Pruning for Performance: Efficient Idiom and Metaphor Classification in Low-Resource Konkani Using mBERT

Harshita Poojary; Kevin Zhu; Pranav Prabhu; Pranav Saran; Sean O'Brien; Timothy Do; Vasu Sharma

REVIEW 4 major objections 6 minor 20 references

Pruning for Performance: Efficient Idiom and Metaphor Classification in Low-Resource Konkani Using mBERT

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Pruning zero-importance attention heads keeps Konkani idiom accuracy at 83% while dropping metaphor accuracy from 88% to 78%, and the paper attributes the difference to how the two tasks distribute attention across layers.

desk verdict New Konkani metaphor annotations are a real but small resource; the pruning-sensitivity conclusion rests on a four-sentence accuracy drop in a single 40-sentence test split. read the letter →

arxiv 2506.02005 v2 pith:OOZ2RNPF submitted 2025-05-24 cs.CL

Timothy Do , Pranav Saran , Harshita Poojary , Pranav Prabhu , Sean O'Brien , Vasu Sharma , Kevin Zhu This is my paper · ORCID

classification cs.CL

keywords KonkaniNLPmetaphorclassificationidiomattentionheadpruningmBERTlow-resourcelanguagesfigurativelanguageBiLSTM

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 paper tries to establish that gradient-based attention head pruning is a viable efficiency trick for Konkani figurative-language models, but only when the pruning target is chosen per task. The authors introduce a 500-sentence metaphor-annotated extension of the Konidioms corpus and fine-tune an mBERT+BiLSTM classifier on it and on idiom classification. After removing the 12 attention heads with zero gradient importance, idiom accuracy holds at 83% (up from 82%), while metaphor accuracy drops from 88% to 78% on a 200-sentence balanced subset. The paper attributes the asymmetry to attention structure: idioms rely on localized lower-layer heads, metaphors on a distributed set across layers. If correct, this gives low-resource Konkani a cheaper idiom model, a first metaphor benchmark, and a warning that one-size-fits-all pruning will not work for figurative language.

What carries the argument

The load-bearing mechanism is a gradient-based attention-head importance metric: for each of the 144 heads in mBERT, the model computes $I_h = \mathbb{E}_{(x,y)\sim D}|\partial L / \partial h(h)|$, the expected absolute gradient of the loss with respect to the head's output. Heads with $I_h = 0$ are removed post hoc (12 heads, 8.33% of Transformer parameters), while the BiLSTM is left unchanged. The argument turns on the contrast between the two tasks' importance heatmaps: idiom classification shows high scores mainly in the lower layers, while metaphor classification shows salient heads across all layers; the paper interprets this as the reason one task tolerates pruning and the other does not.

What would settle it

Re-run the same pruning experiment on the released dataset with multiple 80/20 splits and several seeds; if the pruned metaphor accuracy gap versus the original is not consistently near 10 points and often falls within a couple of sentences' errors, the claim that metaphor classification is inherently more sensitive to attention-head pruning is refuted.

Watch

Extended reading notes

Core claim

The central claim is that pruning all mBERT attention heads whose gradient importance is exactly zero, 12 of 144 heads, has task-specific consequences for Konkani figurative-language classification. For idioms, the pruned mBERT+BiLSTM matches or slightly beats the original: accuracy goes from 0.82 to 0.83, recall from 0.89 to 0.91, and F1 stays at 0.88. For metaphors, the same pruning costs ten accuracy points (0.88 to 0.78), with F1 falling from 0.86 to 0.74. The paper explains the difference structurally: idiom-relevant heads cluster in the lower layers, while metaphor-relevant heads are spread across all layers, so removing any heads disrupts a more diffuse circuit. It also claims to release the first metaphor-annotated Konkani dataset and to apply attention head pruning for the first time to any Konkani NLP task.

Load-bearing premise

The load-bearing premise is that the reported accuracy gap is real, but it rests on a single 40-sentence test set where exactly four errors separate 88% from 78%, so a different split could erase the claimed difference.

