Here is a query for just those tags that aren’t currently associated with topics.

prefix tag: \
prefix rdfs: \
select ?label
where {
?tagging tag:associatedTag ?tag .
?tag rdfs:label ?label .
optional { ?tag rdfs:isDefinedBy ?topic } .
filter (!bound(?topic))
}

each tag appears as many times as its been used so far. Like I said, aggregate functions are on the way. (This takes several minutes to run on the Joseki server, but is quite snappy on our current toy Virtuoso setup) with a bit less data (use ‘sandbox’ for Default Graph URI ). Notice that you can get json objects out if you ask for them.

Currently the (non-standard) aggregate operations, like count, that are theoretically supported by our RDF server aren’t working, but the following will get all the data needed to count them up in a client. We create a distinct URI for each user/tag combination and then associate that with a label. This is done to account for different users using the same string in different ways (one of the reasons we are interested in freebase!)

prefix tag:
prefix rdfs:
select ?label ?topic
where {
?tagging tag:associatedTag ?tag .
optional { ?tag rdfs:isDefinedBy ?topic } .
?tag rdfs:label ?label
}

Its quite slow today, but by the end of the week will be several orders of magnitude faster thanks to a new server and a new RDF repository (Virtuoso).

I’m no SPARQL expert but I tried playing around with this query

prefix tag: <http://bioinfo.icapture.ubc.ca/Resources/SemanticTagging/>
prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#>
select ?tagging ?tag ?topic ?user
where {
?tagging tag:associatedTag ?tag .
optional { ?tag rdfs:isDefinedBy ?topic } .
?tagging tag:taggedBy ?user .
filter (isBlank(?topic))
}

I hit two problems: isBlank() doesn’t work as I expected and ?tag is in the form username_tag so I can’t sort and then count tags from different users.

How would you like the commonly used non-freebase topic tag stream represented? It would be pretty easy to pull them out. In fact, this could be done using the SPARQL endpoint with very little work (feel free to play). I’ll give you a start, I use “ontology evaluation” quite a bit, but haven’t bothered to create the topic yet. We’re going to be moving the endpoint to a different server this week so it should get much faster.

Small problem: the bookmarklets on this page are broken.

The widget idea for adding new tags is a good idea but would require careful implementation to avoid flooding Freebase with topics which are too lightweight.

It would be really interesting to get a feed of commonly used tags which aren’t in Freebase.

The tags are stored in Connotea as normal strings, but they are also stored in our own RDF database (soon to be publicly reachable as a SPARQL endpoint – maybe today). ( We aren’t Connotea by the way, we are a grad student and a programmer at the university of british columbia). In the RDF database we store the links between the URIs for the tags in Connotea, the Freebase URIs and the Freebase Types. Hopefully we’ll be using this data to produce a more semantic experience when browsing tagged collections.

I’d also like to use freebase (or some version thereof) as the tagging database… how about it? Want to compete with ZigTag?

Are you still storing the tags as freetext or are you associating them with a Freebase id?

We went ahead and conducted your exercise for the Connotea social bookmarking service. Let us know what you think!
http://www.entitydescriber.org