THE COMPARISON OF QUANTITATIVE METHODS FOR OPEN-ENDED QUESTIONS ANALYSIS

J. Haviger

University of Hradec Králové (CZECH REPUBLIC)
The paper deals with analysing the several methods that are suitable for processing unstructured open-ended questions. The questionnaire was focused on the implicit concept of family. Research sample consist of students of education (N=219). This paper is pointed into open-ended question: „Imagine that you meet someone who does not know the concept of "family", try to explain it to him/her the best you can.” Answers of individual student were lemmatized and desambiguated (for Czech language is not suitable stemming), transformed into a term-document matrix (TDM), and the words were divided into autosemantic and synsemantic group. TDM matrix was analysed by several methods based on quantitative linguistics, multidimensional statistics and text-network analysis. Result: for automatic processing of unstructured open-ended questions we can recommend two methods: 2h-point and lexical networks methods based on individual responses bag-of-words.