# Introduction

> **LINGULIA** is a standalone, offline-first **mental-health language platform** for Ethiopian
> health workers, in **English, Amharic, and Afaan Oromo**.

## Why LINGULIA exists

Mental health care in Ethiopia is delivered largely by non-specialists at the lower tiers of the
health system — Health Extension Workers (HEWs) and health-center nurses and officers. Two gaps
make their job harder:

1. **Language.** Clinical knowledge, screening tools, and training materials are overwhelmingly
   in English. The vocabulary of mental health in Amharic and Afaan Oromo is fragmented, and
   *local idioms of distress* ("thinking too much", `buda`, `zar`) are central to how patients
   actually describe suffering.
2. **Connectivity.** Rural health posts cannot rely on the internet. A tool that needs a live
   connection is a tool that is not there when it is needed.

LINGULIA addresses both: a carefully built **trilingual mental-health vocabulary**, screening
tools, a clinician–patient phrasebook, and danger-sign guidance — all working **100% offline**,
installed directly on a phone.

## Why mental health first

Rather than building shallow coverage across every medical domain, LINGULIA v1 is **deep and
focused on mental health**. This is a deliberate choice:

- It is high-need and under-served at the primary-care level.
- It has a clean, internationally recognised clinical anchor — the **WHO mhGAP Intervention
  Guide** — already adopted by the Ethiopian Federal Ministry of Health for non-specialist care.
- Language and cultural nuance matter more in mental health than almost anywhere else in
  medicine, so it is exactly where a *language* platform adds the most value.

Other domains (maternal & child health, infectious disease, emergency care) follow later, reusing
the same architecture.

## What LINGULIA is (and is not)

- ✅ A **language and job-aid tool**: lookups, screening, phrases, referral guidance.
- ✅ **Standalone** — it does not depend on any third party. It works with no internet.
- ✅ **Installable** to the phone home screen as a Progressive Web App (PWA).
- ❌ **Not a diagnostic device.** Screening scores are not diagnoses. It supports — never
  replaces — a trained clinician's judgement.

## Building blocks (today, and the roadmap)

LINGULIA is organised the way a language-AI platform is, so it can grow from a content app into a
full product:

| Layer | v1 (now) | Later |
|------|----------|-------|
| **Vocabulary** | ~90+ curated trilingual mental-health terms incl. local idioms | thousands of reviewed terms across domains |
| **Tools** | PHQ-9, GAD-7 screening; phrasebook; danger signs | more validated instruments, structured assessments |
| **Voice** | on-device TTS (where a voice exists) | high-quality neural Amharic/Afaan Oromo TTS & STT |
| **AI** | optional, pluggable chat (off by default) | LINGULIA's own fine-tuned medical model + hosted API |
| **Languages** | English · Amharic · Afaan Oromo | Tigrinya, Somali, Sidaama, … |

## How these docs are organised

- **[Get Started](get-started.md)** — install on a phone, run locally, use offline, configure the
  optional AI.
- **[Architecture](architecture.md)** — how the offline PWA and the pluggable AI layer fit together.
- **[Data Schema](data-schema.md)** — the JSON formats and how to add or correct terms.
- **[Clinical Content](clinical-content.md)** — sources, the review process, confidence flags, and
  limitations. **Read this before any field use.**
- **[Roadmap](roadmap.md)** — where LINGULIA goes next.
