Testing

I have been conducting some tests today on different types of datasets, and the first and most troubling thing that I have discovered is on large blocks of search text I have been getting server not responding timeout errors. At this point I dont know if this is a problem with some browser settings on my end or if this is a setting on my webhosts server that is cutting short a process that seems to be taking to long... hummm. One way around this for the industrious, is to install a server on your local machine and ask me for a copy of the source, you will then be able to run the heck out of some large files without the server timing out.

Also it looks like analysis on complicated datasets where primary concepts have many subject compliments there is a tendency for lots of self relation between sets of dependent attributes. This is showing an extremely large amount of intersection concepts. I think that this problem is mainly one of visualization but perhaps it would be beneficial to track polyline paths and display singular routes through the cloud. this may relieve the current sense of confusion when looking at large dataset displays.

Intersection and instructions

I have added an area off of the Pointsware.php page which has some instructions on how to use the software and a description of some of the ideas behind the project.
I also added some function to what gets visualized from the text, now there is a thin magenta line that traces back to any primary concept parents from the set of concepts that were found to be in common between the two query trees.

Also I have done away with the square grid and in its place have allowed the unique concepts to arrange in a cloud, this relieves the tendency for the query tree lines to stack up on top of one another.

I am thinking about maybe looking into a circular presentation of the results but perhaps I should leave it alone and think about the LSA, or writing an interface for the dataset that can run on an Android handheld...

Really what is needed now is for me to spend some time working with creating a lot of different result SVG images and do a little user testing to see where any problems might be showing up. Also I want to look at the different types of data that might be useful to visualize with this type of application, maybe make some test sets and example outputs.

My thought being there is useful information to be found in the network of connections around how we think about the thing we are focusing on.
I cant help but think about the effected range of industry that surrounds automotive manufacturing. There is a nodework of connected buisnesses which have been effected by the loss of jobs due to declining sales in automobiles. each of these buisnesses can be thought of as a line in the dataset where the primary concept is the buisness and the subject compliments are its affiliates, parent companies, products, locations, etc. An analysis of a list of players in the automotive industry might show some interesting dependancies.

All technical analysis aside, I am interested in showing a few examples of Pointsware being used to solve creative problems. After all this whole project stems out of my own desire to understand and systematize the analysis I have observed myself using during fruitful creative studio practice.

Pointsware online is UP!!

I have made some real progress and the interface and query functionality are up and running!
go take a look here

Currently the analyze.php page is still displaying all my debugging output so the SVG is at the bottom of the screen after all of the text, so if you go try it out please scroll to the bottom to see your image.

The SVG looks great courtesy of an iframe tag,
things still left to do in the short term
- highlight linkage path intersections
- make SVG XML file downloadable
- maybe do some tinkering with the display colors

All in all this application has come together faster than I had thought. Which is great because now I can start looking into taking the intersection concepts and providing web resources based off of these terms. I think that the LSA will add a nice level of value for Pointsware users

Very cool...

Started work on the web application

I put in place some ground work for the web application today, there is a few new pages on the site and a link to the new interface for Pointsware.
I have been working during this rewrite to take better advantage of nested custom functions to speedup the render rate of the SVG output. Nothing is working yet so that is still to be seen. However I think the code will benefit from the clarity of a more modular use of functions and for anyone else who might want to build on the code I have written the functions will help as well.
once the functions are working I will post some more code and talk a little about the mechanics I am working with.

.SVG is the new black... er .PS

In conjunction with a switch from Ruby to PHP, I am going to make the obvious step from Postscript output to Scalable Vector Graphics format. This makes sense as it will work very well with web application output and is XML based so it is well supported and forward thinking as concerns standardization for the semantic web. (O semantic web, when will you arrive). Writing SVG is a little more complicated than writing PS but you will never have to see the dirty dirty XML needed to create the Pointsware output so lucky you.

However if you are interested in SVG the most useful and technical site I have found with all the specs for the markup can be located at

http://zvon.org/xxl/svgReference/Output/index.html

This site has a comprehensive look at the tags and all their attributes. The W3C has a good overview of SVG as they have for all their compliance standards but the depth of usage for the information is not at the level of the link above.
Also if I might fantasize for a moment... the notion of being able to mashup SVG data from any number of sites around the web to create toolpaths for fabricators makes my rapidprototyping sense tingle. You never know when you might have need to cut google map directions into a pancake with a highpowered router.

More web app stuff

As I have been working on my website I am beginning to wonder if I should just go ahead and implement the whole Pointsware software in PHP. I keep coming back to this as a way of allowing people to access the mechanisms of the software with ease and without need to install Ruby and go through the whole process of running it from the command line. The implementation as web application is beneficial for access and storage of notes via Internet and mobile devices. Allows for latent semantic analysis to be conducted as an integrated part of the processing and display via easy access to google. Accounts, sharing, and posting of output can also create some community between users. I need to shift my efforts I think... and I have had the nagging suspicion that I need to do an over haul of the basic mechanisms that the code employs anyway, and switching languages would give me this opportunity. However I do think that I will finish implementing the second degree subject compliment linkage paths in the Ruby version and package it all up for people who might be interested in the application as it stands now.