anvay: A Web-based Tool for Interpretive Topic Modelling in Bengali

anvay is a web-based tool for topic modelling in Bengali, developed for exploratory reading and interpretive analysis.It provides a full pipeline for Latent Dirichlet Allocation (LDA) (Blei et al., 2003)-from corpus ingestion and preprocessing to model configuration and visual output-within a lightweight, browser-based interface.The tool foregrounds user interpretation: rather than providing coherence scores or fixed topic labels, anvay presents the model output to be read, interpreted, and adjusted by the user.Designed for literary, journalistic, and historical corpora in Bengali, anvay supports a range of language-specific preprocessing functions including lemmatisation, frequency filtering, and stopword pruning.The outputs, ranging from topic-word networks to document-level previews, are rendered with clarity and designed to enable close reading.Each topic is accessible through multiple lenses: top words, paragraph-level examples, document weights, and corpus-wide distribution.Beyond its technical function, anvay is an intervention in how we teach and understand computational methods within the humanities in low-resource contexts.