• community discovery
• network analysis
• network science
• nlp
• semantic spaces
• word2vec

It is proposed a complex network approach to the self-organization of word meaning in semantic spaces, NLP techniques capable to model theories of distributional semantics: words link each others if they can be substituted in the same linguistic context, shaping clusters representing semantic fields.
This perspective can be used to model a mental lexicon of word similarities, supporting psycholinguistics or emerging area of research in Cognitive Science.
Extracted networks are analyzed in terms of global properties and meso-scale structure: they have small-world properties and they are assortative by degree; probability of degree distributions follows a truncated power law.
Through three community discovery methods, clusters have been identified (1) as strict semantic fields (2) as different linguistic contexts in which the same word can appear (3) as semantic fields conveyed by a single word, highlighting polysemy.