hrvatski

Automatsko stvaranje konceptualne mape iz nestrukturiranoga teksta na hrvatskome jeziku

This paper discusses the problem of the automatic creation of conceptual maps from documents written in Croatian language. In the first part of the thesis, theoretical assumptions of research are described, and an overview of previous studies in this area is given. In the second part of the thesis one method suitable for the creation of concept maps from unstructured textual sources in Croatian language is described in detail. That method uses statistical and data mining techniques enriched with domain thesaurus and linguistic tools. Based on this method, a prototype system for the automatic creation of conceptual maps from documents in Croatian language is created. Content and application validity of generated conceptual maps were measured using an online questionnaire. Based on the responses of 538 participants, content validity is rated good to very good. During the evaluation of application validity, participants who used concept maps as supporting material achieved better results than participants who used the summaries of documents or complete documents. The obtained results confirm that using the described method, it is possible to create concept maps that illustrate well the contents of the source documents from which they were created, while quality and usefulness of created maps in specific uses are comparable or slightly better than the summaries or full documents.

konceptualna mapa; automatsko stvaranje konceptualne mape; tezaurus; obrada teksta; sažimanje teksta; dubinska analiza podataka; vrednovanje konceptualnih mapa

nije evidentirano