Hindi English AI Dataset
Budget / SalaryHourly project
TypeFreelance project
LocationRemote
Posted1 hour ago
I’m putting together a parallel Hindi-English dataset that can be fed straight into an AI language model. What I need from you is a clean, well-aligned corpus that blends creative writing (short stories, poems, lively dialogues) with everyday conversational snippets, yet keeps the Hindi side in straightforward, commonly spoken language.
Scope
• Collect or craft sentence-level pairs in both directions (Hindi→English and English→Hindi).
• Keep roughly half the material imaginative and literary, the other half drawn from daily life situations—greetings, questions, small talk, etc.
• Use simple, standard Hindi vocabulary and syntax; avoid highly formal or regional expressions.
• Ensure every Hindi line has a precise, natural English counterpart and vice-versa, with no machine-translation artifacts.
Deliverables
1. A CSV or TSV file with three columns: ID, Hindi, English.
2. At least 10,000 unique sentence pairs (feel free to propose a larger number if you already have the resources).
3. A short README noting your data sources, any preprocessing steps, and basic statistics (total pairs, average token length, genre split).
Acceptance Criteria
• Alignment accuracy above 98 % on a random 500-pair spot check.
• No duplicate pairs.
• Text is free of offensive or copyrighted material that cannot be openly shared.
Feel free to rely on CAT tools, Python scripts, or manual curation—whatever ensures quality. Please outline your approach and estimated turnaround in your proposal.
Scope
• Collect or craft sentence-level pairs in both directions (Hindi→English and English→Hindi).
• Keep roughly half the material imaginative and literary, the other half drawn from daily life situations—greetings, questions, small talk, etc.
• Use simple, standard Hindi vocabulary and syntax; avoid highly formal or regional expressions.
• Ensure every Hindi line has a precise, natural English counterpart and vice-versa, with no machine-translation artifacts.
Deliverables
1. A CSV or TSV file with three columns: ID, Hindi, English.
2. At least 10,000 unique sentence pairs (feel free to propose a larger number if you already have the resources).
3. A short README noting your data sources, any preprocessing steps, and basic statistics (total pairs, average token length, genre split).
Acceptance Criteria
• Alignment accuracy above 98 % on a random 500-pair spot check.
• No duplicate pairs.
• Text is free of offensive or copyrighted material that cannot be openly shared.
Feel free to rely on CAT tools, Python scripts, or manual curation—whatever ensures quality. Please outline your approach and estimated turnaround in your proposal.
Apply on Freelancer →
Project sourced from Freelancer.com. Applications happen directly on the original platform — we never collect your data.