Native Japanese NLP Data Trainer
Budget: ₹100 – ₹400 INR
I need a native Japanese speaker with solid linguistic expertise to help me train a Natural Language Processing module. Your main responsibility will be to create, review, and refine Japanese language data so the model can understand real-world usage, nuances, and edge cases.
You will work inside a web-based annotation platform, tagging intent, correcting tokenisation errors, and suggesting natural rephrasings where the model falls short. Whenever you spot ambiguity, please add contextual notes in Japanese and English so the engineering team can trace issues quickly.
Deliverables must include:
• A clean, well-annotated corpus in UTF-8 CSV or JSON, ready for immediate ingestion
• A brief report that explains any linguistic conventions, slang, or regional expressions you introduced, plus guidance on handling polite versus casual forms
• Final verification checklist confirming at least 98 % accuracy across provided test prompts
If you already have experience with tools like Prodigy, Label Studio, or similar, let me know; otherwise I will provide a short walkthrough. Turnaround is flexible but I’d like to see the first batch of annotations within one week of project start so we can calibrate quality early.
You will work inside a web-based annotation platform, tagging intent, correcting tokenisation errors, and suggesting natural rephrasings where the model falls short. Whenever you spot ambiguity, please add contextual notes in Japanese and English so the engineering team can trace issues quickly.
Deliverables must include:
• A clean, well-annotated corpus in UTF-8 CSV or JSON, ready for immediate ingestion
• A brief report that explains any linguistic conventions, slang, or regional expressions you introduced, plus guidance on handling polite versus casual forms
• Final verification checklist confirming at least 98 % accuracy across provided test prompts
If you already have experience with tools like Prodigy, Label Studio, or similar, let me know; otherwise I will provide a short walkthrough. Turnaround is flexible but I’d like to see the first batch of annotations within one week of project start so we can calibrate quality early.