PDF Annotation & Transcription Experts – Bengali | $12.68/hr Remote
Listing checked October 16, 2026 · pay as published by Mercor
Overview
Neon builds training data for document AI in languages that parsing and vision-language models often miss. This project covers Bengali plus four other Indic scripts, Japanese and Korean. You will map real public PDF pages into structural components and transcribe every text region in original script. Source documents include newspapers, textbooks, exams, flyers, forms, manuals, menus, brochures, notices and worksheets, with a focus on handwriting, dense columns, tables, diagrams and mixed-script layouts. All component identification, typing, reading order and transcription come from human experts.
What You'll Do7
- 1Review each assigned PDF page and confirm it uses Bengali, remains legible, contains real content, and excludes personal details before annotation starts.
- 2Map every meaningful region on the page: document title, section heading, paragraph, list, table, figure, diagram, caption, formula, question, and answer field.
- 3Assign a component type and reading-order index to each bounded region.
- 4Link each region to its parent figure or table through a parent component identifier.
- 5Transcribe all text as it appears in Bengali script, including handwritten content, and flag any region that is not legible.
- 6Capture page metadata: language, document type, source, page dimensions, plus flags for tables, formulas, and handwriting.
- 7Review a colleague's completed task from start to finish, and take on second-expert review as your experience grows.
Requirements5
- 1Native Bengali fluency with full command of Bengali script, diacritics, and conjunct forms.
- 2Hands-on experience with documents in annotation, transcription, translation, localization, subtitling, proofreading, journalism, or regional-language data review.
- 3Character-level accuracy, because one wrong diacritic counts as a defect.
- 4A systematic approach to applying a taxonomy across hundreds of pages instead of improvising per document.
- 5Comfort with unfamiliar layouts: multi-column newspapers, exam papers, and handwritten forms.
Who Should Apply
Native Bengali speakers with document work in annotation, transcription, localization, subtitling, proofreading, or regional-language data review will fit. The role suits people who catch diacritic-level errors and apply a taxonomy across hundreds of pages without drift. Newspapers, exam papers, handwritten forms, and multi-column layouts will appear often.
Salary Insight
The role pays $10.00 to $11.00 per hour. That range fits entry-level annotation work, and the project values character-level accuracy over output speed. No other compensation details appear in the listing.
Pay and demand for AI Training roles
AggregatedTypical pay
$70/hour
This role
$10–$11/hr
Most AI Training roles pay $34–$95 per hour. This role's pay sits below that range.
Based on 999 similar roles that publish pay · 301 publish only a top rate; those count at the rate they gave
Rates shown per hour. Yearly and monthly pay converted; one-time fees and non-USD pay are not included.
- Live similar roles
- 1,138
- Listed in last 30 days
- 394
- Remote
- 99%
Hiring most right now: micro1 (308) · SME Careers (253) · AfterQuery (135)
Most requested skills · share of roles
- ai training14%
- ai evaluation12%
- llm evaluation12%
- data annotation10%
Figures from AI Training roles live on NearSkill when this page loaded. A role can close before you apply, so check the listing itself.
Compare your resume against these rolesLocation
Compensation
$10–11/hr
Required Skills
Application Tip
Lead with native Bengali fluency and full command of the script, diacritics, and conjunct forms. Add concrete examples of PDF annotation, transcription, OCR correction, or bilingual QA work, plus numbers such as pages completed, pass rate on second-expert review, or error rate. Mention any experience with newspapers, exam papers, handwritten forms, tables, or multi-column layouts.
How NearSkill helps you apply
1. Check your job match
We compare your resume with the job requirements to help you understand how well your experience fits.
2. Continue to the employer’s website
We’ll direct you to the employer’s application portal, where you can complete the application.
3. Get updates by email
If NearSkill is handling follow-up for this role, we’ll email you when there’s an update on your screening.
Want to check your fit first?
See how your resume matches this role before applying.
Check my job matchSimilar open positions
Explore active roles that match your skills and interests.
Mercor
VerifiedPDF Annotation & Transcription Experts – Malayalam
Neon builds training data for document-understanding models that handle Malayalam and other low-resource scripts poorly. You will work through real published PDF pages, from newspapers and exam papers to forms, menus and worksheets, and turn each one into a complete structural map paired with a faithful transcription in Malayalam script. People do all of the work here: region identification, component typing, reading order and transcription. Handwriting, dense multi-column layouts, tables and mixed-script pages make up much of the corpus, so expect careful editorial judgment rather than bulk labeling.
Mercor
VerifiedPDF Annotation & Transcription Experts – Odia
Neon builds training data for document AI in the scripts that parsing and vision-language models handle worst. Odiya sits at the center of this project, next to four other Indic scripts plus Japanese and Korean. You would take real published Odiya PDFs, drawn from newspapers, textbooks, exam papers, flyers, forms, manuals, menus and worksheets, and map every meaningful region on each page before transcribing the text exactly as it appears in Odiya script. All component identification, typing, reading order and transcription come from people rather than models. A second Odiya expert reviews each task end to end, and the pages chosen lean toward the hard cases: handwriting, tight multi-column layouts, tables, diagrams and mixed scripts.
Mercor
VerifiedPDF Annotation & Transcription Experts – Gujarati
Neon needs native Gujarati speakers to build document AI training data in scripts that parsers and vision-language models handle badly. Each assignment hands you a real, publicly available PDF page drawn from newspapers, textbooks, exam papers, flyers, forms, manuals, menus, and worksheets. You map every meaningful region on that page, tag it by type, set the reading order, and transcribe the text in Gujarati script. No parser output stands in for your work: people handle component identification, typing, ordering, and transcription, and a second Gujarati expert checks each task before it ships.
Mercor
VerifiedPDF Annotation & Transcription Experts – Korean
Neon needs Korean speakers to build training data for document AI that struggles with Korean PDFs. You will work from real published pages, not synthetic samples, and map each region, transcribe text in Hangul, and flag anything illegible. The project centers on the layouts parsing and vision-language models handle worst: handwriting, dense columns, tables, diagrams, and mixed-script pages. Documents come from newspapers, textbooks, exams, flyers, forms, manuals, menus, brochures, notices, and worksheets. Every task receives review from a second Korean expert.
Mercor
VerifiedPDF Annotation & Transcription Experts – Japanese
Neon builds human-authored training data so document AI can read the languages that parsing and vision models handle worst. This project covers Japanese alongside Korean and five Indic scripts, and each task maps the structure of one public PDF page while transcribing its text in the original script. The corpus leans on the material models struggle with: handwriting, vertical text, dense multi-column layouts, tables, diagrams, and pages that mix scripts. Documents come from newspapers, textbooks, exams, and everyday items such as flyers, forms, manuals, menus, and worksheets. Component identification, typing, reading order, and transcription all come from people, never from a model.
Mercor
VerifiedPDF Annotation & Transcription Experts – Telugu
Neon seeks Telugu-fluent annotators to build training data for document AI in scripts that vision-language models handle worst. You will work from real, published PDF pages, mapping each page's structure and transcribing every text region in Telugu script. The corpus covers newspapers, textbooks, exams, flyers, forms, manuals, menus, brochures, notices, and worksheets, with a focus on handwriting, dense columns, tables, diagrams, and mixed-script layouts. All component identification, typing, reading-order assignment, and transcription come from human annotators. A second Telugu expert reviews every task from start to finish.