| The Johns Hopkins SENSEVAL2 system descriptions (2001) | |||||||||||||||
Abstract | |||||||||||||||
| This article describes the Johns Hopkins University (JHU) sense-disambiguation systems that participated in seven SENSEVAL2 tasks: four supervised lexical choice systems (Basque, English, Spanish, Swedish), one unsupervised lexical choice system (Italian) and two supervised all-words systems (Czech, Estonian). The common core supervised system utilizes voting-based classier combination over several diverse algorithms, including decision lists (Yarowsky, 2000) and decision stumps, a cosinebased vector model and two Bayesian classiers. The classiers utilized a rich set of features, including words, lemmas and parts-of-speech modeled in several syntactic relationships (e.g. verbobject), bag-of-words context and local collocational n-grams. The all-words systems relied heavily on morphological analysis in the two highly inected languages. The unsupervised Italian system was a hierarchical class model using the Italian WordNet. | |||||||||||||||
Publication details | |||||||||||||||
| |||||||||||||||