I assume you have read the Methodology section which explains the procedure of the first steps of our project: extraction of massive amout of information from Google and its analysis. The graph we obtain here is based on two quantities, Google Similarity (semantic relatedness) and homophony that can be tuned thanks to the cursors.

In the interactive version of the graph (which doesn't work online yet) if you push cursor 1 to the left, you select words that are semantically more related to the central word. If you push cursor 2 to the left, you select words that are homophonically related to the central word. You may navigates by clicking on words or choose automatic navigation.