just label noise. The authors should re-evaluate on a test set drawn from the original 12,791 instances only, and/or on a randomly sampled (unfiltered) subset of the newly collected comments; the difference in macro F1 should be reported.
[Section 6.1 and Section 5] The error analysis reports that 18% of 300 reviewed misclassifications have problematic annotations, and Section 4.1 does not describe any re-annotation of the new test entries beyond the blind voting process. In addition, no confidence intervals, standard deviations, or significance tests are provided for any F1 score in Tables 6 or 7. The ensemble's improvement over the best single model is 0.03 (0.63 vs 0.60), and the difference between mean value and voting is at most 0.01; given the measured label noise and the absence of uncertainty quantification, this improvement is within plausible noise. The authors should provide bootstrap confidence intervals or multiple-seed results, and ideally re-annotate a held-out test set to measure the effect of label noise.

minor comments (7)

[Section 4.1, class counts] Section 4.1 states that 1,794 new entries were added, but the listed class counts sum to 1,881 (1,073 + 508 + 300); please reconcile.
[Section 8 vs Table 6] The Conclusion states the macro F1 improved from 0.36, while Table 6 reports 0.35 for the 'BanglaBert Baseline [No Class Weight]' row; please harmonize.
[Table 6] The relationship between 'BanglaBert Baseline [No Class Weight]' and 'BanglaBert Baseline' is unclear; state explicitly which rows use class weights.
[Section 5.1] Section 5.1 says 'F1 scores improved for all classes except Religio communal' after adding paraphrased data, but Table 6 shows Noncommunal staying at 0.44; correct the wording.
[Section 4.2] Removing stop words is generally not recommended for transformer models such as BERT; justify this preprocessing choice or remove it.
[Table 10] Table 10 reports cosine similarities but does not specify how the word vectors were extracted (which layer and pooling method); add this detail for reproducibility.
[Section 6.2.1] The claim that 'the model perceives them as being contextually similar' based on static cosine similarity of word embeddings should be softened, since such similarities do not directly explain the model's contextual predictions; a contextualized embedding analysis would be more appropriate.

Circularity Check

1 steps flagged · score 6.0 of 10

Model-assisted augmentation contaminates the test split: the 0.63 macro-F1 gain is partly self-agreement with the BanglaBERT pre-screener used in Section 4.1, not an independent generalization estimate.

fitted input called prediction [Section 4.1 "Data augmentation"; Section 4.2 "Data preprocessing"; Section 5.2 "Ensemble model"]
"Finally, using a fine-tuned BanglaBERT model, we classified these large amounts of comments as communally violent or not. This classification process was followed by manual annotations through a blind voting system. ... Through this process, we added 1,794 new entries, including 1,073 to the Nondenominational communal class, 508 to the Ethno communal class, and 300 to the Religio communal class. ... Initially, we divided our dataset into 80% train and 20% test data. ... In our research, we used an ensemble model that consisted of five fine-tuned models."

The pipeline makes the evaluation set depend on the model being evaluated. Section 4.1 uses a fine-tuned BanglaBERT to decide which of 300,000 comments are 'communally violent or not'; only pre-screener-positive comments reach manual annotation, so the 1,794 added examples are drawn from one side of BanglaBERT's decision boundary. Section 4.2 then splits the augmented dataset 80/20, putting those model-selected examples into the test set used for the reported F1. Section 5.2's ensemble is four BanglaBERT variations plus a fifth BERT variant, i.e., the same model family as the pre-screener. The 0.36-to-0.63 gain therefore partly measures agreement with the pre-screener's own boundary, not generalization to a random sample of real social-media comments.

full rationale

The only load-bearing circularity I can exhibit is in the data-augmentation/evaluation pipeline. The core modeling recipe—fine-tuning BanglaBERT and other BERT variants on an expert-annotated Bengali corpus—is otherwise self-contained and standard; the dataset citation [33] is provenance, not a self-referential derivation, and no uniqueness theorem or ansatz is smuggled in via citation. However, the central empirical claim (macro F1 0.63, up from 0.36) is not measured on a test set independent of the model family being evaluated. Because a fine-tuned BanglaBERT pre-screened the candidate pool for the 1,794 augmented instances, and because the 80/20 split occurred after augmentation, the test set contains examples selected by the same kind of model whose F1 is reported. This is a selection-bias/self-training circularity: it does not make the score mathematically determined, but it invalidates the claim that the improvement is a clean out-of-sample result. The paper's own error analysis (18% problematic annotations among 300 reviewed misclassifications) further shows the labels are noisy, though that is a correctness issue, not circularity. No significance tests or confidence intervals are reported, so the 0.03 ensemble-over-single-model gap may also be noise. Because the central result still has independent components (original 13k dataset, human labels, comparisons across architectures), I score this as partial circularity rather than complete self-definition.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

