Hi. Does the Relcat feature exist today. Can one create relation annotation models. The documentation talks about it as a ‘coming soon’ feature.
Hi @sangeetabose that is correct.
Glad that you are following the repository closely! However we have not released the training methodology/pipeline yet.
You may have noticed that you can annotate relationships on MCTrainer. This annotation task can be set via the following:
Task | Description |
---|---|
Relations | Select from the list of ‘Relation Annotation’ tasks that will appear for a given concept. |
Do you have a particular use case in mind? The majority in my experience do not actually require relationship extraction.
I was hoping to use relations for
1> key value pairs
Blood Pressure: 180/90 - (relation between Blood Pressure and ‘180/90’)
2> temporal statements
Patient underwent surgery in 2012 - (relation between surgery and 2012)
If I do not need to use relations for such extractions, could you please suggest ways to do this.
@sangeetabose This is a use case that requires relational extraction.
You can either:
- Create a window around a concepts and use REGEX to extract relevant patterns.
- Wait for the MedCAT tutorials to come out on relational extraction.
There is no exact date unfortunately. There are other tutorials which are prioritised over this one. Deidentification, Large language models etc.