The labels here occupy more space, how to make it small and give more space to text that we annotate. Thank You
You seem to have a lot of labels and I'm wondering if the best solution is maybe not to worry about CSS, but to reduce the number of labels in the first place.
I can imagine that you could design a pipeline that takes a two-step approach that would require you to have less labels. For example, you might primarily worry about detecting dates. Once you have a pipeline that can detect dates you can build a step that uses this information to figure out if the date is a start/end date and if the date is about job experience or academic experience.
Does this make sense?
I think this makes sense to me. but how to get start with this?
You would start by just annotating dates, possibly with ner.correct because this would allow you to re-use a pretrained spaCy model that can detect dates. I might also suggest annotating the entity for academic institution as well as companies because these entities will be useful later.
After that you can run the train recipe to train a spaCy pipeline that can detect dates. This will train a system that's saved on disk that's hopefully pretty accurate at detecting dates in your domain.
From here you can build a custom spaCy component that can look at a sentence and when there are two dates, can assign the dates with a start and end entitiy. Furthermore, if the sentence also contains an academic institution or a company reference then you can also indicate if the date belongs to an "academic start date" or a coporate one.
Does this help?
1 Like