span segmentation containing large labels and sublabels

I have almost 20 labels for span segmentation. I have labels for example as EXPERIENCE as it has more in numbers in a specific resume. Do i need to specify all EXPERIENCE as a single span or individual span having sentence as EXPERIENCE. Also EXPERIENCE has sub-labels as:


Do labeling a sentences as a Single EXPERIENCES affects other sub-labels? Please enlighten me more.
P.S: EXPERIENCE is taken as an example here. We have other labels and sub-labels as of EXPERIENCE. Thank you

Hi Kushal,

I might recommend taking a two-step approach. If you can first detect each paragraph that describes a single experience at an employer then it will be fairly easy to associate a date/adress in a second step.

Is that something you've considered? Is there a reason why it might not work well? In general I usually recommend splitting the problem into more manageable chunks because it helps keep the subproblems, as well as the annotation interfaces, simpler.

Came up with this problem after using two step approaches and training the annotated data.