PDF Annotation & Transcription Experts – Bengali

What you would do

• Identify and bound every meaningful region on the page (titles, sections, paragraphs, lists, tables, figures)
• Classify each region consistently using a defined component taxonomy
• Document reading order and relationships between regions
• Transcribe all text exactly as it appears in native Bengali script and diacritics
• Flag regions with handwriting or legibility issues and record page metadata

Who they want

• Native Bengali speaker with complete command of Bengali script and diacritics
• Background in annotation, transcription, translation, localization, or proofreading
• Character-level accuracy orientation; precision valued over speed
• Comfort with unfamiliar layouts including multi-column pages, handwritten forms, and mixed-script documents
• Familiarity with document production, OCR, or regional-language AI data work preferred

What the interview asks about

1. 1.Bengali script and orthography mastery

   AI models learn from transcription accuracy; this tests whether your Bengali command includes diacritics, conjunct forms, and numeral variants.

   For example: “A Bengali page contains a word with a complex conjunct form and multiple diacritics. You're unsure whether you're reading it correctly. How would you verify the correct transcription?”

2. 2.Complex page layout navigation

   Real Bengali documents have multi-column layouts, sidebars, figures with captions, and mixed scripts; this tests whether you can handle that complexity.

   For example: “A newspaper page has three columns of text, a sidebar with a different font, a photograph with a caption in Bengali, and an embedded advertisement. How would you annotate the reading order and region boundaries?”

3. 3.Taxonomy consistency over hundreds of pages

   Dataset quality requires consistent classification; this tests whether you apply the same rules even when documents vary widely.

   For example: “You've annotated 200 pages. You notice that what you called a 'caption' on page 1 could also be classified as a 'note' or 'label.' How would you decide and how would you handle earlier pages?”

4. 4.Handwriting recognition and legibility judgment

   Training data must accurately reflect what can and cannot be read; this tests whether you distinguish handwriting confidence from evasion.

   For example: “A page has handwritten Bengali text in the margin. The handwriting is legible but ambiguous in a few characters. Do you transcribe your best guess, flag it as unclear, or skip it?”

5. 5.Quality review and error identification

   Every annotation is peer-reviewed; this evaluates whether you can spot errors in others' work systematically.

   For example: “You're reviewing a colleague's annotations of a table-heavy page. You notice they've classified one column header as a separate region, but you think it belongs to the table. How would you document that concern?”

A task you may get

Annotate a 1-2 page Bengali document sample, identifying regions, applying consistent component classification, and transcribing all text accurately in Bengali script.

How to prepare

• Review a Bengali document style guide or annotation taxonomy to understand classification principles
• Collect 3-4 real Bengali documents in various formats (newspaper, textbook, form) to practice on
• Prepare a checklist of common Bengali orthographic elements to verify your accuracy

The facts

Pay

$10–11/hr

Hours

Hourly, 20 hours a week

Where

Remote

Open to

IND

Field

Miscellaneous

Project name

Neon

Posted

9/2/2026

Places left

10

We wrote this page from the public Mercor listing. It may be out of date, so read the full posting before you apply.