Project Background(Toxic Words Labeling)
This project focuses on cleaning and annotating in-game voice content, with an emphasis on identifying sensitive language such as violent expressions and profanity. The resulting training data will help better distinguish ambiguous content and support overall content safety and user experience improvement.
Responsibilities
1. Download the audio to your local computer through the audio link for playback.
2. Determine whether the audio contains toxic words. If so, mark "Toxic" in the "Is toxic" column of the table. If the audio does not contain toxic words, mark "Normal"
3. For words containing toxic, mark each time point of the toxic word in the "Note" column (if a word is repeated multiple times, the time point of each occurrence must be marked) and the toxic word, and its corresponding English translation
4. List the 20 English translations of the most frequently occurring dirty words (it is not necessary to collect 20, as many as possible)
Requirements
1. Native Portuguese based in Brazil, fluent in both spoken and written English (minimum B2 English level)
2. Available 30-40 hours per week with relatively fixed daily schedule
3. Up-to-date PC (Windows 10 or later) or Mac (Big Sur or later) with stable utility and internet connection
4. Degree and/or working experience in literature, translation, linguistics, teaching is a plus
5. Similar project experience in data collection, annotation, quality control, coordination is a plus
Training & Payment
1. Each participant will attend approximately 1 hour of training
2. Actual work will be paid based on completed cases.
3. Estimated time per case is about 1 minute, based on internal testing.
Additional Notes
1. The project is expected to run for approximately 3–4 weeks. Please ensure that you are able to remain available for the full project duration and meet the output targets as defined by the Project Manager.
2. Once you confirm participation in this project, early withdrawal during the project period is not permitted