The development of a network thesaurus with morpho-semantic word markups
Marko Orešković, Mirko Čubrilo, Mario Essert · Repository of Faculty of Mechanical Engineering and Naval Architecture University of Zagreb · 2016
This paper presents a part of the network frame of Croatian linguistics which focuses on a new kind of thesaurus, based on morpho-semantic features of words.Instead of the classic (e.g.MULTEXT-EAST) POS tagging of words for grammatical and some semantic categories (e.g.animate), in this paper every word has its hierarchical T-structure which can hold various data types in its branches (string, integer, link, word list, ordered word list etc.), and in that way words and their various occurrence possibilities in a text can be described even better.Moreover, the known WordNet or other semantic structures (e.g. the Croatian Language Portal, terminology repository or a network encyclopedia) can be presented as T-structure nodes in the same way.During this process each word in the definition of an entry is linked to a lexicon, which results in increasing the semantic connectivity of words by at least one order of magnitude (about ten times more semantic relations).Searching through and browsing such a network dictionary brings a new dimension, and words in the dictionary, beside the paradigmatic, also possess all the syntagmatic properties, because the computer processes their appearance in any utterance or sentence as a series of connected nodes (LOD objects).This provides the possibility of storing all data in triplestore (e.g. on the Virtuoso server).