Sequence segmentation

Is there any plan to release sequence segmentation into classes for large text like 400 chars ?

for eg.

In single-label ZSL, each label can be uniquely represented
in a semantics space; in other words, the meaning
of a label and the relatedness between two different labels
are all fixed. In this paper, we refer to such semantics as
global semantics. To obtain a global semantic representation,
there are two approaches in general: manually converting
a label into a list of pre-defined attributes that can
characterize all possible labels in a specific domain [5],
and automatically learning a continuous semantic embedding
space from linguistic resources, e.g., semantic
embedding learning from Wikipedia leads to the wellknown
word2vec space.

Tag sub sequence into multiple classes

Could you explain a little further? It would be helpful to have a link to the publication — it’s hard to make anything of just that paragraph!