BanglaVoice(Active, Passive, Middle)

The BanglaVoice dataset contains 4,397 unique Bangla sentences annotated for grammatical voice classification, including 1,531 Passive, 1,459 Active, and 1,407 Middle constructions. The sentences were manually collected and carefully verified to ensure accurate labeling. Each entry includes a cleaned Bangla sentence and its corresponding voice category, with optional English translations where available. Duplicate and incomplete entries have been removed to maintain data quality and consistency.
Basic statistical analysis indicates structural differences across voice types. Passive sentences are generally longer, with an average length of 28.48 characters and 4.89 words, compared to 21.11 characters and 3.84 words for Active sentences and 22.42 characters and 4.04 words for Middle sentences. The dataset is suitable for Bangla grammatical analysis, voice classification research, morphological studies, and machine learning applications in Natural Language Processing.

-------Value of the Data-------
1. One of the first structured sentence-level datasets for Bangla grammatical voice classification.
2.Enables research in Bangla morpho-syntactic analysis, voice detection, and grammatical transformation modeling.
3. Supports machine learning benchmarking for text classification in an under-resourced language.
4. Provides balanced class distribution suitable for unbiased model evaluation.