Label Comparison Analysis Using Hybrid Similarity for Labeling Systems of Indonesian University Websites (English Menu Version)

Abstract

Labeling systems are an important component in designing a university website. In order to avoid misinformation, website developers need to design a website labeling system that can represent information that a website owner wants to deliver to the users of her website. One way of designing a labeling system is by comparing and studying the labeling system used on competitor’s websites. Syntactic similarity is usually used for comparing the websites’ labels. However, it has a limitation that can only calculate similarities based on strings only, so that the possibility of two compared labels having the same meaning will be considered as different labels. To provide the meaning of the label, the comparison using semantic similarity is used. However, semantic similarity has limitations such as not being able to process two words or more and possibly having no path exists between two word senses. Therefore, this research proposes a hybrid similarity where the highest result of syntactic and semantic similarity comparison between two labels is combined and able to cover the limitation of two methods. The hybrid similarity shows the results of possibility of higher score between combined syntactic and semantic similarity score with suggestion on the list label (based on semantic result). The expert suggest output of hybrid list label to help developer create and reviewing existing label, and then participant who answers the questionnaires 82.16% agree with label that show the acceptance of the label. The list label based on hybrid score can help and assisting the web developer to create new labeling systems for a new website. It is also help developer to improving an existing website’s labeling systems. Keywords: university website, labeling systems, syntactic similarity, semantic similarity, hybrid similarity, Levensthein Distance, Wu and Palmer.