Twitter Sentiment Analysis in pt-BR (extended to all languages) is a project to make a real time analysis over tweets using PHP and external tools to train a model and apply over texts we are intended to categorize.
I'm testing 2 platforms:
Apache OpenNLP: http://http://opennlp.apache.org/ libSVM: http://www.csie.ntu.edu.tw/~cjlin/libsvm/
and use the PHP-SVM interface from Ian Barber: https://github.com/ianbarber/php-svm
more information can be found in: http://www.eddieoz.com/us/projects/twitter-sentiment-analysis (under construction)
2013-01-15 Collecting a bunch of tweets, choosed by emotion.
It uses liblinear-php for machine learning directly from php: https://github.com/eddieoz/liblinear-php