Contextual document spans

In contextual top2vec there are multiple document vectors per document. Is there a way to recover which documents these vectors represent? Or what document span (snippet of text) corresponds to each document vector?

