I am trying build a domain specific question answering system using the Wikidata that already has the rdf/json dump format.
There are pros and cons between triplestore (e.g. virtuoso) and graph (e.g. neo4j), I made a lot of reading but still can't decide which.
People said triplestore is good for inferencing/reasoning, while graph is not but I thought with graph in neo4j we can write Cypher in such a way that can do inferencing too such as this , so what I am not aware of ? Is it easier to query with SPARQL than Cypher for inferencing ?
I am leaning towards using graph database as it has a couple of advantages as explained here but maybe it has bias because it's written by neo4j. But if I choose this route, means I have to sanitize a lot of rdf data when importing to neo4j with the neosemantic plugin ?! I am not sure how direct it's and if it will be painful to import them.
I would prefer to have everything stored in a single db that include the user's login system, application specific data as well.
so what is more suitable in my situation, triplestore or graph ? Thanks