Bengali (Bangla) is used across Bangladesh and parts of India, with regional and social variation that can matter for language technology. A project should distinguish its target location, spoken variety, writing conventions, domain, and intended application before data collection begins.
Speech and transcription
For speech recognition or conversational systems, the specification may need to cover recording devices, background noise, speaker coverage, code-switching, pronunciation, and whether transcripts should preserve disfluencies. Define spelling conventions, punctuation, treatment of English terms, timestamps, speaker labels, and rules for unclear audio before annotators begin.
Text and annotation
Bengali text projects may involve sentiment, intent, entities, moderation, search relevance, or translation evaluation. Each task requires a clear label schema, representative examples, edge-case guidance, adjudication rules, and a way to measure reviewer agreement. Script normalization and punctuation handling should be documented rather than silently assumed.
A useful brief
When sourcing Bengali language data collection, Bengali transcription, or Bengali annotation, describe the target task and variety, source and consent requirements, sample size, quality checks, delivery format, and rights to use the data. Langigs can discuss a proposed scope, but actual language and task availability must be confirmed individually.
