Bengali Document Annotation & Transcription Expert

Mercor; Helsinki; On-site; EUR 25,000 - 42,000; Full time

Job summary

Obsidian is seeking a detail-oriented Bengali annotator to process training data for multilingual document understanding. You will source Bengali documents, bound page regions, and create precise transcriptions in Bengali script for complex layouts.

The role emphasizes high-precision transcription, correct reading order, and strict adherence to a consistent taxonomy across hundreds of pages. Collaboration and peer review are core parts of the workflow.

Qualifications

• You are a native Bengali speaker with full command of the Bengali script, its diacritics and conjunct forms.
• You have experience with annotation, transcription, localization, proofreading, or related document tasks.
• Character-level accuracy is essential; careful attention to diacritics and orthography is required.

Responsibilities

• Source Bengali PDFs or publicly accessible Bengali documents for annotation tasks.
• Annotate page structure: bound and label regions like title, headings, paragraphs, lists, tables, diagrams, captions, formulas, questions and answers.
• Record relationships by linking regions to their parent figures or tables.
• Transcribe all Bengali text exactly as it appears, including handwritten content.
• Capture page metadata: language, document type, source, dimensions, and flags for tables, formulas and handwriting.
• Review a colleague's work as part of a two-expert quality control process.

Skills

Native Bengali speaker

Full command of Bengali script

Job description

Obsidian is seeking a detail-oriented Bengali annotator to process training data for multilingual document understanding. You will source Bengali documents, bound page regions, and create precise transcriptions in Bengali script for complex layouts.

The role emphasizes high-precision transcription, correct reading order, and strict adherence to a consistent taxonomy across hundreds of pages. Collaboration and peer review are core parts of the workflow.