Editorial extensions

If this is right

Konkani idiom classification can be compressed by 12 attention heads with no accuracy loss: the pruned model reaches 83% accuracy and 0.88 F1, so the released pruning recipe is directly usable for idiom tasks.
Metaphor classification should not be pruned with the same zero-importance threshold: the same 12-head removal costs about 10 accuracy points, meaning pruning policies need to be task-specific.
Attention-head importance maps can act as a diagnostic: early-layer heads carry idiom signals while metaphor signals are distributed, giving a cheap way to predict whether a task will tolerate compression.
The 500-sentence metaphor-annotated subset, once released with the framework, gives future work a first Konkani benchmark for metaphor detection and a direct comparison point for other low-resource models.

Reading between the lines

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

I would not equate a zero gradient with functional redundancy: the paper's own numbers show that removing zero-score heads can still cost 10 accuracy points, so in small-data settings gradient magnitude at convergence is a weak proxy for a head's contribution.
A natural extension the paper does not run is a threshold sweep: pruning heads with small but nonzero importance values and tracking accuracy would show whether the metaphor drop is specifically about zero-score heads or about any head removal.
If the structural contrast (early-layer idioms, diffuse metaphors) replicates in other low-resource languages, attention-head importance maps could become a cheap pre-pruning diagnostic for figurative-language tasks.

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

The one thing to know: this paper contributes a genuinely new, manually verified metaphor-annotated dataset for Konkani, but its central scientific claim—that metaphor classification is more pruning-sensitive than idiom classification—is not supported by the reported evaluation. The entire signal is four sentences on a single 40-sentence test split.

What is new and good: the paper provides the first metaphor annotations for Konkani, extending the Konidioms corpus, with three native-speaker validators. The code and data are released, the limitations section is honest, and the ablation table comparing mBERT, IndicBERT, and XLM-R+BiLSTM+Attn gives useful context. Applying Michel et al.'s gradient-based head pruning to a low-resource figurative language task is a reasonable extension, and the paper is transparent that the architecture was chosen after seeing ablation results.

The load-bearing problem: Table 2's metaphor accuracy drop is 0.88 to 0.78, which on a 40-item test set is 35/40 to 31/40 correct—four examples. There are no confidence intervals, no repeated runs, no seed variance, and no cross-validation. The 200-sentence balanced subset was 'curated' from 500 without describing the selection rule, so the single split may not be representative. The limitations section candidly says 'evaluation on a single test split necessitates further validation,' but Sections 6 and 7 state the conclusion categorically. Also, the importance metric's dataset D is never specified (training set? validation set?), which matters for reproducibility. The efficiency gain is only 8.33% of heads (12 of 144), so calling the model 'efficient' oversells it.

For a reader: this is a resource/application paper, not a methodological breakthrough. The dataset, while tiny, may be useful for Konkani NLP and for studying figurative language in low-resource settings. The pruning-sensitivity conclusion should be treated as a hypothesis, not a finding. The paper deserves a serious referee because the resource contribution is real and the fragility can be addressed with variance reporting and cross-validation.

Referee Report

4 major / 6 minor

Summary. The paper introduces a metaphor-annotated Konkani dataset built on the Konidioms corpus, and proposes a hybrid mBERT+BiLSTM model for metaphor and idiom classification. It applies a gradient-based attention-head pruning method (Michel et al., 2019) to both tasks. The authors report that after removing all heads with zero gradient importance (retaining 132 of 144 heads), idiom classification accuracy is stable (0.82 to 0.83), while metaphor classification accuracy drops from 0.88 to 0.78. From these results and the attention-head importance heatmaps, the paper concludes that idiom classification relies on localized lower-layer heads, whereas metaphor classification relies on a broader, more distributed attention profile and is therefore more sensitive to pruning. The paper releases its code and dataset.

