Contextual Self-Organizing Map: Semantic Space of Italian Words (?)

The aim of this study is to test the lexical-semantic representation of 300 Italian words by means of a contextual self-organizing map (Zhao, Li & Kohonen, 2011). In distributed models, the lexical-semantic representation of a word is conceived as a series of points in a high dimensional space. By using a corpus-based approach words meaning is constructed through an algorithm which extracts the statistical co-occurrencies of words present in similar context. In the study, the algorithm generated a matrix of the words starting by an extract of the Co.L.F.I.S. corpus and calculated for each word the averages of the preceding and following words. The lexical-semantic representations obtained were used as inputs to generate the self-organizing map. The resulting map captured the conceptual and semantic properties of words (Zhao et al., 2011). More surprisingly, it captured also their syntactic features, such as grammatical class and gender. One possible interpretation of the results is that words occurring in the same semantic context occupy the same syntactic slots in phrases, blurring the distinction between semantic and syntactic levels of processing. Implications for the theories on the organization of the mental lexicon will be discussed.