Doctor Tanaka-Ishii from the Electrotechnical Laboratory (ETL) in Tsukuba has worked on building words co-occurrence graphs in order to get a representation of concepts. That approach was first aimed at comparing a non-aligned multi-lingual corpus. Indeed, with the help of a bilingual dictionary and co-occurrence graphs it is possible to find a mapping from one graph to the other that can be used for disambiguation of words in the framework of machine translation.
The graph is thus automatically built and is supposed to represent relationships between concepts, no matter what the language is, just like the EDR dictionary does.