Benchmarking BERT-based Models for Sentence-level Topic Classification in Nepali Language

Many sentiment datasets use one annotator for each item, so the test labels show the conventions of that person. We build a Nepali benchmark from two annotators who labelled the same sentences blind. They labelled 5,614 sentences from tweets of Nepali public figures on a five-point scale. We keep only the 3,189 sentences where both give the same polarity. On the three-class benchmark, NepBERTa with a fixed tokenizer gets a macro-F1 of 0.877 and IndicBERTv2 0.867. Three commercial large language models (LLMs) get 0.882 to 0.896 zero-shot, and these differences are not significant. On the five-point scale, 766 sentences have two adjacent labels, for example Pos and SPos. We accept both labels and train on both. On this task, Claude Opus 5.5 beats the best encoders by 0.10 to 0.12 macro-F1. A tokenizer audit adds a second result. The default tokenizers of three Nepali BERT models delete the non-spacing Devanagari vowel signs and the virama. When we restore the marks, macro-F1 increases by 0.084 for NepBERTa but decreases by 0.095 for a model pretrained on stripped text.

Nepal Bhasha (Newari), an endangered language of the Kathmandu Valley, remains digitally marginalized due to the severe scarcity of annotated speech resources. In this work, we introduce Nwāchā Munā, a newly curated 5.39-hour manually transcribed Devanagari speech corpus for Nepal Bhasha, and establish the first benchmark using script-preserving acoustic modeling. We investigate whether proximal cross-lingual transfer from a geographically and linguistically adjacent language (Nepali) can rival large-scale multilingual pretraining in an ultra-low-resource Automatic Speech Recognition (ASR) setting. Fine-tuning a Nepali Conformer model reduces the Character Error Rate (CER) from a 52.54% zero-shot baseline to 17.59% with data augmentation, effectively matching the performance of the multilingual Whisper-Small model despite utilizing significantly fewer parameters. Our findings demonstrate that proximal transfer from Nepali language serves as a computationally efficient alternative to massive multilingual models. We openly release the dataset and benchmarks to digitally enable the Newari community and foster further research in Nepal Bhasha.

Topic modeling is a natural language processing technique used to discover latent themes and abstract topics from text corpora by grouping co-occurring keywords. Although widely researched in English, topic modeling remains understudied in Bengali due to a lack of adequate resources and initiatives. Existing Bengali topic modeling research lacks exploration of modern methodological approaches, reproducible implementations, diverse datasets, and systematic benchmarking. To bridge these gaps, this study presents Graph-based Hybrid Topic Model (GHTM), a novel architecture that synthesizes the distinct strengths of multiple paradigms by unifying statistical (TF-IDF), semantic (GloVe), and structural (GCN) modalities to construct semantically aligned document representations. GHTM later employs matrix factorization for topic extraction, offering a computationally lightweight alternative to resource-intensive neural and graph-based approaches. Through comprehensive benchmarking across three Bengali datasets of varying scales, GHTM consistently surpasses both traditional and contemporary topic modeling approaches, achieving improved topic coherence (NPMI: 0.27-0.28) and high topic diversity while maintaining computational efficiency. The model also outperformed established graph-based models on the English 20Newsgroups benchmark. Additionally, we introduce NCTBText, a diverse Bengali textbook-based dataset comprising 8,650 text documents, providing much-needed topical diversity beyond newspaper-centric Bengali corpora.