RDF Graphs...

User 534 | 7/26/2014, 11:27:58 PM

Hi, Popular ontology data like dbpedia are available as N3 triples in rdf framework. Usually these are processed in graph databases like virtuoso and queried using sparql. With the speed advantage of GraphLab, can we load the rdf triples and do some text analytis? Text analytics could include all connections to a node, connections between nodes. Inheritance, reasoning etc could come later. The vuirtuoso graph database building usig triples was harder last time I tried.


User 3196 | 4/17/2016, 3:51:49 PM

Please Dato Team can you respond to Hsuvarna, it's very important for me also ?

User 3196 | 4/17/2016, 3:57:32 PM

It's possible to use RDFlib in Dato ? https://rdflib.readthedocs.org/en/stable/introtosparql.html

User 4 | 4/18/2016, 2:52:19 AM

Hi @mbaroudi, RDFlib appears to be a Python library, as is Dato's flagship product GraphLab Create. Since both are Python libraries I'm sure it is possible for them to coexist in the same Python process and with your own code you can use them together.

Although we have no special integration points specifically with RDFlib that I am aware of, it should work to glue the two modules together using Python code. To work with data in GraphLab Create, you need to get it into the native format (SArray, SFrame, or in this case, SGraph). I don't know of a faster way to create an SGraph than to simply add_vertices and add_edges -- see these pages:

https://dato.com/products/create/docs/generated/graphlab.SGraph.addvertices.html https://dato.com/products/create/docs/generated/graphlab.SGraph.addedges.html

Using RDFlib in combination with these methods you should be able to ingest data into an SGraph and then use the SGraph methods as well as Dato's full toolkits library (including our text analytics toolkit).