SPARK NLP 3.4.1Annotation Lab 2.7.0 includes Spark NLP 3.4.1 and Spark NLP for Healthcare. Model training is now significantly faster and issues related to Rule-based annotation have been solved. The Models Hub has increased the list of models and old incompatible models are now marked as “incompatible”. If there are any incompatible models downloaded on the machine, we recommend deleting them.
SPANISH, GERMAN MODELS IN MODELS HUBIn previous versions of the Annotation Lab, the Models Hub only offered English language models. But from version 2.7.0. models for two other languages are included as well, namely Spanish and German. It is possible to download or upload these models and use them for preannotation, in the same way as for English language models.
FLEXIBLE ANNOTATIONS FOR VISUAL NER PROJECTS
CHUNK ANNOTATIONThe chunk annotation feature added to Visual NER projects allows the annotation of several consecutive tokens as one chunk. It also supports multiple lines selection.
How to create multiple chunks?
To annotate a multi-token chunk follow the steps below:
- Activate the label to use
- Click on the first token
- Press the ‘a’ or shift key once [Do not keep on pressing the hotkey] or click on the ‘Select all’ button on the right-hand side.
- After releasing the hotkey select the last token on the chunk
This action will create one or several annotated regions which are linked by a Connected relation. The connected relations are organized in groups and they act as one annotation.
- Users can now select multiple tokens and annotate them together in Visual NER Projects.
- The label assigned to a connected group can be updated. This change will apply to all regions in the group.
- Edit the connected group: It is possible to remove one or several parts of the connected group and/or add new regions.
- Connected relations are independent of regular relations.
CONSTRAINTS FOR RELATION LABELINGWhile annotating projects with
Entities, defining constraints (the direction, the domain, the co-domain) of relations is important. Annotation Lab 2.7.0 offers a way to define such constraints by editing the Project Configuration. The Project Owner or Project Managers can specify which
Relationneeds to be bound to which
Labelsand in which direction. This will hide some
Relationsin Labeling Page for NER Labels which will simplify the annotation process and will avoid the creation of any incorrect relations in the scope of the project.
To define such constraint
allowedattribute to the
- L1>L2 means Relation can be created in the direction from Label L1 to Label L2, but not the other way around
- L1>L2 means Relation can be created in either direction between Label L1 to Label L2
- If the a
allowedattribute is not present in the tag, there is no such restriction
<Header value="Sample Project Configuration for Relations Annotation"/>
<Relation value="Was In" allowed="PERSON>LOC"/>
<Relation value="Has Function" allowed="LOC>EVENT,PERSON>MEDICINE"/>
<Relation value="Involved In" allowed="PERSON<>EVENT"/>
<Relation value="No Constraints"/>
<Labels name="label" toName="text">
<Text name="text" value="$text"/>
Security issues related to SQL Injection Vulnerability and Host Header Attack were fixed in this release.
GET & INSTALL IT HERE.
FULL FEATURE SET HERE.