The paper introduces no new theoretical entities. Its central empirical claims rest on the training data and the choice of model hyperparameters.

free parameters (4)

Learning rate = 2e-5
Set via Bayesian optimization; standard for BERT fine-tuning.
Batch size = 32
Selected via Bayesian optimization.
Class weights = Inverse frequency
Applied to minority classes to address imbalance.
Early stopping patience = 2 epochs
Used to prevent overfitting.

assumptions (3)

domain assumption The manually augmented labels are correct.
Section 4.1 describes blind voting annotation but also reports 18% of reviewed errors are due to incorrect annotations, so this assumption is partially violated.
domain assumption The dataset from [33] is a reliable ground truth for communal violence.
The paper builds directly on the annotation framework of their own prior work without independent validation.
domain assumption Macro F1 is an appropriate evaluation metric.
This is a reasonable choice for imbalanced classes, but the paper does not justify it.

how reviews work

0 comments

Cite this review

Pith. "Pith review of How Effectively Can BERT Models Interpret Context and Detect Bengali Communal Violent Text?." pith.science (2026). https://pith.science/paper/MJF64AF2

@misc{pith2026250619831,
  author       = {Pith},
  title        = {Pith review of: How Effectively Can BERT Models Interpret Context and Detect Bengali Communal Violent Text?},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MJF64AF2}},
  note         = {Machine review of arXiv:2506.19831}
}

read the original abstract

The spread of cyber hatred has led to communal violence, fueling aggression and conflicts between various religious, ethnic, and social groups, posing a significant threat to social harmony. Despite its critical importance, the classification of communal violent text remains an underexplored area in existing research. This study aims to enhance the accuracy of detecting text that incites communal violence, focusing specifically on Bengali textual data sourced from social media platforms. We introduce a fine-tuned BanglaBERT model tailored for this task, achieving a macro F1 score of 0.60. To address the issue of data imbalance, our dataset was expanded by adding 1,794 instances, which facilitated the development and evaluation of a fine-tuned ensemble model. This ensemble model demonstrated an improved performance, achieving a macro F1 score of 0.63, thus highlighting its effectiveness in this domain. In addition to quantitative performance metrics, qualitative analysis revealed instances where the models struggled with context understanding, leading to occasional misclassifications, even when predictions were made with high confidence. Through analyzing the cosine similarity between words, we identified certain limitations in the pre-trained BanglaBERT models, particularly in their ability to distinguish between closely related communal and non-communal terms. To further interpret the model's decisions, we applied LIME, which helped to uncover specific areas where the model struggled in understanding context, contributing to errors in classification. These findings highlight the promise of NLP and interpretability tools in reducing online communal violence. Our work contributes to the growing body of research in communal violence detection and offers a foundation for future studies aiming to refine these techniques for better accuracy and societal impact.

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 28 canonical work pages

[33]

Research progress of zero- shot learning

Sun, X., Gu, J., Sun, H., 2020. Research progress of zero- shot learning. Applied Intelligence 51, 36003614. doi:10.1007/ s10489-020-02075-7

work page 2020
[1]

New bounds on the accuracy of majority voting for multiclass classiﬁcation

Aeeneh, S., Zlatanov, N., Yu, J., 2024. New bounds on the accuracy of majority voting for multiclass classiﬁcation. IEEE Transactions on Neural Networks and Learning Systems , 1–15doi:10.1109/TNNLS. 2024.3387544

arXiv 2024
[2]

Emoji dictionary

Agrawal, D., 2020. Emoji dictionary. URL: https://www.kaggle.com/ divyansh22/emoji-dictionary-1

work page 2020
[3]

BanglaParaphrase: A high-quality Bangla paraphrase dataset, Associ- ation for Computational Linguistics, Online only

