I think what you're looking for is described in this paper http://jair.org/vol/vol34.html. Basically a single LDA like Model generates text as well as document level tags. The model can also Auto suggest tags given a document and the topics are sensitive to tags in an existing corpus. The software is also open sourced.

I think what you're looking for is described in Learning Document-Level Semantic Properties from Free-text Annotations. Basically this paper http://jair.org/vol/vol34.html. Basically a single LDA like Model is an LDA-variant which jointly generates text as well as document level tags. The up-shot is that when you do topic modeling documents with similar tags get similar topics and that when you have text by itself, the model can also Auto suggest predicts tags given a document and with strong accuracy. The code for the topics are sensitive to tags in an existing corpus. The software project is also open sourced. available here.