Pramit Bhattacharyya

Pramit Bhattacharyya, Joydeep Mondal, Subhadip Maji et al.

Bangla (or Bengali) is the fifth most spoken language globally; yet, the state-of-the-art NLP in Bangla is lagging for even simple tasks such as lemmatization, POS tagging, etc. This is partly due to lack of a varied quality corpus. To alleviate this need, we build Vacaspati, a diverse corpus of Bangla literature. The literary works are collected from various websites; only those works that are publicly available without copyright violations or restrictions are collected. We believe that published literature captures the features of a language much better than newspapers, blogs or social media posts which tend to follow only a certain literary pattern and, therefore, miss out on language variety. Our corpus Vacaspati is varied from multiple aspects, including type of composition, topic, author, time, space, etc. It contains more than 11 million sentences and 115 million words. We also built a word embedding model, Vac-FT, using FastText from Vacaspati as well as trained an Electra model, Vac-BERT, using the corpus. Vac-BERT has far fewer parameters and requires only a fraction of resources compared to other state-of-the-art transformer models and yet performs either better or similar on various downstream tasks. On multiple downstream tasks, Vac-FT outperforms other FastText-based models. We also demonstrate the efficacy of Vacaspati as a corpus by showing that similar models built from other corpora are not as effective. The models are available at https://bangla.iitk.ac.in/.

Pramit Bhattacharyya, Arnab Bhattacharya

In this paper, we present a comprehensive corpus-driven analysis of Bangla literary and newspaper texts to investigate their lexical diversity, structural complexity and readability. We undertook Vacaspati and IndicCorp, which are the most extensive literature and newspaper-only corpora for Bangla. We examine key linguistic properties, including the type-token ratio (TTR), hapax legomena ratio (HLR), Bigram diversity, average syllable and word lengths, and adherence to Zipfs Law, for both newspaper (IndicCorp) and literary corpora (Vacaspati).For all the features, such as Bigram Diversity and HLR, despite its smaller size, the literary corpus exhibits significantly higher lexical richness and structural variation. Additionally, we tried to understand the diversity of corpora by building n-gram models and measuring perplexity. Our findings reveal that literary corpora have higher perplexity than newspaper corpora, even for similar sentence sizes. This trend can also be observed for the English newspaper and literature corpus, indicating its generalizability. We also examined how the performance of models on downstream tasks is influenced by the inclusion of literary data alongside newspaper data. Our findings suggest that integrating literary data with newspapers improves the performance of models on various downstream tasks. We have also demonstrated that a literary corpus adheres more closely to global word distribution properties, such as Zipfs law, than a newspaper corpus or a merged corpus of both literary and newspaper texts. Literature corpora also have higher entropy and lower redundancy values compared to a newspaper corpus. We also further assess the readability using Flesch and Coleman-Liau indices, showing that literary texts are more complex.

Pramit Bhattacharyya, Arnab Bhattacharya

Large language models (LLMs) have achieved remarkable success across various natural language processing tasks. However, most LLM models use traditional tokenizers like BPE and SentencePiece, which fail to capture the finer nuances of a morphologically rich language like Bangla (Bengali). In this work, we introduce BanglaByT5, the first byte-level encoder-decoder model explicitly tailored for Bangla. Built upon a small variant of Googles ByT5 architecture, BanglaByT5 is pre-trained on a 14GB curated corpus combining high-quality literary and newspaper articles. Through zeroshot and supervised evaluations across generative and classification tasks, BanglaByT5 demonstrates competitive performance, surpassing several multilingual and larger models. Our findings highlight the efficacy of byte-level modelling for morphologically rich languages and highlight BanglaByT5 potential as a lightweight yet powerful tool for Bangla NLP, particularly in both resource-constrained and scalable environments.