Significance. If the reported pruning-sensitivity asymmetry is real, the finding that idiom and metaphor classification have distinct attention-head dependencies would be a useful contribution to low-resource figurative-language NLP. The paper also provides a new manually annotated Konkani metaphor dataset, with labels verified by native speakers, and it publicly releases code and resources, which are concrete assets for a severely under-resourced language. The ablation study across mBERT, IndiC-BERT, and XLM-R variants is a useful engineering comparison. However, the headline scientific claim rests on a very small test set and a four-sentence accuracy difference, so the significance is currently conditional on additional statistical validation.

major comments (4)

[Section 3, Table 2] The central claim that metaphor classification is more pruning-sensitive than idiom classification rests on a single test set of 40 sentences. The reported metaphor accuracy drop from 0.88 to 0.78 corresponds to exactly 4 sentence-level errors (35/40 to 31/40 correct). No confidence intervals, seed variance, cross-validation, or repeated splits are reported, and the 200-sentence balanced subset is described only as "curated" from 500 annotations without specifying the selection procedure. Because the qualitative asymmetry could invert with a small number of borderline test sentences, the paper must provide a more robust evaluation: e.g., k-fold cross-validation, bootstrap confidence intervals, or a paired McNemar test, and a precise description of how the 200-sentence subset was chosen.
[Section 5] The gradient-based importance score I_h is defined as an expectation over a dataset D, but the paper never specifies whether D is the training set, the validation set, or the test set. This is not a minor implementation detail: using the test set to compute importance and then evaluating on the same test set would leak information, while using the training set is the more defensible choice. The paper should state the exact composition of D and justify its use.
[Section 6, Appendix D (Table 3)] The mBERT+BiLSTM architecture is selected after inspecting the ablation results in Table 3, which report both original and pruned accuracy for the chosen model and its baselines. Selecting the best-performing architecture on the same split that is later used for the headline comparison risks optimism bias and weakens the claim that the pruning effect is intrinsic to the task rather than an artifact of model selection. The authors should clarify whether the ablations were conducted on a held-out validation set and how the final test evaluation was kept separated from the selection process.
[Title, Abstract, Section 4] The paper consistently describes the work as improving "efficiency," yet no runtime, latency, memory, or parameter-count measurements are reported. The pruning removes 12 of 144 attention heads (8.33% of heads), which is a small fraction of the mBERT architecture, and the paper does not report how many parameters this actually removes or how it affects inference speed. The efficiency claim should either be backed by direct measurements or explicitly softened to "small head-count reduction" rather than general efficiency.

minor comments (6)

[Section 5.1] The sentence "The contrasting patterns observed in the two classification tasks, suggests fundamental differences..." contains a subject-verb agreement error; it should be "suggest."
[Section 6] The phrase "after pruning 8.33% of parameters" is imprecise; the paper prunes 8.33% of attention heads, not 8.33% of the total parameters. This distinction matters because attention heads are only a subset of the model's parameters.
[Appendix D, Table 3] The model names in the first column are inconsistent: "XLM-R + BiLSTM + AttnIdiom" and "XLM-R + BiLSTM + AttnMetaphor" appear to combine the task name into the architecture name. This should be separated into a Model column and a Task column for clarity.
[Section 5] The importance formula I_h = E_{(x,y)~D} |∂L/∂h(h)| is written in a confusing way: the argument of the gradient should be a head-indexed quantity, and the notation h(h) is unclear. Please define the head output function h and write the derivative either as ∂L/∂h_i or with an explicit notation such as h^{(i)}(x).
[References] The reference "Paul Michel, Omer Levy, and Graham Neubig. 2019a. Are sixteen heads really better than one?" is missing venue and page information; the 2019b entry is complete, but 2019a should be completed for consistency.
[Appendix C] The caption for Figures 6 and 7 states that numerical decimal values are displayed, but the referenced figures are not visible in the text and their heatmap values are not otherwise provided; please include the figures or make them available in a readable resolution.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: reported pruning results are empirical measurements, not constructed predictions.

