External Recommender for INCEpTION

[Tool] Run your own recommenders using Python

This tool is provided by its authors outside the Eclipse Foundation and is not part of the Eclipse INCEpTION project.

External recommenders are web services that provide annotation suggestions to INCEpTION. They let you integrate custom classifiers or pre-trained models.

The inception-external-recommender repository provides ariadne, a Python library for running and implementing external recommenders. It also comes with ready-to-use recommenders, for example based on spaCy.

To get started, see the tutorial. Installation instructions are in the repository.

Connecting a recommender to your project

1. Start the recommender service. Its README has an example that serves spaCy named entity and part-of-speech recommenders at URLs such as http://localhost:5000/spacy_ner.
2. In INCEpTION, open the project settings, choose Recommenders and click Create.
3. Choose the layer and feature, select the external recommender as the tool, enter the URL of the service and save.
4. Open a document on the annotation page. The suggestions appear after your next annotation action.

Further reading

• External recommenders in the User Guide, including HTTPS connections
• External recommender API in the Developer Guide, for implementing your own service in any language