The accessibility to social media platforms can be improved with the use of machine translation (MT). Non-standard features present in user-generated on social media content such as hashtags, emojis, and alternative spellings can lead to mistranslated instances by the MT systems. In this paper, we investigate the impact of MT on offensive language identification in Indo-Aryan languages. We use both original and MT datasets to evaluate the performance of various offensive language models. Our evaluation indicates that offensive language identification models achieve superior performance on original data than on MT data, and that the models trained on MT data identify offensive language more precisely on MT data than the models trained on original data.

This paper introduces and investigates the problem of Task and Language Incremental Continual Learning (TLCL), wherein a multilingual model is systematically updated to accommodate new tasks in previously learned languages or new languages for established tasks. This significant yet previously unexplored area holds substantial practical relevance as it mirrors the dynamic requirements of real-world applications. We benchmark a representative set of continual learning (CL) algorithms for TLCL. Furthermore, we propose Task and Language-Specific Adapters (TLSA), an adapter-based parameter-efficient fine-tuning strategy. TLSA facilitates cross-lingual and cross-task transfer and outperforms other parameter-efficient fine-tuning techniques. Crucially, TLSA reduces parameter growth stemming from saving adapters to linear complexity from polynomial complexity as it was with parameter isolation-based adapter tuning. We conducted experiments on several NLP tasks arising across several languages. We observed that TLSA outperforms all other parameter-efficient approaches without requiring access to historical data for replay.