Bengali Transcription and Annotation Expert

Obsidian; København; On-site; DKK 165,000 - 276,000; Full time

Job summary

Obsidian seeks a native Bengali speaker to annotate and transcribe Bengali-language documents with exacting character-level accuracy. You will identify and bound meaningful regions, record reading order, and ensure faithful transcription of handwritten and printed text.

This role supports training multilingual document AI by handling diverse layouts and dense scripts, with expert review to ensure consistency across pages.

Qualifications

• Native Bengali speaker with full command of Bengali script, diacritics, and conjunct forms.
• Experience with document annotation, transcription, localization, proofreading, or related data review is preferred.
• Character-level accuracy is essential; attention to diacritics and script details is critical.

Responsibilities

• Source Bengali PDFs or documents containing multimodal elements and record source details.
• Annotate the page structure: bound regions, types, and reading order indices.
• Link regions to parent figures/tables and transcribe Bengali text faithfully, including handwritten content.
• Capture page metadata: language, document type, source, page dimensions, and handwriting flags.
• Undergo second-expert review to ensure accuracy and layout diversity.

Job description

Fluent Language Skills Required: Bengali. Native fluency in Bengali, including full command of Bengali script and orthography, is required for this position. All annotation and transcription work is performed in Bengali.

Why This Role Exists

Document understanding breaks down fastest in the languages that parsing and vision-language models rarely see. This project builds training data for exactly those languages: Bengali, alongside four other Indic scripts, Japanese and Korean. Each task takes a real, publicly available PDF page and produces a complete structural map of that page, paired with a faithful transcription of every text region in the original script.

The dataset deliberately concentrates on the material models handle worst: handwriting, dense multi-column layouts, tables, diagrams, and mixed-script pages. Documents are drawn from newspapers, textbooks, examinations, and everyday formats such as flyers, forms, manuals, menus, brochures, notices and worksheets, so that the corpus reflects the real diversity of Bengali documents rather than a narrow band of easily parsed ones.

Delivered work is human-authored throughout. Component identification, component typing, reading order and all transcription are performed by people, not generated by parsing models.

What You'll Do

• Source documents: find a publicly accessible Bengali PDF in an assigned document type, containing at least one multimodal element (images, tables, diagrams, or handwriting), and record where you obtained it

• Annotate structure: identify and bound every meaningful region of the page - document title, section heading, paragraph, list, table, figure, diagram, caption, formula, question, answer field - and assign each a component type and a reading-order index

• Record relationships: link each region to the figure or table it belongs to through a parent component identifier

• Transcribe faithfully: reproduce all text exactly as it appears in Bengali script, including handwritten content, flagging any region where the source is not legible

• Capture page metadata: language, document type, source, page dimensions, and flags for tables, formulas and handwriting

• Review a colleague's work: every task is reviewed end to end by a second Bengali expert, and experienced annotators take on that review

Who You Are

• You are a native Bengali speaker with full command of the script, its diacritics and its conjunct forms

• You have worked with documents: annotation, transcription, translation, localization, subtitling, proofreading, journalism, or regional-language data review

• You are exact: character-level accuracy matters more here than speed, and a single wrong diacritic is a defect

• You are systematic: you apply a taxonomy consistently across hundreds of pages rather than improvising per document

• You are comfortable with unfamiliar layouts: multi-column newspapers, exam papers, handwritten forms

Nice-to-Have Specialties

• Regional-language AI data: annotation, labeling, grading, or bilingual evaluation for training datasets

• Transcription and localization: MTPE, subtitling, bilingual QA, OCR correction or post-editing

• Document production: typesetting, copy-editing, proofreading, or digitization of Bengali-language material

• Script and encoding: Unicode normalization, Bengali input methods, numeral-form and character-form accuracy

What Success Looks Like

• Every meaningful region on the page is captured, correctly bounded and correctly typed

• Reading order reflects how the page is actually read, including across columns

• Transcriptions match the source character for character, in Bengali script rather than transliteration

• Your tasks pass second-expert review the first time

• The documents you bring in add layout diversity rather than repeating templates already in the corpus

Why Join Mercor

• Build the training data that makes document AI work in scripts it currently handles badly

• Work from real published Bengali documents rather than synthetic or templated pages

• Quality leads on this project: accuracy is the first measure, with handling time tracked alongside it