arXiv subjects

Soham Khisa

Publications and source records attributed to Soham Khisa.

3 recordsLinked to original sources

arXiv2026

Re-educating Educated Ones: A Case Study on Chakma Language Revitalization in Chittagong Hill Tracts

Indigenous languages face significant cultural oppression from official state languages, particularly in the Global South. We investigate the Bangladeshi Chakma language revitalization movement, a community grappling with language liquidity and amalgamation into the dominant Bengali language. Our six-month-long qualitative study involving interviews and focus group discussions with Chakma language learning stakeholders uncovered existing community socio-economic challenges and resilience strategies. We noted the need for culturally grounded digital tools and resources. We propose an ICT-mediated community-centric framework for Indigenous language revitalization in the Global South, emphasizing the integration of historical identity elements, stakeholder-defined requirements, and effective digital engagement strategies to empower communities in preserving their linguistic and cultural heritage.

cs.HC↗

arXiv2026

ChakmaNMT: Machine Translation for a Low-Resource and Endangered Language via Transliteration

We present the first systematic study of machine translation for Chakma, an endangered and extremely low-resource Indo-Aryan language, with the goal of supporting language access and preservation. We introduce a new Chakma-Bangla parallel and monolingual dataset, along with a trilingual Chakma-Bangla-English benchmark for evaluation. To address script mismatch and data scarcity, we propose a character-level transliteration framework that exploits the close orthographic and phonological relationship between Chakma and Bangla, preserving semantic content while enabling effective transfer from Bangla and multilingual pretrained models. We benchmark from-scratch MT, fine-tuned pretrained models, and large language models via in-context learning. Results show that transliteration is essential and that fine-tuning and in-context learning substantially outperform from-scratch baselines, with strong asymmetry across translation directions.

cs.CL↗

arXiv2025

VaCDA: Variational Contrastive Alignment-based Scalable Human Activity Recognition

Technological advancements have led to the rise of wearable devices with sensors that continuously monitor user activities, generating vast amounts of unlabeled data. This data is challenging to interpret, and manual annotation is labor-intensive and error-prone. Additionally, data distribution is often heterogeneous due to device placement, type, and user behavior variations. As a result, traditional transfer learning methods perform suboptimally, making it difficult to recognize daily activities. To address these challenges, we use a variational autoencoder (VAE) to learn a shared, low-dimensional latent space from available sensor data. This space generalizes data across diverse sensors, mitigating heterogeneity and aiding robust adaptation to the target domain. We integrate contrastive learning to enhance feature representation by aligning instances of the same class across domains while separating different classes. We propose Variational Contrastive Domain Adaptation (VaCDA), a multi-source domain adaptation framework combining VAEs and contrastive learning to improve feature representation and reduce heterogeneity between source and target domains. We evaluate VaCDA on multiple publicly available datasets across three heterogeneity scenarios: cross-person, cross-position, and cross-device. VaCDA outperforms the baselines in cross-position and cross-device scenarios.

cs.LG↗