V. S. D. S. Mahesh Akavarapu, Hrishikesh Terdalkar, Pramit Bhattacharyya et al.

Large Language Models (LLMs) have demonstrated remarkable generalization capabilities across diverse tasks and languages. In this study, we focus on natural language understanding in three classical languages -- Sanskrit, Ancient Greek and Latin -- to investigate the factors affecting cross-lingual zero-shot generalization. First, we explore named entity recognition and machine translation into English. While LLMs perform equal to or better than fine-tuned baselines on out-of-domain data, smaller models often struggle, especially with niche or abstract entity types. In addition, we concentrate on Sanskrit by presenting a factoid question-answering (QA) dataset and show that incorporating context via retrieval-augmented generation approach significantly boosts performance. In contrast, we observe pronounced performance drops for smaller LLMs across these QA tasks. These results suggest model scale as an important factor influencing cross-lingual generalization. Assuming that models used such as GPT-4o and Llama-3.1 are not instruction fine-tuned on classical languages, our findings provide insights into how LLMs may generalize on these languages and their consequent utility in classical studies.

Pramit Bhattacharyya, Arnab Bhattacharya

Large Language Models (LLMs) perform exceedingly well in Natural Language Understanding (NLU) tasks for many languages including English. However, despite being the fifth most-spoken language globally, Grammatical Error Correction (GEC) in Bangla remains underdeveloped. In this work, we investigate how LLMs can be leveraged for improving Bangla GEC. For that, we first do an extensive categorization of 12 error classes in Bangla, and take a survey of native Bangla speakers to collect real-world errors. We next devise a rule-based noise injection method to create grammatically incorrect sentences corresponding to correct ones. The Vaiyakarana dataset, thus created, consists of 5,67,422 sentences of which 2,27,119 are erroneous. This dataset is then used to instruction-tune LLMs for the task of GEC in Bangla. Evaluations show that instruction-tuning with \name improves GEC performance of LLMs by 3-7 percentage points as compared to the zero-shot setting, and makes them achieve human-like performance in grammatical error identification. Humans, though, remain superior in error correction.

Utsav Singh, Pramit Bhattacharyya, Vinay P. Namboodiri

Large language models (LLMs) have shown remarkable abilities in logical reasoning, in-context learning, and code generation. However, translating natural language instructions into effective robotic control policies remains a significant challenge, especially for tasks requiring long-horizon planning and operating under sparse reward conditions. Hierarchical Reinforcement Learning (HRL) provides a natural framework to address this challenge in robotics; however, it typically suffers from non-stationarity caused by the changing behavior of the lower-level policy during training, destabilizing higher-level policy learning. We introduce LGR2, a novel HRL framework that leverages LLMs to generate language-guided reward functions for the higher-level policy. By decoupling high-level reward generation from low-level policy changes, LGR2 fundamentally mitigates the non-stationarity problem in off-policy HRL, enabling stable and efficient learning. To further enhance sample efficiency in sparse environments, we integrate goal-conditioned hindsight experience relabeling. Extensive experiments across simulated and real-world robotic navigation and manipulation tasks demonstrate LGR2 outperforms both hierarchical and non-hierarchical baselines, achieving over 55% success rates on challenging tasks and robust transfer to real robots, without additional fine-tuning.

Pramit Bhattacharyya, Raghava Mutharaju

Building an ontology is not only a time-consuming process, but it is also confusing, especially for beginners and the inexperienced. Although ontology developers can take the help of domain experts in building an ontology, they are not readily available in several cases for a variety of reasons. Ontology developers have to grapple with several questions related to the choice of classes, properties, and the axioms that should be included. Apart from this, there are aspects such as modularity and reusability that should be taken care of. From among the thousands of publicly available ontologies and vocabularies in repositories such as Linked Open Vocabularies (LOV) and BioPortal, it is hard to know the terms (classes and properties) that can be reused in the development of an ontology. A similar problem exists in implementing the right set of ontology design patterns (ODPs) from among the several available. Generally, ontology developers make use of their experience in handling these issues, and the inexperienced ones have a hard time. In order to bridge this gap, we propose a tool named OntoSeer, that monitors the ontology development process and provides suggestions in real-time to improve the quality of the ontology under development. It can provide suggestions on the naming conventions to follow, vocabulary to reuse, ODPs to implement, and axioms to be added to the ontology. OntoSeer has been implemented as a Protégé plug-in.