Advantages and Limits of Text Mining Software for Analysis of Students' Satisfaction in Online Education
Zdena Lustigova, Charles University in Prague (Czech Republic)
Veronika Novotna, Charles University in Prague (Czech Republic)
Abstract
Presented article deals with the possibilities of statistical software packages, which include data mining and text mining tools, for the analysis of unstructured text available at different Open Educational Resources. The information about satisfaction is often available in the form of multilingual information, hidden in users reviews, chat rooms, tea rooms and other unspecified and unstructured ways of feedback. To read such responses is often time-consuming and almost impossible. Text mining helps us to 1/classify and to categorize the type of responses (complaints positive, negative, irrelevant, disease, etc..), usually on the base of sentiment analysis, 2/ to reveal the most frequent problems, 3/ to discover similarities and patterns and/or 4/ to identify similar text records (clusters).
Authors do not present the "big data" approach, based on powerfull (and expensive software). They focus just on part of the whole large scale of users' reflection to present the basic problems researchers might meet while working with available software packages like Statistica, etc. The unstructured text they processed, (mostly users remarks and reviews), was created and published in 16 languages. Authors describe the problems they met, especially in the area of multilingual information processing. Despite all effort, nearly half of languages, used by users, failed to process.