Evaluace kvality strukturovaných dokumentací s využitím metod umělé inteligence

The motivation of this work is to provide evaluation simplicity and speedup of an extensive number of documentations with the same subject. The tool for automate evaluation was created and tested on the set of structured documentations originated from student´s projects with the same assignment. The tool uses an existing methods of the artificial intelligence for the purpose of the supervised learning. The source of the first attributes group are string-kernel functions using the criterion of the text strings similarity. Another group of attributes makes provision for properties like the structure of documentations, the stylistic, the length of chapters etc., which are statistically evaluated. These groups of attributes are joined together to achieve higher diversity of classified objects. The experiments present results achieved by using of neural networks, naive Bayes and SVM methods. The analysis of the data has been performed by data mining studio RapidMiner. The contribution of this work is to find out the classification usage possibilities in the case when classified objects have very similar properties in the feature space.

evaluation documentationartifical intelligencevalidation