Akil, A., Sultana, N., Bhattacharjee, A., Shahriyar, R., 2022. BanglaParaphrase: A high-quality Bangla paraphrase dataset, Associ- ation for Computational Linguistics, Online only. pp. 261–272. URL: https://aclanthology.org/2022.aacl-short.33

work page 2022
[4]

Hundreds protest in bangladesh over deadly religious violence

Al Jazeera, 2021a. Hundreds protest in bangladesh over deadly religious violence. URL: https://aje.io/7ndrbq. retrieved January 23, 2024

work page 2024
[5]

Two hindu men killed, temples vandalised in bangladesh violence

Al Jazeera, 2021b. Two hindu men killed, temples vandalised in bangladesh violence. URL: https://aje.io/uy9gmz. retrieved Febru- ary 18, 2024

work page 2024
[6]

Robust hate speech de- tection in social media: A cross -dataset empirical evaluation

Antypas, D., Camacho -Collados, J., 2023. Robust hate speech de- tection in social media: A cross -dataset empirical evaluation. URL: https://arxiv.org/abs/2307.01680, arXiv:2307.01680

arXiv 2023
[7]

An- alyzing abusive bangla comments on social media: Nlp & explainable ai, pp

Ashraf, K., Hosen, M.H., Asgar, S., Islam, M.T., Nawar, S., 2024. An- alyzing abusive bangla comments on social media: Nlp & explainable ai, pp. 1–6. doi:10.1109/iCACCESS61735.2024.10499547

work page arXiv 2024

Show all 36 references

[8]

Bangladeshs forgotten crisis: Land, ethnicity, and violence in chittagong hill tracts

Bashar, I., 2011. Bangladeshs forgotten crisis: Land, ethnicity, and violence in chittagong hill tracts. Counter Terrorist Trends and Analyses 3, 1–5. URL: http://www.jstor.org/stable/26350972

2011
[9]

Interpretable multi labeled bengali toxic comments classiﬁcation using deep learning, pp

Belal, T.A., Shahariar, G.M., Kabir, M.H., 2023. Interpretable multi labeled bengali toxic comments classiﬁcation using deep learning, pp. 1–6. doi:10.1109/ECCE57851.2023.10101588

2023
[10]

Bhattacharjee, A., Hasan, T., Ahmad, W., Mubasshir, K.S., Islam, M.S., Iqbal, A., Rahman, M.S., Shahriyar, R., 2022. BanglaBERT: Language model pretraining and benchmarks for low -resource lan- guage understanding evaluation in Bangla, Association for Compu- tational Linguisti...

2022
[11]

Communal conﬂict, civil war, and the state: Complexities, connections, and the case of sudan

Brosché, J., Elfversson, E., 2012. Communal conﬂict, civil war, and the state: Complexities, connections, and the case of sudan. African Journal on Conﬂict Resolution 12, 33 –60. URL: https://www.ajol. info/index.php/ajcr/article/view/78700

2012
[12]

Smote for imbalanced classiﬁcation with python

Brownlee, J., 2021. Smote for imbalanced classiﬁcation with python. URL: https://machinelearningmastery.com/ smote-oversampling-for-imbalanced-classification/

2021
[13]

A weakly supervised Bayesian model for violence detection in social media, Asian Federation of Natural Language Processing, Nagoya, Japan

Cano Basave, A.E., He, Y., Liu, K., Zhao, J., 2013. A weakly supervised Bayesian model for violence detection in social media, Asian Federation of Natural Language Processing, Nagoya, Japan. pp. 109–117. URL: https://aclanthology.org/I13-1013

2013
[14]

Bangla hate speech detection on social media using attention -based recurrent neural network

Das, A.K., Asif, A.A., Paul, A., Hossain, M.N., 2021. Bangla hate speech detection on social media using attention -based recurrent neural network. Journal of Intelligent Systems 30, 578 –591. doi:10. 1515/jisys-2020-0060

2021
[15]

BERT: pre- training of deep bidirectional transformers for language understand- ing

Devlin, J., Chang, M., Lee, K., Toutanova, K., 2018. BERT: pre- training of deep bidirectional transformers for language understand- ing. CoRR abs/1810.04805. URL: http://arxiv.org/abs/1810.04805

2018 arXiv
[16]

Bengali Hate Speech Detection Dataset

Dey, K., Sumon, Cochez, M., Karim, M.R., 2022. Bengali Hate Speech Detection Dataset. UCI Machine Learning Repository. URL: https://doi.org/10.24432/C5PD07

