engleski

Preliminary Report on the Structure of Croatian Linguistic Co-occurrence Networks

. In this article, we investigate the topological properties of word co-location networks generated from Croatian texts and, in particular, present a study of networks based on a different word co-location window size and on a inclusion/exclusion of stopwords. Our experiments expose properties of twenty- five co-location networks generated from three different corpora. In our network models, each word is linked to the exact specified number of following words if they co- occur within a sentence. Comparations are made for both networks with included and excluded stopwords, and for various word co-location window sizes. We conclude that the results indicate properties of a small world networks, with emphasis on the importance of word co-location window size and on the inclusion of stopwords.

complex networks, Croatian texts, co-location, stopwords

nije evidentirano