Similarity Research at Stirling


Something like 90% of all computer power in the modern world is used to search for things, web services and documents in particular. Search is so successful, it’s changed the way we behave.

How do we do search? In a small way, by using similarity. Typing “similarity search entropic distance” into the Google Chrome menu bar is instructing the browser to list those documents on the Internet which are similar to the microcosmic document you’ve just typed in. But, in the developing Internet, we should start to expect much more.

This is where technology ends and research begins. The photo above is a picture of my favourite watch; it’s broken. I want to search the Internet for a new one, and what I want is a watch that’s has a nice plain dial, metal bracelet, day and date probably, and not too expensive - that is, I want to find one which is very similar to this one, but I don’t care much about which attributes are a little different. That ability, in general, is beyond the scope of our current understanding: just one reason we need to do research.

Or, I could search for similar images using Google: but the results aren’t too encouraging as you can see... it seems that searching huge collections with a good similarity algorithm isn’t yet possible.


You’ve just used similarity, probably, to get to this page...