Would it make sense to add LSA/LDA vec to TF IDF representation?

[pommedeterresautee]

I was wondering if using both a topic vector (LSA/LDA based, or even paragraph2vec...) plus tf idf would improve results.
Topic vector based score would be added to tf idf based score with a low weight so common words (with high tfidf weight) are very important, but topic would be taken into account to probably affect document order.

What do you think?