LSI - Latent Semantic Inferencing
A college at work pointed me at this interesting paper about getting meaning from unstructured data using LSI.
A college at work pointed me at this interesting paper about getting meaning from unstructured data using LSI.
Have been writing an rdfstore in python based on Steve Harris' excellent 3store. It is designed to handle IFPs efficiently, by replacing all the hashes in 3store with ids. These then link to a URI table enabling a 1-many resource to URI mapping.
I got married on 14th August to the lovely Claire Beckett. My mum has put some photos of the happy event on her website.
FOAF and DOAP both use IFPs instead of URIs to identify resources. Have been considering how to do this in veudas.
Am currently thinking will provide a new version of the resource view which uses IFP prop/object pairs to identify the resource. Also will look for IFPs in bnodes objects when displaying the resource. If they exist, the view display will be a link to the bnode resource, otherwise bnode is displayed inline with the resource (as it is currently).
Have been playing with the 3store db format and have written an experimental rdfsql library in python. I did this because I'm not very hot with c and can knock things up much more quickly in python. I've implemented optional queries and a remove statement feature (doesnt handle inferred triples), which I needed for my veudas browser.
Released version 0.5.2 of veudas
The RAP rdf netapi is about to be released (Chris is putting the finishing touches to the RAP 0.8). My work on this api has taught me that I like working with python a lot more than I like working with PHP.
With that in mind, I did some simple performance comparisons between a php page and a python cgi on phildawes.net.
Results of 4 runs are:
0.133100032806 Python 0.111065864563 PHP
0.154880046844 0.104358196259
0.142001867294 0.0995299816132
0.134169101715 0.0949969291687
PHP is consistently faster, sometimes as much as 0.06 sec faster. Personally that's not a big enough speed improvement to warrant me switching all my development to php. Having said that, I need to test DB speed from python cgi vs. pooled connections from php.
Veudas is currently all clientside javascript. Unfortunatly it is becoming difficult to manage due to deficiencies in the javascript environment.
Specifically:
No support for threading, no asynchronous primitives.
Poor tool support - no refactoring browser etc...
Poor support for RDF, especially RDF/XML
I am thinking of splitting veudas into a client/server system, using python cgi. I've experimented with PHP, and have gained a little experience with it while coding the RAP joseki net-api, but I much prefer python - specifically its brevity of syntax. I also know bicyclerepairmanback to front ;-)
Have spent my spare time in the last few days writing an implementation of the Joseki net-api for RAP.
An early demo is available here
Chris Bizer of (author of RAP) emailed me to ask if he can add it to RAP when its finished - praise indeed.
Spilt orange juice on my work laptop (powerbook g5) today. The keys have gone all tacky - it's put me in mind of the spectrum plus I had in my teens.