Leyu Platform

Leyu is an open source crowdsourcing platform that allows you to collect high-quality data from users in a scalable and efficient way.

The platform:

Can collect multi-modal data (audio, text) from contributors via a Flutter mobile app
Provides a web dashboard to manage projects, tasks, users, and track progress with advanced reporting
Supports multiple languages for both the web dashboard and mobile app
Enables the creation of complex data collection workflows with distribution control
Validates submitted data automatically using AI-assisted tools or manually by reviewers
Supports automated or manual reward distribution to contributors

How Does the Platform Work?

Leyu consists of two major components:

The Web Dashboard – The desktop component to create projects, assign tasks, manage users, review submissions, generate reports, and manage rewards.
The Mobile App – The cross-platform Flutter app that contributors use to collect data and submit tasks.

Example: Speech-to-Text Task

Let's take an example of a speech-to-text task from English to Amharic to see how Leyu handles it end-to-end.

Scenario: A company wants to collect audio data from contributors and transcribe it into Amharic.

Key Components

The following components work together to create this workflow:

Dataset → Task → Mobile App → Output Dataset

┌───────────────────────┐                                 ┌───────────────────────┐
│    Input Dataset      │                                 │    Output Dataset     │
│   (raw audio clips)   │                                 │ (transcribed texts)   │
├───────────────────────┤                                 ├───────────────────────┤
│ 1. Audio Clip A       │         ┌──────────┐            │ 1. EN: Audio Clip A   │
│ 2. Audio Clip B       │◄───────►│   Task   │◄──────────►│    AM: Transcription │
│ 3. Audio Clip C       │         │          │            │ 2. EN: Audio Clip B   │
│                       │         └────┬─────┘            │    AM: Transcription │
│                       │              ▲                  │ 3. EN: Audio Clip C   │
│                       │              │                  │    AM: Awaiting...   │ ← waiting for contributor
│                       │              ▼                  │                       │
└───────────────────────┘         ┌────┴─────┐            └───────────────────────┘
  ↑ each item = 1 datapoint       │   App    │              ↑ each item = 1 datapoint
                                  │          │
                                  └──────────┘

What is a Dataset?

A dataset is a collection of individual data points to be processed. It can contain:

Raw audio clips or text entries to be collected from contributors
Completed submissions that have been reviewed and validated

What is a Task?

A task defines the action required for each data point. It:

Takes input from one dataset (e.g., raw audio)
Guides contributors to process it (e.g., transcribe or label it)
Produces output in another dataset (e.g., validated transcription)

Example: A speech-to-text task instructs the contributor to listen to an audio clip and provide a text transcription.

How Does the Mobile App Work?

The mobile app presents tasks one at a time in a chat-like interface:

Contributors see a single audio clip or text item
They complete the task (e.g., record audio or transcribe text)
Move on to the next task until all items are complete

What This Looks Like in Practice

Step 1: Contributor performs the task
The contributor listens to the audio clip and types the transcription.

Step 2: Contributor submits completed tasks
Once all clips are processed, the contributor reviews and submits the batch.

Summary

Leyu combines web and mobile components to create a seamless data collection workflow:

Web Dashboard – Manage projects, tasks, users, and rewards
Mobile App – Collect multi-modal data efficiently, even offline
Scalable Datasets & Tasks – Organize, distribute, and process data efficiently

How Does the Platform Work?​

Example: Speech-to-Text Task​

Key Components​