2022 doi
[17]

Bengali stop word and phrase detection mecha- nism

Haque, R.U., Mridha, M.F., Hamid, M.A., Abdullah -Al-Wadud, M., Islam, M.S., 2020. Bengali stop word and phrase detection mecha- nism. Arabian Journal for Science and Engineering 45, 33553368. doi:10.1007/s13369-020-04388-8

2020 doi
[18]

Few -shot learning

IBM, 2024. Few -shot learning. URL: https://www.ibm.com/topics/ few-shot-learning. retrieved January 29, 2024

2024
[19]

Classiﬁcation benchmarks for under -resourced bengali language based on multichannel convolutional-lstm network

Karim, M.R., Chakravarthi, B.R., McCrae, J.P., Cochez, M., 2020. Classiﬁcation benchmarks for under -resourced bengali language based on multichannel convolutional-lstm network. URL: https: //arxiv.org/abs/2004.07807 , arXiv:2004.07807

2020 arXiv
[20]

Digital 2023: Bangladesh

Kemp, S., 2023. Digital 2023: Bangladesh. URL: https:// datareportal.com/reports/digital-2023-bangladesh

2023
[21]

G-bert: An eﬃcient method for identifying hate speech in bengali texts on social media

Keya, A.J., Kabir, M.M., Shammey, N.J., Mridha, M.F., Islam, M.R., Watanobe, Y., 2023. G-bert: An eﬃcient method for identifying hate speech in bengali texts on social media. IEEE Access 11, 79697– 79709. doi:10.1109/ACCESS.2023.3299021

2023
[22]

Khang Hsien, Y., Arabee Abdul Salam, Z., Kasinathan, V., 2022. Cyber bullying detection using natural language processing (nlp) and text analytics, in: 2022 IEEE International Conference on Distributed Computing and Electrical Circuits and Electronics (ICDCECE), pp. 1–4. doi:1...

2022
[23]

Smote: Practical consideration and limitations

Kim, M., 2023. Smote: Practical consideration and limitations. URL: https://medium.com/@minjukim023/ smote-practical-consideration-limitations-f0d926b661a8

2023
[24]

ComMA@ICON: Multi- lingual gender biased and communal language identiﬁcation task at ICON-2021, NLP Association of India (NLPAI), NIT Silchar

Kumar, R., Ratan, S., Singh, S., Nandi, E., Devi, L.N., Bhagat, A., Dawer, Y., Lahiri, B., Bansal, A., 2021. ComMA@ICON: Multi- lingual gender biased and communal language identiﬁcation task at ICON-2021, NLP Association of India (NLPAI), NIT Silchar. pp. 1–

2021
[25]

URL: https://aclanthology.org/2021.icon-multigen.1

2021
[26]

Noorana, M., 2018. Pre electoral violence in 2001 parliamentary elec- tions in bangladesh: Empirical evidence from bagerhat and narayan- ganj districts, social science review: The dhaka university stud- ies 35, 231–246. URL: https://www.researchgate.net/publication/ 353636999

2018
[27]

Deep learning algorithms for cyber -bulling detection in social media platforms

Obaida, M.H., Elkaﬀas, S.M., Guirguis, S.K., 2024. Deep learning algorithms for cyber -bulling detection in social media platforms. IEEE Access 12, 76901–76908. doi:10.1109/ACCESS.2024.3406595

2024
[28]

Zero-shot learning (zsl) explained

Polat, G., 2023. Zero-shot learning (zsl) explained. URL: https: //encord.com/blog/zero -shot-learning-explained/

2023
[29]

Myanmars religious and ethnic conﬂicts: No end in sight

ReliefWeb, 2017. Myanmars religious and ethnic conﬂicts: No end in sight. URL: https://reliefweb.int/report/myanmar/ myanmars-religious-and-ethnic-conflicts-no-end-sight. retrieved January 17, 2024

2017
[30]

Hate speech detection in the bengali language: A dataset and its baseline evaluation

Romim, N., Ahmed, M., Talukder, H., Islam, M.S., 2020. Hate speech detection in the bengali language: A dataset and its baseline evaluation. arXiv preprint arXiv:2012.09686 URL: https://arxiv. org/abs/2012.09686

2020 arXiv
[31]

