KUET Whispers Dataset

The KUET Whispers Text Dataset is a curated collection of 2,485 Bangla and English text submissions, originally sourced from the KUET Whispers online community. Each entry consists of short, anonymous, emotion-rich text snippets, representing natural whispered communication in informal, user-generated contexts.

The dataset includes:

Raw and cleaned text versions for each post, enabling direct NLP preprocessing or model-ready usage.
Metadata featuressuch as timestamps, anonymized sender tags, post engagement (likes, comments), and original source URLs.
Language labels for multilingual text processing (English dataset).

This dataset is designed for natural language processing (NLP) tasks focusing on whispered or low-volume communication, particularly:

* Sentiment and emotion classification
* Whispered text normalization
* Social interaction and engagement modeling
* Low-resource text data research in Bangla and English

All entries are anonymized and non-identifiable. Data is shared under a **Creative Commons Attribution 4.0 (CC BY 4.0)** license, allowing academic and commercial usage with appropriate citation.