SMART User Docs

SMART is an open source application designed to help data scientists and research teams efficiently build labeled training datasets for supervised machine learning tasks.

Feature Highlights

• Active Learning algorithms for selecting the next batch of data to label.
• Inter-rater reliability metrics to help determine a human-level baseline and the understand the test validity of your labeling task.
• Admin dashboard and other project management tools to help oversee the labeling process and coder progress.
• Multi-user coding, for parallel annotation efforts within a project.
• Self-hosted installation, to keep sensitive data secure within your organization’s firewall.

Quick Start

$ git clone [github repo]
$ cd smart/envs/dev/
$ docker-compose build
$ docker volume create --name=vol_smart_pgdata
$ docker volume create --name=vol_smart_data
$ docker-compose up -d

Open your browser to http://localhost:8000

Tutorial

• Part 1: Installation
• Part 2: Creating a New Project
• Part 3: Reviewing Projects & Editing Project Settings
• Part 4: Annotating Data
• Part 5: Administrator Dashboard
• Part 6: Downloading Labeled Data and/or Model

Advanced Features

• Advanced Feature Details
  • Active Learning
  • Inter-rater Reliability (IRR)
  • Fix Skewed Label Distributions

About SMART

• Frequently Asked Questions (FAQs)
  • General Questions
  • Technical Questions
• Release Notes and Change Log
  • Release v.0.0.1
  • Contributors
• License