The Communal Conﬂict In India Causes And Peaceful Solutions

Sahai, P., 2020. The Communal Conﬂict In India Causes And Peaceful Solutions. Rajiv Gandhi Institute for Contemporary Stud- ies, New Delhi. URL: https://www.rgics.org/wp-content/uploads/ Communal-Conflict-in-India.pdf

2020
[32]

Changes in social and religious prac- tices of disputing communities after riot: A case study on communal violence in ramu, in: Journal of Philosophy, Culture and Religion, p

Soruar, W., Uddin, M.M., 2021. Changes in social and religious prac- tices of disputing communities after riot: A case study on communal violence in ramu, in: Journal of Philosophy, Culture and Religion, p. 13. doi:10.7176/JPCR/51-03

2021 doi
[34]

Mapping violence: Developing an extensive framework to build a bangla sectarian expression dataset from so- cial media interactions

Tasnim, N., Gupta, S.S., Juee, F.I., Tahsin, A., Ghum, P., Fatema, K., Haque, M., Farzana, W., Nasir, P., KhudaBukhsh, A., Sadeque, F., Sushmit, A., 2024. Mapping violence: Developing an extensive framework to build a bangla sectarian expression dataset from so- cial media int...

2024 doi
[35]

Automated bengali abusive text clas- siﬁcation: Using deep learning techniques, pp

Titli, S.R., Paul, S., 2023. Automated bengali abusive text clas- siﬁcation: Using deep learning techniques, pp. 1 –6. doi:10.1109/ ICAECIS58353.2023.10170294

2023
[36]

Ensemble MLP Classiﬁer Design

Windeatt, T., 2008. Ensemble MLP Classiﬁer Design. Springer Berlin Heidelberg, Berlin, Heidelberg. pp. 133 –147. doi:10.1007/ 978-3-540-79474-5_6

2008

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.

[1] [33]

Research progress of zero- shot learning

Sun, X., Gu, J., Sun, H., 2020. Research progress of zero- shot learning. Applied Intelligence 51, 36003614. doi:10.1007/ s10489-020-02075-7

work page 2020

[2] [1]

New bounds on the accuracy of majority voting for multiclass classiﬁcation

Aeeneh, S., Zlatanov, N., Yu, J., 2024. New bounds on the accuracy of majority voting for multiclass classiﬁcation. IEEE Transactions on Neural Networks and Learning Systems , 1–15doi:10.1109/TNNLS. 2024.3387544

arXiv 2024

[3] [2]

Emoji dictionary

Agrawal, D., 2020. Emoji dictionary. URL: https://www.kaggle.com/ divyansh22/emoji-dictionary-1

work page 2020

[4] [3]

BanglaParaphrase: A high-quality Bangla paraphrase dataset, Associ- ation for Computational Linguistics, Online only

Akil, A., Sultana, N., Bhattacharjee, A., Shahriyar, R., 2022. BanglaParaphrase: A high-quality Bangla paraphrase dataset, Associ- ation for Computational Linguistics, Online only. pp. 261–272. URL: https://aclanthology.org/2022.aacl-short.33

work page 2022

[5] [4]

Hundreds protest in bangladesh over deadly religious violence

Al Jazeera, 2021a. Hundreds protest in bangladesh over deadly religious violence. URL: https://aje.io/7ndrbq. retrieved January 23, 2024

work page 2024

[6] [5]

Two hindu men killed, temples vandalised in bangladesh violence

Al Jazeera, 2021b. Two hindu men killed, temples vandalised in bangladesh violence. URL: https://aje.io/uy9gmz. retrieved Febru- ary 18, 2024

work page 2024

[7] [6]

Robust hate speech de- tection in social media: A cross -dataset empirical evaluation

Antypas, D., Camacho -Collados, J., 2023. Robust hate speech de- tection in social media: A cross -dataset empirical evaluation. URL: https://arxiv.org/abs/2307.01680, arXiv:2307.01680

arXiv 2023

[8] [7]

An- alyzing abusive bangla comments on social media: Nlp & explainable ai, pp

Ashraf, K., Hosen, M.H., Asgar, S., Islam, M.T., Nawar, S., 2024. An- alyzing abusive bangla comments on social media: Nlp & explainable ai, pp. 1–6. doi:10.1109/iCACCESS61735.2024.10499547

work page arXiv 2024