ph3Fluent Language Skills Required: /h3 pJapanese. Native fluency in Japanese, including full command of kanji, hiragana and katakana, is required for this position. All annotation and transcription work is performed in Japanese. /p h3Why This Role Exists /h3 pDocument understanding breaks down fastest in the languages that parsing and vision-language models rarely see. This project builds training data for exactly those languages: Japanese, alongside Korean and five Indic scripts. 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. /p pThe dataset deliberately concentrates on the material models handle worst: handwriting, dense multi-column layouts, vertical text, 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 Japanese documents rather than a narrow band of easily parsed ones. /p pDelivered work is human-authored throughout. Component identification, component typing, reading order and all transcription are performed by people, not generated by parsing models. /p h3What You'll Do /h3 ul lipSource documents: find a publicly accessible Japanese PDF in an assigned document type, containing at least one multimodal element (images, tables, diagrams, or handwriting), and record where you obtained it /p /li lipAnnotate 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 /p /li lipRecord relationships: link each region to the figure or table it belongs to through a parent component identifier /p /li lipTranscribe faithfully: reproduce all text exactly as it appears, including kanji, hiragana, katakana, furigana and handwritten content, flagging any region where the source is not legible /p /li lipCapture page metadata: language, document type, source, page dimensions, and flags for tables, formulas and handwriting /p /li lipReview a colleague's work: every task is reviewed end to end by a second Japanese expert, and experienced annotators take on that review /p /li /ul h3Who You Are /h3 ul lipYou are a native Japanese speaker with full command of kanji, hiragana and katakana, including furigana and variant character forms /p /li lipYou have professional experience in interpretation, journalism, transcription, translation, editorial work, or comparable document-intensive work /p /li lipYou are exact: character-level accuracy matters more here than speed, and a single wrong character is a defect /p /li lipYou are systematic:



you apply a taxonomy consistently across hundreds of pages rather than improvising per document /p /li lipYou are comfortable with unfamiliar layouts: vertical text, multi-column newspapers, exam papers, handwritten forms /p /li /ul h3Nice-to-Have Specialties /h3 ul lipAI training data: annotation, labeling, grading, or bilingual evaluation for training datasets /p /li lipTranscription and localization: MTPE, subtitling, bilingual QA, OCR correction or post-editing /p /li lipDocument production: typesetting, copy-editing, proofreading, or digitization of Japanese-language material /p /li lipScript and encoding: Unicode normalization, Japanese input methods, full-width and half-width forms, and kanji variant handling /p /li /ul h3What Success Looks Like /h3 ul lipEvery meaningful region on the page is captured, correctly bounded and correctly typed /p /li lipReading order reflects how the page is actually read, including vertical text and multi-column layouts /p /li lipTranscriptions match the source character for character, in Japanese script rather than romaji /p /li lipYour tasks pass second-expert review the first time /p /li lipThe documents you bring in add layout diversity rather than repeating templates already in the corpus /p /li /ul h3Why Join Mercor /h3 ul lipBuild the training data that makes document AI work in scripts it currently handles badly /p /li lipWork from real published Japanese documents rather than synthetic or templated pages /p /li lipQuality leads on this project: accuracy is the first measure, with handling time tracked alongside it /p /li /ul /p #J-18808-Ljbffr

📌 Japanese PDF Annotation Specialist (Milano)
🏢 Obsidian
📍 Milano