Alwin de Rooij | Portfolio

In close collaboration with Hugo Schoonewille we have created a layer on top of the usual web search engines like google and altavista. This layer aims to visualize semantic relations using new advances in the field of human computer interaction information retrieval and is loosely based on the semantic network theory. The program we have created makes use of information maintained by people on the wikipedia website. Our program simply takes a search query from the user and then uses a custom keyword matching algorithm and comparisons between super and sub catergories to estimate the semantic relations among keywords found around the query. The acquired information is then fed into a self organizing graph system (wich employs a physics model) that tends to cluster the right connections between keywords based on the strength between keywords and their categorical similarity. The graph is visualized and has an interface similar to google earth for easy browsing. The semantic network can be expanded and keywords can be used for a google search.

Fig 1: Version 1 of our semantic search software “Simple Search”.

Fig. 2: Zoomed in on the search query and its surroundings

As you can see in Figure 1 and 2 the search query is positioned in the middle of the screen. By using our self-organizing graph algorithm related keywords are positioned around the query. To give an example, acousmatic sound, musique concrete and electroacoustic music are closely related to the query. These are closer to the query than keywords like computer, real-time and music composition. These are very much related, but in a totally different semantic scope as the other keywords. One can see that the keywords are not only visualized in relation to the query, but also to eachother.