Wednesday, February 01, 2006
An Auspicious New Year
First things first. For those that are return visitors you may know that last November I married a lovely Chinese woman named Yang Nian. We had both hoped the U of I in Champaign/Urbana Illinois would renew their Scholar exchange program with Jinan University in Guangzhou and we would be reunited permanently in March. That doesn't look likely now, so we have spent the last month gathering and filling out papers to apply for a K3 visa (marriage to American citizen). I sent the papers to the INS Monday overnight mail so they could be pouring over them even now. Nian thinks we are in a for a long approval process, I'm not so pessimistic. We have about 3 steps to complete, and now that we have taken the first one, things should start going faster. How quickly the first set of papers gets approved should be a good indicator of how quickly things should move along (stay tuned).
This last weekend was the Chinese New Year, celebrations last all week in China. It turns out this year, the-year-of-the-dog, is the same animal-year as my birth-year. As such it was important to Nian that I wear special red underwear with a good luck symbol on them to ward off any misfortune in the coming year. Seems in China your birthday or birth-year are not necessarily lucky things unless you take steps to make them so, quite the opposite it would seem. Since it is now Wednesday and I haven't been hit by an errant bus or out of control turnip truck I can only assume the undies did their assigned job of scaring away any malevolent demons wishing to do me harm.
In addition to filled out INS forms in a package Nian sent me recently were special boxes to be filled with candies and given to friends an family to celebrate our marriage. While our marriage was in November there does seem to be good synchronicity with the Chinese New Years and our filing the I-130 for a K3 visa with the INS as the first step towards our life together in the US. Mutual friends of Nian and I gratefully accepted the boxes here State side and a few of my closer coworkers.
Changing gears a bit I had a couple of minor insights in how to code my Monkey-Typewriter application. It isn't in web form yet, but I got a simplified first generation command line version working last week. It was actually a bit of a let down. The transformation of text from one passage into another in an evolutionary driven way was not as amusing or informative has I had hoped. Still it is just one more step along the way to a more ambitious goal dealing with experiments in AI (Artificial Intelligence). In a way Shakespeare-O-Meter and BaBeLiZeR were steps towards Monkey-Typewriter. And BaBeLiZeR and REBUSizer are steps in automatically grabbing needed info directly of from the web to complete a task. All baby steps to be sure.
Anyway, I am constructing a better word-database for Monkey-Typewriter even as I type this. I downloaded the most common 15,000 words used in the English language some time ago as part of the Shakespeare-O-Meter project. Using this list I performed 15,000 web queries from merrium-webster.com to get their definitions. Right now I am doing some ad-hoc automated data analysis to denote the various categories (noun, verb, adjective, etc...) and connections (synonyms, antonyms, related words, etc...) to form the custom database that Monkey-Typewriter will ultimately use. As opposed to its more primitive categorizing methods (similar endings, similar lengths) the first crude version is using now. Hopefully this will result in more natural and interesting looking sentence evolutions.Back when I worked at Wolfram Research I had had in mind to create a neural net based AI engine that used the feedback of thousands of website visitors to train it into giving useful responses. A first step would have been a proof of concept Eliza or Ractor chatterbot like program whose responses became more entertaining with time by allowing users to modify the web-search and query selection rules in a wiki type fashion. Ultimately I had hoped to develop a genetic algorithm to augment this process that learned from how users modified the query selection rules and question-and-answer database and could improve itself. Users would then grade how well the self-modification performed and reinforce modification schemes that lead to better responses.
All of this would require hundreds of feedback responses per day to have any chance of creating some non-trivial AI engine. Given that few of the toys written to date have had much use, I'm not optimistic about realizing these rather grandiose goals. Still I find that as I write my, admittedly simple, toy applications for BNL I end up creating tools that make writing the next toy easier and hopefully more entertaining. With a little luck I might eventually stumble on something that gets a lot of return repeat use. Nian's brother Yi suggested I develop my projects a little further and perhaps submit them along with a job application to Google. Too bad I missed the bandwagon on getting into Google before its recent IPO.
Well I'm running long again, I really do have to get the hang of making shorter more often postings. I haven't even mentioned Nian and my house hunting. But that is as they say a story for another day.