full rationale

I examined the paper's derivation chain for circularity. The paper does not derive its accuracy figures from the pruning importance scores; rather, it computes gradient importance Ih = E|∂L/∂h|, removes heads with zero importance, and then independently evaluates the pruned model on a held-out test split (Table 2). The reported 78% and 83% accuracies are measurements of post-pruning performance, not predictions constructed from the pruning criterion. The central interpretive claim—that metaphor classification relies on a broader, more distributed attention profile—is a post-hoc reading of the measured accuracy drop and the importance heatmaps. This is statistically fragile because the metaphor accuracy difference is only four sentences on a single 40-sentence test split, and the Limitations section candidly flags this ('evaluation on a single test split necessitates further validation with more diverse data to confirm the robustness of our findings'). Fragility and reproducibility concerns are not circularity, however. The importance metric is adopted from external prior work (Michel et al., 2019), and the Konidioms corpus is prior external work that the paper extends; no load-bearing self-citation chain is present. I cannot exhibit any equation or fitted parameter that reduces to the paper's own inputs, so under the required evidentiary standard the finding is no significant circularity.

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

The central claim depends on assumptions about the representativeness of a 40-sentence test set, the reliability of gradient-based importance scores on very small data, the adequacy of mBERT for Devanagari Konkani, and the correctness of native-speaker labels. The paper does not report inter-annotator agreement or dataset selection details, so these are all load-bearing postulates.

free parameters (4)

Zero-importance pruning threshold = 0.0
Heads with gradient importance exactly zero are pruned; this threshold is chosen post hoc and determines the 8.33% pruning rate.
Balanced subset size = 200 sentences (100 metaphor, 100 literal)
The authors selected 200 of 500 annotated sentences for a balanced evaluation set; the selection procedure is not described.
Early stopping patience = 10 epochs
Patience of 10 epochs is a hyperparameter that influences model selection.
Learning rate and batch size = 2e-5, 16
Standard fine-tuning hyperparameters, but the results depend on them.

assumptions (4)

domain assumption The 200-sentence balanced subset is representative of Konkani figurative language use.
The paper uses only 40 test sentences to draw conclusions about metaphor and idiom processing; Section 3 describes the subset creation but not random sampling or stratification beyond balance.
domain assumption Gradient-based importance scores (Michel et al., 2019) reliably identify removable heads when computed on dataset D.
Section 5 adopts the Michel et al. method without noting that importance estimates can be noisy on small data; D is not specified.
domain assumption The mBERT tokenizer and representations are adequate for Devanagari Konkani.
mBERT is used as a fixed feature extractor plus fine-tuned head; Section 2.1 notes Konkani's script diversity but the model only sees Devanagari.
domain assumption Native-speaker binary labels are ground truth.
Three native speakers verified labels, but the annotation guidelines and inter-annotator agreement are not reported.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Pruning for Performance: Efficient Idiom and Metaphor Classification in Low-Resource Konkani Using mBERT." pith.science (2026). https://pith.science/paper/OOZ2RNPF

@misc{pith2026250602005,
  author       = {Pith},
  title        = {Pith review of: Pruning for Performance: Efficient Idiom and Metaphor Classification in Low-Resource Konkani Using mBERT},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OOZ2RNPF}},
  note         = {Machine review of arXiv:2506.02005}
}

read the original abstract

In this paper, we address the persistent challenges that figurative language expressions pose for natural language processing (NLP) systems, particularly in low-resource languages such as Konkani. We present a hybrid model that integrates a pre-trained Multilingual BERT (mBERT) with a bidirectional LSTM and a linear classifier. This architecture is fine-tuned on a newly introduced annotated dataset for metaphor classification, developed as part of this work. To improve the model's efficiency, we implement a gradient-based attention head pruning strategy. For metaphor classification, the pruned model achieves an accuracy of 78%. We also applied our pruning approach to expand on an existing idiom classification task, achieving 83% accuracy. These results demonstrate the effectiveness of attention head pruning for building efficient NLP tools in underrepresented languages.

Figures

Figures reproduced from arXiv: 2506.02005 by the authors.

