Cordelia Nowak Summer 2012
6/12/12 I am now five days into this project and I still feel like I am trying to get my bearings. There is so much code to sort through and my short term goals never seem clear. I finished going through the 1800 words from genesis and I got the manuscript to line of with the transcript yesterday and found the transcription errors. That felt like a nice accomplishment, but now i've abandoned it altogether for this next step.
I don't feel like I accomplished very much today: I got a feel for the the psm works and I found examples of some Syriac letters which were not accounted for in the SyriacExamples.m file. At the moment I am working on understanding FitPsmWordModel.m, but it's kind of a tiring process.
On the bright side, I feel like I've been understanding more MATLAB everyday. I'm still not fond of the syntax. Nor am I a big fan of the rest of the language, but it is less baffling now than it was this time last week (and since I didn't put in any work last week, that's great!).
I'm not sure how well this schedule is working out for me. I might try to come in earlier tomorrow so I will have more time to ask questions and less time sitting here alone. The problem is that I really like to have slow mornings and I'm not sure how long I can keep waking up at 9 o'clock for.
I think today has gone pretty well so far. I got the models up and running. They letters are now in correct alphabetical order and many of them even look like the letter that they represent! In the next couple hours I will try to fix the scaling of the models and find a new model for nun because it is indistinguishable from zain.
In other news, I don't know how to properly update this wiki. Sorry about that.
I made a lot of progress today, which is good. The program can now properly identify a letter once it knows that it's there. I changed a lot of the models so that they are now sized properly and actually look like what they are supposed to look like.
I am currently attempting to compel the little boxes of Syriac in the corner to close after ten seconds, but working with java inside of MATLAB is even more awkward than MATLAB itself. Right now I can get them to close but, they do it almost instantaneously because the wait method is throwing an IllegalMonitorStateException.
For my last hour or so today I am going to look over some files and see if I can figure out how they work.
It has been a while since I have updated this, so let me see if I can summarize what i have been doing for the last two weeks.
I did a lot of research on Support Vector Machines and machine learning in general. I got an SVM implementation (LIBSVM) which now functions on the data which was provided for it. Now I am working on getting the SVM to work on my data, which I will first have to put in a format which it will recognize.
I am currently trying to understand DecisionForestExperiments.m which will help me to see a little more about machine learning and training the SVM. In order to do that, I had to get initGW20 working which took a while. But I figured it out!
Generally when I've gotten stuck in the last couple of weeks, I've tried to go back and figure out how to make the little windows of syriac text close. I have made some progress, but I've been frustrated in odd ways
Also, I met Tracy Kidder today, which was awesome. I got him to sign my book and I listened to him talking about journalism and writing for an hour. It was a good day.