Nominal Semantic Classes Using. Decision Trees and Bayesian Models UPF has focused on nominal semantic classes, performing research on the use of Decision Trees and Bayesian Models to approach this task. See Related Papers section for details about used methods. Nominal lexical semantic classes gather together properties that appear to be linguistically significant for a number of linguistic phenomena. Determiner selection, selectional restrictions and noun collocation have been described in terms of such groupings of properties. In addition, these classes are often used to generalize over particular senses of different words. For instance, Xxxxxx et al. (1990) used a number of lexical semantic classes as features that ordered the nominal meaning hierarchy in WordNet. Applications that use nouns annotated with lexical semantic classes include: machine translation, discrimination of referents in tasks such as event detection and tracking (Xxxxxxxx et al., 2006), question answering (Xxx et al., 2001), entity typing in named entity recognition (Xxxxxxxxx & Xxxxx, 2005; Fu, 2009), automatic building and extending of ontologies (Xxxxxxxxx et al., 2005), textual inference (de Marneffe et al., 2009), etc. Furthermore, nominal lexical semantic classes have also recently proved to be useful information for grammar induction (Xxxxxx et al., 2011), where problems come from the need of generalizing over a high dimensional space. Lexical semantic noun tagging in large lexica is still mostly done by hand, and the high cost of this exercise hinders the production of rich lexica for different languages. In addition, domain tuning of lexica is considered too expensive, and the use of an inadequate lexicon is one of the causes of poor performance of many applications. Thus, current research on automatic production of class-annotated lexica is expected to have a high impact on the performance of most NLP applications. Most critically, it will bring significant improvements in their coverage over different languages and domains. Thus, any reduction in the amount of human work required for the production of these resources can contribute to improve the current situation. UPF has addressed cue-based noun classification in PANACEA WP6.2. Its main objective is to automatically acquire lexical semantic information by classifying nouns into previously known lexical classes. This is achieved by using particular aspects of linguistic contexts where the nouns occur as cues that represent distributional characteristi...
