Shopping for shoes!
Picture this: You walk into a store and tell the sales associate - "I'm looking for shoes". "Why certainly, Sir (or Madam)", says the sales person - and magically the entire store is rearranged for you - 500,000 of their best shoes on racks as far as the eye can see, all arranged according to their popularity, with the ones most customers prefer closest to you, and the least favored ones farthest from you. You don't feel like walking for miles just to find new shoes, so you go back to the sales associate. "Hiking shoes", you say pointedly - and once again the store is magically rearranged, hiking shoes as far as the eye can see. Wait - you forgot to say "brown hiking shoes" - your color of preference, and it appears that most buyers prefer shades of tan and green. Next time, you'll remember to use the whole term - brown hiking shoes - but it would be even better if you could remember to mention your size and the need for rubber soles, at the same time.
Wouldn't it have been easier if, when you first asked for shoes, the sales person had just asked about your preferences in the first place - color, size and everything? In fact, this is how most (well-run) retail stores operate. Just because most customers prefer their shoes in brown, black or tan doesn't mean that you can assume that's what this particular customer wants - this particular customer may want yellow shoes. It's easier to just ask up front.
Why don't you just ask?
What does this have to do with Internet Search? If you try a Google search for "java", it returns results 1-10 of 553,000,000 results, all of the first 10 having to do with the Java programming language. It appears that Java developers are good at SEO, or perhaps there are just more web pages about the programming language linking to each other. But - what if I meant something else? Why can't the search engine just ask me - "Did you mean, the programming language? The island in Indonesia? Coffee? A settlement in Scotland? A town in Georgia?" [Ok, for the curious, I got all these definitions from Wikipedia.] I would much rather get only 5000 results (heck, even 5 would be fine), if they were relevant, authoritative and comprehensive.
Personalization to the rescue
At this point, I'm sure someone will be tempted to say - "Personalization, dude!". Once the search engine watches you for a while, it should just know that you are a super-geek programmer, and of course you mean the programming language, not the island, and certainly not the town in Georgia. [Wait a minute, is that Georgia the state, or Georgia the country?].
I agree. That is certainly a technical possibility, not much of a stretch. But I value my privacy - I find the idea of Google Search watching everything I do and look for, slightly unnerving; and no, I don't have anything to hide. For example, my wife does most of my clothes shopping (bless her!), and so, in a clothing store, I expect her to instinctively pick out something that she knows I would like; but if the sales person were to do that, even with the best of intentions - frankly, I would find that somewhat creepy. I would much rather be asked at search time, than tracked.
A new search engine that just launched officially, MetaMojo.com, aims to do just that. MetaMojo is one more in a long and growing list of Vertical Search Engines. The primary motivation for this particular search engine is corporate: to power the internal search within the growing list of web properties created by MojoSupreme; direct search benefits to the end user, as founder Ashkan Karbasfrooshan writes, are a bonus. Key differentiators for MetaMojo include the ability for a user to specify a "Category" to qualify the entered search terms, and richer results (such as video, reference lists and blogs) as the various properties of MojoSupreme grow. A large number of categories have already been defined, including Video clips, Blogs, video classifieds, video games, beer, travel, movies, food, health, music and wine, as well as "content web sites" that include Top 10 lists, city directories, Rock n' Roll, advice, and (unusually) Alexander the Great. It appears that MetaMojo will only crawl and index the internal content for MojoSupreme, rather than a wider sampling of the web, limiting its usefulness for the moment. This is not a fatal flaw though; as Wikipedia and About.com have shown us, once the content base is large enough, internal search can be just as useful as a search of the broader Internet.
The overall search approach of MetaMojo, from a user perspective, is reminiscent of Kosmix; more on that in a future post. By targeting discrete verticals, these search engines offer key features over mainstream search engines like Yahoo and Google, such as Authority (since only specific sources are allowed) and Relevance (by asking users to specify constraints/vertical).
What about Google?
Why doesn't Google do this? Why doesn't the Google search page ask users to provide a category or otherwise qualify queries, so that returned results would be more relevant? It appears that they have certainly thought about it. Here's a post from Danny Sullivan on Query refinement and another one with a comprehensive look at the various attempts to improve search results presentation. His main argument is that query refinement will only work if it doesn't cause more work for users, a sentiment that any experienced user interface designer would agree with.
So - what do you think? Would you, as a search user, be willing to select/type in a category if it would significantly improve the quality of search results? I think I would make that tradeoff in a heartbeat, but I may be in a minority. In the meantime, the next time you go to Google - make sure you specify all the important terms of interest up front. Or better still, use a vertical search engine in your area of interest.
In an email exchange, the MetaMojo folks have explained that their vertical search engine does crawl and index other high-quality websites outside the MojoSupreme network, for example, sites like Fodors, Zagat, Expedia and others for travel searches. This enables them to maintain the high quality and authority of their search engine (less spam), while providing more comprehensive results. Thanks for the clarification, Ashkan!