• SUMMA - Evaluation of Semantic Similarities
• PISA 2009 - Analysis of Condition Factors of Poor Reading Performance of Young with Non-German Language Origin

A large part of school knowledge is taught through texts, i.e. students learn primarily by reading. It is well known that texts summarization improves the reading skills, and consequently the learning efficiency is also increased. An intelligent tutorial system was developed to assists students writing summaries. The system uses the so-called Latent Semantic Analysis (LSA), a computer-model to evaluate the semantic similarities of words and texts. LSA is a statistical method for analyzing the occurrences of words in large text corpora. It generates a high-dimensional vector space in which words, sentences and paragraphs are represented as vectors. Thus, a word vector can be understood for example as a representation of the meaning of the word. German semantic vector spaces for different topics were generated and these were integrated in tutor systems.

A platform that generates course-related efficiency control's feedbacks was developed, which is able to quickly and easily detect contents plagiarism and to assess open answers automatically. Besides, methods were developed for the classification of text contents.

Accomplished Work

• Investigation and development of methods
• Software architecture design
• Development and implementation of the server
• Implementation of an integrated development system
• Implementation of a system for assessing student task answers

Consortium

Department of Psychology at the University of Würzburg and a close cooperation with the LSA Research Group of "Summary Street" project (Professor Eileen and Walter Kintscher, University of Boulder, Colorado). Also in cooperation with Professor Sandra Jhean-Larose and Guy Denhière (Paris), Professor Spinath (Heidelberg) and "Strategic Innovation Center of the Bavarian criminal investigation."

Press

The following press news and radio features appeared in December 2006 in various regional and national press institutions reporting the research project.

Selected newspaper articles:

• Bayerischer Rundfunk Online (December, 4th 2006) 47KB
• Die Welt (December, 4th 2006) 9KB
• Süddeutsche newspaper (December, 5th 2006) 46KB
• Mainpost and associated newspapers in Lower Franconia region (December, 6th 2006) 281KB

Radio Posts:

• Bavarian Radio 2 (11.12.2006) and Internet news at Bayerischer Rundfunk 3 (December, 17th 2006)
  bayerischer_rundfunk-32bits.mp3
• Deutschlandfunk Forschung aktuell (December, 27th 2006)
  dlf_200612271645-32bits.mp3

Studies on the development of reading skills in German-speaking areas have shown that children and adolescents of non-German language origin are highly overrepresented in the group of weak readers. The differences in the analysis between good and poor readers have been investigated, without the consideration of migration and linguistic background of the pupils. The profiles of poor readers in primary and comprehensive schools in different German federal states were analyzed in depth and specified taking into account their linguistic backgrounds. These results allowed to forecast the results of the PISA 2009 reading tests of the same scholars.

A system was developed that allows the acquisition of phonological awareness, working memory, listening, and language structural skills.

Accomplished Work

• User requirement analysis and software architecture design
• Development and implementation of the server
• Design and development of the user interfaces