You can run one pre-annotator on documents only. If you run one pre-annotator, and then run a second pre-annotator, the second run will strip the annotations that were added by the first pre-annotator from the documents. Pick the pre-annotation method that best fits your use case, and use that one only.
This is huge limitation because in most production use Watson Knowledge Studio the entity will need to leverage ALL pre annotation techniques. So entity 'Person Name' is best handled by Natural Language Understanding while entity 'Appliance' is best handled by Dictionary. The true ease of use will happen when I can set both and then pre-annotate using both Dictionary and Natural Language Understanding.
Allow combining multiple pre-annotators of any kind (Natural Language Understanding, Dictionary, Rules, or Custom Model) to the same set of documents without wiping out previous pre-annotations or human annotations. (NOTE: Would require adjudication in case of annotation conflicts.)
Different designs possible. E.g., running multiple pre-annotators simultaneously (one run) or sequentially (one after the other).
Now only 1 type of pre-annotating can be used (Dictionary, Rules or NLU). It limits the pre-annotating volume and coverage while several entities should be annotated via dictionaries and rules at the same time.
It's possible also via merging Rules and Dictionaries in some way - i.e. the ability to add terms-based and rules-based pre-annotating techniques to the given entity.
Why is it useful?
|Who would benefit from this IDEA?||Faster and complete pre-annotating|
How should it work?