Swedish Sentiment Annotation Project

I have a dataset of Swedish-language social media posts that must be annotated for sentiment. Every line of text needs a clear label—positive, negative or neutral—so that I can train and benchmark a new classifier.

You will work inside my web-based annotation tool (Label Studio is already set up) and follow the short guideline document I supply. Native-level Swedish comprehension is essential because the posts contain slang, emojis and informal grammar.

Deliverables
• Completed annotations for all assigned posts, exported as CSV or JSON from the tool
• A brief log noting tricky examples or guideline ambiguities you encounter

Acceptance criteria
• At least 95 % inter-annotator agreement on the overlap set I provide for quality checks
• No empty or ambiguous labels in the final export
• File structure exactly matches the template so I can plug it straight into my pipeline

I will batch the work in tranches of roughly 1,000 posts; once each tranche passes the quality gate, I release the milestone and open the next set. If you are ready to start right away, let me know your estimated throughput per day so we can map the schedule together.