Email Inquiry Annotation Project
Budget / SalaryHourly project
TypeFreelance project
LocationRemote
Posted2 hours ago
I have a large batch of CRM-exported emails and I need every customer inquiry inside them clearly annotated. The corpus is plain text; no images or audio are involved. For each email, I want the freelancer to mark the exact sentence(s) or phrase(s) where the customer is asking something—nothing else for now. (Later we may look at sentiments or topic labels, but this phase is inquiry-only.)
Scope
• Roughly 12,000 email threads supplied as UTF-8 text or CSV.
• You will follow a short guideline I provide that defines what qualifies as an inquiry, how to tag multi-sentence questions, and how to deal with quoted replies.
• Preferred tools are Label Studio, Prodigy, or a custom Python script that outputs start–end character offsets plus the inquiry text; I’m open to alternatives if the final export is a clean CSV/JSON with email_id, span_start, span_end, inquiry_text.
Acceptance Criteria
1. ≥95 % agreement with a 500-email gold-standard sample I already annotated.
2. No data leakage: all files stay on the private, access-controlled workspace I set up.
3. Deliverables include the annotated dataset, the project-level config file/tool preset, and a two-paragraph summary of edge cases you encountered.
Timeline is flexible, though I’d like the first 1,000 emails annotated within one week so we can align on quality before you proceed with the full set.
Scope
• Roughly 12,000 email threads supplied as UTF-8 text or CSV.
• You will follow a short guideline I provide that defines what qualifies as an inquiry, how to tag multi-sentence questions, and how to deal with quoted replies.
• Preferred tools are Label Studio, Prodigy, or a custom Python script that outputs start–end character offsets plus the inquiry text; I’m open to alternatives if the final export is a clean CSV/JSON with email_id, span_start, span_end, inquiry_text.
Acceptance Criteria
1. ≥95 % agreement with a 500-email gold-standard sample I already annotated.
2. No data leakage: all files stay on the private, access-controlled workspace I set up.
3. Deliverables include the annotated dataset, the project-level config file/tool preset, and a two-paragraph summary of edge cases you encountered.
Timeline is flexible, though I’d like the first 1,000 emails annotated within one week so we can align on quality before you proceed with the full set.
Apply on Freelancer →
Project sourced from Freelancer.com. Applications happen directly on the original platform — we never collect your data.