SinJax’s House-o-Stuff

re: Jay-Z - 99 Problems

I couldn’t help but notice the amount of problems are grossly overestimated.

1. Overly zealous censors policing his lyrics regarding his love of late 19th century rapid-fire weapons
2. violent foes
3. critics
4. grew up in a bad area
5. Famous radio stations sullying his name due to contractual obligations he would not honour
6. Subsequently they don’t play his music
7. Writers and reviewers in his field displaying his image without paying royalties
8. Illegal contents in the trunk of his car
9. Being stopped by police
10. Young, black and poor choice of haberdasher
11. Not a mind reader, definitely doesn’t look like one
12. Speeding
13. Lost the keys to various storage compartments of his car
14. Hasn’t passed the bar exam regardless of a seemingly large knowledge base in traffic law procedure
15. Imminent arrival of seemingly unwelcome dogs
16. Had to forcefully resolve a situation involving a female.
17. Not being able to make the distinction between women and weaker men
18. Paranoid delusions of the existence of a ancestral paternal figure with whom he presumably seeks solace occasionally
19. Often takes part in fruit fights, presumably low vitamin levels
20. Paranoid Hallucinations involving membership of early 16th century pirate crew or military unit
21. Fear of incarceration in confectionary
22. Superiority complex
23. Slight fear of mythical creatures scratching his floor quelled only by his deification of the paparazzi, who he assures will dispatch of these fiends
24. Sexual abuse from high ranking holders of legal office
25. Loss of 50% of his stake in a local mill due to race
26. Harassed by various african americans
27. Savoury behaviour
28. Equally savoury weaponry

I’ve been annoyed with apple’s Mail app for a while now. For an operating system whose apps usually play so well with each other, Mail is a heaving pile of ass sometimes. There are several keyboard shortcut inconsistencies and wierdnesses. Furthermore, solving these problems is a rather non trivial process seeing as the applescript library for Mail is also thoroughly underpowered. Basically there are things that you want to do in mail without taking your hands off the keyboard that … well… you can’t! Turns out you can though… heres how

(more…)

Over the past month or so I’ve been re-evaluating my paper submitted to SAMT2008 . Specifically i’ve been looking at the Singular Value Decomposition and how its used for Latent Semantic Indexing. Up until now my understanding of the process has been fairly superficial, and over the past few weeks i’ve started some more in depth analysis of what it does, what the answers actually mean, how to make it work faster and how to make the answers more accurate. Thought I’d write up my findings so far, but what ended up happening is that i wrote up what the SVD and LSI are in…what i would call plain english! Enjoy

(more…)

Ehm! So! I had a thought that one could use some generic Lossless Compression Algorithms to make more use of the 140 characters provided by twitter. Can we squeeze out a few more characters if we’re clever?

After some fiddling, it turns out the command line tools on most unix platforms let you compress text pretty easily and pump the output directly into the stdio. Useful! The commands go along the lines of:

echo "text to compress" | gzip -cf 

Using this command, and the equivalent for bzip, i went about testing. So the baseline is 140 characters 1:1 mapping. The first set of tests concerned themselves with random strings. These have no structure so are especially challenging for algorithms like Huffman code which try to take advantage of some innate characteristics of the data. As you’ll see in the results, these performed predictably badly, with the “compressed” data size being larger than the number of characters basically throughout, though with some clear convergence well past our point of interest (140 characters).

This, of course, is not the basis case on twitter, where text is primarily (he said tentatively) in the form of structured English. Using The Mountains of Madness as the text source, i randomly selected varying length strings and compressed them as tweets. The results are promising. In the best case, using gzip we can get whopping extra 30 characters, no less than a 21 percent increase!

Interesting things to note include the poor performance on anything lower than roughly 90 characters on gzip, and also the poor performance of bzip as compared to gzip. Now this is expected as these compression algorithms aren’t really meant for this tiny amount of data. However, wikipedia seems to think Bzip uses Huffman Codes … so you’da thought it woulda been good at small text, though you can imagine if the dictionary is thrown in then its gonna be a massive overhead on small amounts of data.

Hooray. How can this approach be utilised in practise? Well the answer is some sort of syntactic agreement (#g?) that a message is gzipped, with some client side application decompressing and rendering.

now the next question: is this worthwhile? No doubt there will be backlash! This is against the inherent simplicity of twitter after all. Mac described it as talking in haiku. If you are forced to use 140 characters you tend to get to the bones of the thing more efficiently. Also, requirement for a special client to see what you’ve written? this sounds gay… too gay

Basically… i don’t think this is going anywhere … but it was fun to investigate the nature of structured text, and generating graphs is always fun.

Can you tell my PhD has hit a slow patch?

hugs

So, It would seem that on Friday, 13 Febuary 2009 at 23:31:30 GMT, UTC will be 1234567890. IS this a reason to get excited? No… not really… But its a good excuse to get together, drink some beers… maybe geek out

Seems an event that deserves note right?

Thoughts?

Also, completely unrelated but i saw this and laughed