Doccano tutorial: Docker setup, NER and text classification

This tutorial covers a self-hosted Doccano workflow for preparing NLP training data. Uploaded on May 5, 2021, it shows the interface and installation approach used at that time; the supplied material names no Doccano version. The speaker describes pip installation as experimental and recommends Docker for this demonstration. Those installation comments belong to the 2021 tutorial.

The setup walkthrough uses Docker Desktop for Windows, an administrator command prompt and a browser connection to localhost on port 8000. The speaker explains pulling the container image and setting login credentials, but skips the full installation process because of the time it takes.

The main example creates a sequence-labeling project for named entity recognition. The presenter imports a text file, defines person, location, miscellaneous and designation labels, assigns shortcut keys and colors, then selects text spans to annotate. The demonstrated NER export offers JSONL, and the presenter explains how the saved labels record character positions.

A second example creates a text-classification project with positive, negative and neutral sentiment labels. It follows the import, annotation and export workflow and inspects the saved data. Project members, annotation guidelines and statistics also receive a brief walkthrough. Sequence-to-sequence tasks are mentioned as a supported project type, but the tutorial does not demonstrate that workflow.