ሰው

ሰው helps detect potentially inflammatory or hate-related terms while users type in web text
fields.

It uses a curated lexicon based on CIR’s inflammatory keyword research for Ethiopia-focused
online language contexts (including English, Amharic, Afaan Oromo, and Tigrigna terms). When
a flagged term appears, the extension highlights it and provides a short reason so users can
revise wording before posting.

Key features:
- Real-time detection while typing
- On-page visual highlights and threat summary
- Per-site enable/disable controls
- Local history and page-level stats
- Optional “Suggest a Term” form for missing entries

Privacy:
- Text analysis is performed locally in the browser
- Typed page content is not sent to our backend for analysis
- Only data entered in the optional Suggest form is sent when the user submits it