I am annotating a corpus of Dutch text. In Dutch words are often joined together, for example ‘contrastnefropathie’ rather than ‘contrast nephropathy’ as you would see in English. In MedCATTrainer if you try to annotate a span that is immediately adjacent to another span (for example, ‘contrast’ and then ‘nefropathie’), the annotation on the first span will disappear. Has anyone else experienced this?
Oh that is potentially an issue. This will always occur unless there is at least a whitespace separation (MedCATtrainer does not support subword annotations).
I haven’t come across many of these types of examples in English. In our case a concept is always one word or longer in length.
To overcome this in the meantime. Have you thought about creating and adding a single custom concept called “contrastnefropathie” rather than annotating “contrast” and “nefropathie”?