In case you missed it earlier this week Google confirmed that it licensed a search algorithm and hired the algorithm’s creator, Ori Allon, a 26 year old PhD student from the University of New South Wales. The basic concept behind the new search engine is that it extracts portions of a web page that are most relevant to a given query and returns that relevant text. It also provides results for keywords related to the original keywords. There has been a lot of speculation going on about what this means for Google and search and I thought I would add my thoughts to the mix.
I’ve been reading as many articles as I could find about the new algorithm and have posted links to some at the bottom of this post. The algorithm is part of the Orion search engine and before anyone jumps to any conclusions this new algorithm isn’t anything that will be replacing the current Google algorithm anytime soon. So what does this new search algorithm do?
The Orion Search Algorithm
So what is it that Orion does? According the the original Press Release from last September Orion finds pages where the content is about a topic strongly related to the keyword. Ok don’t all search engines do that or at least try to? Orion then returns a section of those pages and lists other topics related to the keyword so the user can pick the most relevant. By returning the text it extracts Orion will let you see more of a page without having to click on the link to find out what it’s about. The results for other related keywords will give you additional information you may not have thought to ask about which could be very useful when researching a particular topic. Kind of like asking the librarian for a book and getting a excerpts of other books you may find relevant along with the book you first requested.
Very useful stuff here though in all honestly not entirely new. Most search engines give you a portion of the page it considers relevant as part of the results. And many search engines will provide related words you might want to search for and let you see sites related to the ones in the results. Apparently though Orion does this better. And it provides more relevant text from web pages than any search engine does currently.
How Google Might Incorporate Orion
Like I mentioned there’s mostly speculation where Orion is concerned at the moment. And again this isn’t anything that will be replacing the current Google algorithm we’ve all grown to either love or hate, depending on where our sites appear for our favorite keywords. There are some ways that Google might use this new algorithm to improve it’s own.
Google can use the algorithm simply to display better descriptions for the sites in the results pages. I really like Google and have always felt I got relevant sites to my searches, but I’ve never been all that crazy about the descriptions I see below the links. Google will sometimes use the meta description or the description for the DMOZ listing if there is one, but most of the time there’s some incomplete sentences with my keywords highlighted. Generally not the most useful information and I’ve often clicked on pages that barely had anything to do with my search.
Orion can perhaps be used simply for the text it extracts to provide better information about a given site before having to click on the link. This would allow searchers to eliminate a few more irrelevant links and even better find what they want on the first try. While that might mean less overall traffic for many sites it will generally mean more targeted traffic for all sites. The extra text would keep people on the Google site longer which will certainly appeal to anyone who may end up advertising on those results pages like AdWords subscribers and some may see that as bad for web sites other than Google and it’s advertisers. If it helps me determine which of the 10 links on the page has the best chance of having the information I want I’m all for it.
If Orion is good really good at determining how closely pages are related to each other then maybe Orion could be used after the initial Google algorithm to reorganize some of the results into more relevant groups. Essentially running the Google results through Orion. Often search results will yield pages that we weren’t looking for though someone else using the same query might find relevant. For example if someone searches with the query ‘computer repair’ are they looking for sites that offer computer repair services or are they looking for information as to how they can repair their own computer. Maybe the related functionality of Orion will be able to cluster the results into different subcategories to queries helping searchers zero in on the sites they want. This might be offered as a separate service to the basic search.
Taking the clustering concept further it might be possible to let searchers seed their queries either by providing additional information or automatically through the use of past searches as a way to personalize search results. I’m more of a do it yourselfer so chances are I was looking for the information to repair my own computer. Knowing this either by asking me or knowing more about my past searches might enable Google to provide results more relevant to me and a different set of results that would be more relevant to someone else. Could Orion be the part of the solution to personalized search that so many search experts see as the future of search?
Given that Orion also provides results to other related keywords it’s certainly possible that it’s main use will be to help less skilled searches in providing them results related to their query that they may otherwise not have found themselves. Not everyone who goes to Google is an expert searcher and helping them get results that an expert searcher might would attract them to Google and keep them loyal. Kind of like having that librarian there with you when you look for information for your research paper.
How about an idea that’s probably more from my own imagination than anything found in the current information. If Orion is really good at determining other related keywords could those keywords somehow be used to form a concept of a theme for a particular query. For example that searcher looking for information on computer repair might also be interested in results for specific computer parts like motherboards or CPUs. Is it possible this new algorithm can look beyond the page and at the site as a whole to determine how much of a theme the site has and whether a site that also contains pages on motherboards and CPUs and soldering chips might end up being more relevant than a single page on computer repair on a site that otherwise features content on general electronic repair. I admit this is some creative speculation on my part.
The truth is the only known quantities are that Google has hired Ori Allon and has also purchased the Orion Search Algorithm. Outside of the original September 2005 press release most everything is speculation. I doubt we’ll be seeing any major changes to Google or it’s algorithm anytime soon and all of this could have just been a way to keep both Ori and his algorithm away from Microsoft as they were apparently interested in the technology. Google may have been more interested in Ori than in his algorithm and purchasing it was a way to ensure his employment.
In any even it’s fun to speculate from time to time and from me this was one of those times. Please keep in mind that most of this post is pure speculation on my part. Perhaps more to do with what I’d like to see from Google than anything the Orion algorithm can do. Feel free to comment with your own ideas about how the Orion algorithm may be incorporated into Google or just in how you thinking Google and search in general might be improved.