NATURAL LANGUAGE PROCESSING METHODS FOR CONCEPT MAP MINING: THE CASE FOR ENGLISH, KAZAKH AND RUSSIAN TEXTS

Abstract

Concept maps are used for knowledge visualization via representing an input text or domain at the conceptual level. Concept maps reflect the systemic relations between key concepts of a text/ domain and thereby contribute to a deeper understanding of text/domain ideas, save time spent on reading and analysis. However, the process of concept maps construction is laborious and time consuming. Currently, there is a lot of research on the idea of automatic generation concept map from natural language texts. The problem has a high practical value, but in theoretical terms, methods for its solution are mainly language-dependent. Such methods require high-quality annotated linguistic resources, which is a serious problem for low-resource languages like Kazakh. In this work, we analyze the issues related to language-dependent approaches and present our experimental work on automatic generating concept maps from English, Kazakh and Russian texts. We use a well-known language-dependent method called ReVerb which was originally developed for English, and on the example of this method we explore the issues that we have encountered in the case of Kazakh and Russian languages.

Keywords

• concept maps
• concept map mining
• natural language processing
• low-resource languages
• r language