Hi, I would like to announce my Master's Dissertation, which is now available online. The title is "Refining the CoGrOO Grammar Checker". CoGrOO is a Portuguese grammar checker built on top of OpenNLP.
The text might be interesting for users of OpenNLP, specially the sections from 6.1 to 6.9, which describe the process of tuning the NLP models using the OpenNLP framework. For each module I measured how each OpenNLP tool can help, for example, by introducing an abbreviation dictionary in the Sentence Detector, or a lexeme dictionary in POS Tagger. I also evaluated the impact of a custom feature generators or a custom sequence validator, and how to customize the tools for other needs, for example a I built a contraction finder, word featurizer, chunk head finder and shallow parser by modifying the existing tools. I also compared the effectiveness of Maxent and Perceptron models and how "cutoff" impacts each algorithm As required by my University, the title and the thanks pages are in Portuguese, but all the other content is translated to English. http://www.teses.usp.br/teses/disponiveis/45/45134/tde-02052013-135414/en.php Thank you, William
