The new Semantic Hacker match server exploits the fact that Semantic Signatures® are mathematically related. The original example tools we provided have a “similarity” function. This produces a “score” of how closely related two signatures are. Often times when writing applications a large set of documents is in hand, and one wishes to find the most closely related documents to some other document. This is exactly how the Wikipedia extension (and the front page demonstration) work. We used the API to generate a signature for every single Wikipedia page. All 2 million of them. Then, we took those signatures and added them to a match server. Once that’s all done we can get the most closely related Wikipedia articles to any document.
The concept of the match server is the same as using the similarity tool and then sorting by which ones had the highest score. That’s tedious work and error prone code. We’re providing the match server to speed up application development of ideas that require it.
The match server is also very very fast. It can sift through all those 2 million Wikipedia pages and grab the top matches in less then 10 milliseconds.