The “suggestion engine“ is a component that by default runs in the background, constantly updating a model based on your annotations in the documents. The objective is to speed up annotations by offering suggestions to the user that they then only need to accept or reject rather than adding annotations manually.
By default a solid and fast algorithm is used for this job, but it is possible to finetune and build your own suggestion model.
- Go to the Settings | Suggestions view
- The default suggestion engines are:
- CRF-Suite engine for Entity detection project (name=Sherpa)
- Sklearn for Text classification project (name=Sklearn)
It is possible to duplicate the engine, modify its parameters or create a new one to address specific needs.
- Create a new Suggestion producer
- Select CRFSuite
- Define parameters
- Save and launch the suggestion producer
- Go to the Suggestions view when job is completed
It is also possible to produce suggestions with Wikidata:
- Create a new suggestion producer with the Wikidata EntityFishing engine
- Define parameters
- Save and launch the suggestion producer
- Go to the Suggestions view when job is completed.
- See also: