This paper presents the submission of IIITH-BUT to the IWSLT 2025 shared task on speech translation for the low-resource Bhojpuri-Hindi language pair. We explored the impact of hyperparameter optimisation and data augmentation techniques on the performance of the SeamlessM4T model fine-tuned for this specific task. We systematically investigated a range of hyperparameters including learning rate schedules, number of update steps, warm-up steps, label smoothing, and batch sizes; and report their effect on translation quality. To address data scarcity, we applied speed perturbation and SpecAugment and studied their effect on translation quality. We also examined the use of cross-lingual signal through joint training with Marathi and Bhojpuri speech data. Our experiments reveal that careful selection of hyperparameters and the application of simple yet effective augmentation techniques significantly improve performance in low-resource settings. We also analysed the translation hypotheses to understand various kinds of errors that impacted the translation quality in terms of BLEU

This paper describes the systems submitted for Marathi to Hindi low-resource speech translation task. Our primary submission is based on an end-to-end direct speech translation system, whereas the contrastive one is a cascaded system. The backbone of both the systems is a Hindi-Marathi bilingual ASR system trained on 2790 hours of imperfect transcribed speech. The end-to-end speech translation system was directly initialized from the ASR, and then fine-tuned for direct speech translation with an auxiliary CTC loss for translation. The MT model for the cascaded system is initialized from a cross-lingual language model, which was then fine-tuned using 1.6 M parallel sentences. All our systems were trained from scratch on publicly available datasets. In the end, we use a language model to re-score the n-best hypotheses. Our primary submission achieved 30.5 and 39.6 BLEU whereas the contrastive system obtained 21.7 and 28.6 BLEU on official dev and test sets respectively. The paper also presents the analysis on several experiments that were conducted and outlines the strategies for improving speech translation in low-resource scenarios.

This paper presents our proposed method for SemEval-2023 Task 12, which focuses on sentiment analysis for low-resource African languages. Our method utilizes a language-centric domain adaptation approach which is based on adversarial training, where a small version of Afro-XLM-Roberta serves as a generator model and a feed-forward network as a discriminator. We participated in all three subtasks: monolingual (12 tracks), multilingual (1 track), and zero-shot (2 tracks). Our results show an improvement in weighted F1 for 13 out of 15 tracks with a maximum increase of 4.3 points for Moroccan Arabic compared to the baseline. We observed that using language family-based labels along with sequence-level input representations for the discriminator model improves the quality of the cross-lingual sentiment analysis for the languages unseen during the training. Additionally, our experimental results suggest that training the system on languages that are close in a language families tree enhances the quality of sentiment analysis for low-resource languages. Lastly, the computational complexity of the prediction step was kept at the same level which makes the approach to be interesting from a practical perspective. The code of the approach can be found in our repository.