**Figure 2.** Heatmaps showing attention head importance scores across layers for idiom (left) and metaphor (right) [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Linguistic tree showing Konkani’s classifi [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Figures from the paper (4 more)

**Figure 4.** Figure 4: Geographic distribution of Konkani speakers [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

**Figure 5.** Figure 5: Flowchart outlining our experimental pipeline. [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]

**Figure 6.** Figure 6: Heatmap visualization of attention head importance across model layers for idiom [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]

**Figure 7.** Figure 7: Heatmap visualization of attention head importance across model layers for metaphor [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

20 extracted references · 17 canonical work pages

[1]

Steven Bird. 2020. https://aclanthology.org/2020.coling-main.313/ Decolonising speech and language technology . In 28th International Conference on Computational Linguistics, COLING 2020, pages 3504--3519. Association for Computational Linguistics (ACL)

work page 2020
[2]

Steven Bird. 2024. https://aclanthology.org/2024.acl-long.797/ Must nlp be extractive? In 62nd Annual Meeting of the Association for Computational Linguistics, ACL 2024, pages 14915--14929. Association for Computational Linguistics (ACL)

work page 2024
[3]

Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. https://doi.org/10.18653/v1/N19-1423 BERT : Pre-training of deep bidirectional transformers for language understanding . In Proceedings of the 2019 Conference of the North A merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long a...

doi:10.18653/v1/n19-1423 2019
[4]

Jovi D’Silva and Uzzal Sharma. 2022. https://doi.org/10.11591/ijece.v12i2.pp1990-2000 Automatic text summarization of konkani texts using pre-trained word embeddings and deep learning . International Journal of Electrical and Computer Engineering (IJECE), 12:1990

work page doi:10.11591/ijece.v12i2.pp1990-2000 2022
[5]

Encyclopedia Britannica . 2025. https://www.britannica.com/topic/Konkani-language Konkani language . Encyclopedia Britannica. Accessed May 11, 2025

work page 2025
[6]

Shi Feng, Eric Wallace, Alvin Grissom II, Mohit Iyyer, Pedro Rodriguez, and Jordan Boyd-Graber. 2018. https://doi.org/10.18653/v1/d18-1407 Pathologies of neural models make interpretations difficult . In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics

work page doi:10.18653/v1/d18-1407 2018
[7]

Palia Tukaram Gaonkar and Andre Rafael Fernandes. 2019. https://ceur-ws.org/Vol-2364/13_paper.pdf Digitization of Konkani Texts, and their Transliteration: An Initiative towards Preservation of a Language Culture . CEUR Workshop Proceedings, 2364:110--117

work page 2019
[8]

Weicheng Ma, Kai Zhang, Renze Lou, Lili Wang, and Soroush Vosoughi. 2021. https://doi.org/10.18653/v1/2021.acl-long.152 Contributions of transformer attention heads in multi- and cross-lingual tasks . In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language P...

work page doi:10.18653/v1/2021.acl-long.152 2021

Show all 20 references

[9]

Paul Michel, Omer Levy, and Graham Neubig. 2019. https://papers.nips.cc/paper/9551-are-sixteen-heads-really-better-than-one Are sixteen heads really better than one? In Advances in Neural Information Processing Systems, volume 32, pages 14014--14024. Curran Associates, Inc

2019
[10]

Pratik Naik, Nilesh Kamat, Shweta Naik, Prashant Naik, and Rajesh Kamat. 2024. https://aclanthology.org/2024.lrec-main.867.pdf Konidioms corpus: A dataset of idioms in konkani language . In Proceedings of the 2024 International Conference on Language Resources and Evaluation (...

2024
[11]

Nigatu, I

Y. Nigatu, I. D. Raji, M. Choudhury, S. Diddee, G. Le Ferrand, J. Dearden, and A. Tucker. 2024. https://arxiv.org/html/2410.20817v1 The zeno's paradox of 'low-resource' languages . ArXiv preprint arXiv:2410.20817

2024 arXiv
[12]

Annie Rajan, Ambuja Salgaonkar, and Ramprasad Joshi. 2020. https://doi.org/10.1016/j.cosrev.2020.100299 A survey of konkani nlp resources . Computer Science Review, 38:100299

2020
[13]

Naziya Mahamdul Shaikh and Jyoti Pawar. 2024. https://aclanthology.org/2024.icon-1.6/ Identification of idiomatic expressions in K onkani language using neural networks . In Proceedings of the 21st International Conference on Natural Language Processing (ICON), pages 54--58, A...

