Scoble wants more from my book, so he writes it up, and the comments are flowing in. Cool!
On the subject of why he thinks engines are doing a poor job, (using a hotel search as an example), he asks:
So, what COULD search engines do? Well, first, give me some choices at the top of the page. Why couldn’t search engines ask you these questions:
1) “are you looking for hotels in New York or named New York?”
2) Are you looking for hotels with free Wifi?
3) Are you looking for hotels with great views?
4) Are you looking for hotels nearby major tourist destinations?
5) Are you looking for hotels with above average ammenities like super large bathtubs, well stocked minibars, etc.?
This reminds me of the passage in the book with Gary Flake, who is now – huh – at Microsoft (he was at Yahoo when I spoke to him). “If only I could have one more modal dialog….” In other words, he wished searchers were more like Scoble – sure of what they wanted, and willing to engage in a dialog with the search engine. Most, it turns out, are not.
“Why couldn’t search engines ask you these questions”
Won’t happen. Too big, too much, too many variables, too high of an expectation of the searcher. Unless there is metadata that hotels expose (or someone does it for them), the permutations of questioning is far too broad and the results would be less than optimal.
Using your hotel example, I would also want to know tons of other stuff before I make a decision on where to stay. Am I on business or vacation? Are my bride or kids with me? Which hotels have inventory and would make me special offers? The list goes on.
In fact, coming out to Web 2.0 presented me with exactly this dilemma since the Argent was booked. So was the Marriott by Moscone. The Hilton was too (my first choice was conference convenience…the other two were based on my participation in their loyalty programs). I ended up at the Hyatt Embarcadero spending more than I wanted but refused to stay at an unknown hotel in the Mission — or down at the Wharf — for less since I’d have to cab it and wanted to walk to the Argent. Expedia and Orbitz were worthless finding properties so I called and relied on my company travel group to narrow it down for me and book it (didn’t do this first since I like to bypass the travel group and decide on my own…and have them just book my choice).
Until we have Identity Management for humans (are you a one-time-per-year or a frequent traveler? Stay at Super 8’s or Hyatt’s? What’s your budget for this trip?) and open directories that hoteliers (and others) find imperative to participate in (with taxonomies and consumable data that search engines can rely on), expecting users to input innumerable variables and expect an algorithim(s) to decipher — and present nicely for human consumption — the data to make a decision or to shop for a hotel (or anything else frankly) is unlikely.
I’m no expert on any of these topics, but I think that Rollyo.com is a step in an interesting direction. Personalized searches, according to your most-visited/re-visited sites (data gathering techniques aside) should generally give better percieved results to the end users. Regardless of the actual usefullness of the results.
Search engines could build user-profiles based on feedback rankings of various web site search results in a similar fashion to the way Tivo profiles work. Issues with privacy data would obviously be an issue.
Twenty questions was dead by the 1980’s, or at least understood to be a difficult niche to work in, eg for medical diagnosis.
From what I’ve seen, drill-down is the way to go, if there’s a way to get the user to keep inputting attributes. The Information Retrieval community seems to be focused on producing the perfect ranking for a single query. If the user has more interactive feedback, she/he will narrow down the selectivity her/himself. Last week I talked to a company who told me the same thing; one of their most popular features is a drill-down query screen with instant update after each edit.
A simple example is from Shakespeare (from the search doohickey linked on my homepage). One would think the phrase “To be” at the start of a sentence would narrow down to a specific quote, but he uses it 197 times. Adding a comma, though, gets a single best-known result. This process would be hell with a conventional search box.
The obvious limitation of all current search engines is that they are primarily engines with scant attention to how to drive the bloody things.
The interface is now the bottleneck for all search. We (Visual i|o) are focused like a lasar on the interface and we have a prototype for all to checkout called SearchIris here: http://www.visual-io.com/search/index.html
Read the about page before diving in:
http://www.visual-io.com/search/about.html
In short, Google and Yahoo! are merely black boxes (oracles, wizards). SearchIris is an open, discoverable, comprehensible and accessible tool. Users quickly grasp the simple underlying model and are immediately productive. SearchIris introduces a concept that is ubiquitous in the real world but entirely absent in the digital search world: nearby search results.
Enjoy,
Douglass Turner
This is easy to contrive a feasible set of contextual queries for the Airline/Hotel reservation example, but the problem quickly becomes intractable when applied to all the millions of search domains people are interested in. These questions and there answers would have to be automatically derived.
The problem he describes is best left to custom, individual sites that people go to when they know that they want to find something specific.
Search engines are not mind readers, ther’e just a glorified version of Grep.
John – i just finished the book and had written something along the lines of Scoble before i discovered the conversation. I suggest even more – a collaborative industry discussion – “Search for Search”!! See [1]
Great book btw.
Interestingly, when you go to print.google.com and type in “The Search” (or your name) – the book doesn’t appear 😉
[1] http://stevenr2.blogspot.com/2005/10/final-comments-on-search.html