Table of Contents
Getting Started
Guidance Lab is power tools for creating data labeling projects for benchmarking and training machine learning models.
The main idea is your likely have a large number of samples to go through. we want to make it easy on the hands to go from sample to sample. Step by step:
- Create a project. All you need is a name.
- Upload your samples. You can add these in a variety of ways. See Samples for details on format.
- If you have trusted annotators, add them to the project by their email address. They will receive an email with a link to sign in.
- Click "Start Labeling" to begin. This interface works on desktop and mobile devices.
- When you have another labels, you can export the results to CSV or OpenAI JSONL format.
Guidance Lab is a living project. If you would like anything in particular, contact me; we are easy to reach and eager to help.
Samples
Sample Types
Each sample added to a project is stored with its type, which is used to determine how to display it in the labeling interface.The following sample types are supported:
- text/plain
- text/html
- text/markdown
- text/json
Importing Samples
Import samples by file upload on the project detail page. Let's have a look.
File upload types are not one-to-one with sample types; uploads are intended to bundle samples together.
Plain Text (Lines)
Sample type: text/plain
If you have a text file, you can upload it and Label Studio will label each line as a separate sample.
CSV (Two-Column)
Sample type: text/plain
If you have a CSV file with two columns, the first column will be used as the sample text and the second column will be used as the label.
Redundant samples with different labels are allowed; each label will be recorded into the system under a sample together.
Zipped Sample Sets
Upload type: application/zip
Sample type: text/plain; text/html; text/markdown; text/json
If you have a zip file containing supported files, each file will be imported as a separate sample.
For example, a zip containing several HTML files will be imported as a set of HTML samples.
JSONL
Upload type: application/jsonl
Sample type: application/json
JSONL is a JSON Lines format; each line is a separate JSON object. Each JSON object will be imported as a separate sample.
Changelog
A detailed history of updates, improvements, and fixes to Team Label Studio.
JSON View
Expanded import options.
- Feature: JSON samples render with pretty-print.
- Feature: File import accepts JSONL files as sets of JSON samples.
Export Options
We've expanded the export options.
- Feature: OpenAI JSONL format
- Feature: Export only samples with confirmed labels
UI Improvements
The labeling interface now supports mobile devices.
Language Settings Update
We've added new language and translation features to make Team Label Studio more accessible to international teams.
What's New
- Feature: Configurable user interface language preferences
- Feature: Support for English, Russian, Arabic, and Chinese
- Improvement: Language settings accessible from both project list and project detail pages
These updates make Team Label Studio more accessible for international teams collaborating on multilingual datasets.
General Availability Release
Today marks the official public release of Team Label Studio! We're happy to make our collaborative data labeling platform available to everyone.
What's New
- Feature: Team collaboration with real-time project updates
- Feature: Streamlined label assignment workflow
- Feature: Comprehensive project analytics and reporting
- Feature: Email-based authentication system
- Feature: Quality control with label verification
- Improvement: Optimized sample assignment based on workload distribution
Contact
Guidance Lab is a project of Joseph Cooper Consulting. If you have any questions, please email me.