2024
[14]

Pawar, and Mubarak Banu Sayed

Naziya Mahamdul Shaikh, Jyoti D. Pawar, and Mubarak Banu Sayed. 2024. https://aclanthology.org/2024.lrec-main.867/ Konidioms corpus: A dataset of idioms in K onkani language . In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Reso...

2024
[15]

Ekaterina Shutova. 2015. https://doi.org/10.1162/COLI_a_00233 Design and evaluation of metaphor processing systems . Computational Linguistics, 41(4):579--623

2015 doi
[16]

u shurescue: Reviving the endangered n \

Ivory Yang, Weicheng Ma, and Soroush Vosoughi. 2025 a . https://aclanthology.org/2025.coling-main.468/ N \"u shurescue: Reviving the endangered n \"u shu language with ai . In Proceedings of the 31st International Conference on Computational Linguistics, pages 7020--7034

2025
[17]

Ivory Yang, Weicheng Ma, Chunhui Zhang, and Soroush Vosoughi. 2025 b . https://aclanthology.org/2025.naacl-short.24/ Is it N avajo? accurate language detection for endangered athabaskan languages . In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of...

2025
[18]

Arnav Yayavaram, Siddharth Yayavaram, Prajna Devi Upadhyay, and Apurba Das. 2024. https://aclanthology.org/2024.mwe-1.26 BERT -based idiom identification using language translation and word cohesion . In Proceedings of the Joint Workshop on Multiword Expressions and Universal ...

2024
[19]

URL: " 'urlintro :=

ENTRY address author booktitle chapter edition editor howpublished institution journal key month note number organization pages publisher school series title type volume year eprint doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before...
[20]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

Steven Bird. 2020. https://aclanthology.org/2020.coling-main.313/ Decolonising speech and language technology . In 28th International Conference on Computational Linguistics, COLING 2020, pages 3504--3519. Association for Computational Linguistics (ACL)

work page 2020

[2] [2]

Steven Bird. 2024. https://aclanthology.org/2024.acl-long.797/ Must nlp be extractive? In 62nd Annual Meeting of the Association for Computational Linguistics, ACL 2024, pages 14915--14929. Association for Computational Linguistics (ACL)

work page 2024

[3] [3]

Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. https://doi.org/10.18653/v1/N19-1423 BERT : Pre-training of deep bidirectional transformers for language understanding . In Proceedings of the 2019 Conference of the North A merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long a...

doi:10.18653/v1/n19-1423 2019

[4] [4]

Jovi D’Silva and Uzzal Sharma. 2022. https://doi.org/10.11591/ijece.v12i2.pp1990-2000 Automatic text summarization of konkani texts using pre-trained word embeddings and deep learning . International Journal of Electrical and Computer Engineering (IJECE), 12:1990

work page doi:10.11591/ijece.v12i2.pp1990-2000 2022

[5] [5]

Encyclopedia Britannica . 2025. https://www.britannica.com/topic/Konkani-language Konkani language . Encyclopedia Britannica. Accessed May 11, 2025

work page 2025

[6] [6]

Shi Feng, Eric Wallace, Alvin Grissom II, Mohit Iyyer, Pedro Rodriguez, and Jordan Boyd-Graber. 2018. https://doi.org/10.18653/v1/d18-1407 Pathologies of neural models make interpretations difficult . In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics

work page doi:10.18653/v1/d18-1407 2018

[7] [7]

Palia Tukaram Gaonkar and Andre Rafael Fernandes. 2019. https://ceur-ws.org/Vol-2364/13_paper.pdf Digitization of Konkani Texts, and their Transliteration: An Initiative towards Preservation of a Language Culture . CEUR Workshop Proceedings, 2364:110--117

work page 2019

[8] [8]

Weicheng Ma, Kai Zhang, Renze Lou, Lili Wang, and Soroush Vosoughi. 2021. https://doi.org/10.18653/v1/2021.acl-long.152 Contributions of transformer attention heads in multi- and cross-lingual tasks . In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language P...

work page doi:10.18653/v1/2021.acl-long.152 2021