What is the algorithm for searching users in the social. networks by keywords?

Question

Algorithms

Please advise what approaches can be used to search for users in social networks (facebook, linkedin) on request. The query represents a failure of one or more keywords. Informally, I want to find users with the interests specified in the request. At the same time, I want the search to be non-boolean, i.e. even those users who did not have an exact match, but had "close" terms, got into the result.
What can be done here? I have such ideas:
- model topics through LSA, then map to a new user space and the query itself
- somehow use collaborative filtering (how?)
- use boolean search, expanding the search query
The problem is complicated by the fact that there is not much data, and they are not marked. But the latter can be done.
Maybe someone faced similar, please share.

Answer 1

I'm answering myself if anyone is interested.
Since there is little data, and there are no marked ones at all, I made a feint with my ears. I took the word2vec utility from Google developers and set it on a piece of Wikipedia. After that, I mapped the profiles and queries into a new space and applied the cosine distance. It turned out tolerably, although with anomalies.