Google Killers? Maybe Text Search Killers

Every time there is news of a new search engine, pundits start asking the question, “Is it the Google killer?” Although this seems like more of the tired horse race journalism that James Fallows described in his 1996 book, Breaking the News, there actually is something to it. Only the target of the crosshairs is not Google, but text search itself, because it certainly cannot continue as we know it.

Let’s make it clear at the top: I’m not suggesting a cessation of all text searching. Too many people are looking for too much information that is contained in text. But the actual expression, no matter what intelligence built into delivering a set of pages to a reader, is a hidebound habit. Online hyperlinked search goes back to the 1970s, with tools mostly named for characters from the Archie comics series.

There have been tremendous advances over the years, but largely in the sense of more thoroughly indexing materials and more cleverly suggesting what might be to someone’s interest. Look at Google’s front page today. If you substituted hitting the tab key repeatedly instead of indicating a choice with a mouse click, the expression would have been self-explanatory to an early ARPANET/Internet audience. We can quickly see the fundamental aspects of the habit:

1. No matter what you look for — whether writing, video, images, or audio — it all comes down to text descriptions.
2. You will likely receive multiple pages of results, only one or two of which you are likely to view, showing the false assumption that more is better.
3. Depending on the search engine, you will have varying degrees of precision in narrowing your interests. Now instead of ignoring hundreds of pages of results, you’ll ignore perhaps only a dozen.
4. Accompanying any results will be some ads, so that someone can afford to provide the search service.

Such assumptions become self-fulfilling prophesies. The one thing that has remained constant is the assumption that to find something, the best approach is to describe what you want by words. But these assumptions face disruption, given that text search alone cannot adequately address how online media and information are developing.

When news of Wolfram Alpha first started to appear, the immediate reaction in press and blogs was speculation whether it could be a Google killer. And in that context, the site is disappointing. But it wasn’t designed to retrieve text. Instead, it focuses on the world of numbers: quantities, formulae, relationships. It can understand how numbers work with each other, which is something that is foreign to a text search engine like Google’s. The real threat that Wolfram Alpha offers Google is not replacing what the search king does, but offering a new way to think about finding information. Get people considering other approaches and options, and suddenly a dominant company no longer has so certain a lock on the public.

The association between search and verbal expression is an invisible burden, because it effectively forces you to use a single tool, no matter what job you’re trying to accomplish. How do you describe the first few bars of the major theme of Mozart’s Jupiter Symphony if you couldn’t remember the composer or the piece? The answer is inaccurately, because music does not lend itself to verbal description, unless you are using words to name the key and the notes, including their pitch and duration, then you are talking about the music, not of it.

What you would need to do is reproduce the tune – maybe by humming or possibly by typing it into a music keyboard. The search engine would need the capability to take in a melodic line and compare it to an index of a large body, possibly with everything reduced to the same key for ease of searching and the initial note acting as a reference line, with further notes expressed as intervals from it and some tolerance to get music that is “close to” what was described. Can you imagine such capabilities added on to an online music store? “I want the song that goes something like this.”

We’re already beginning to see such changes in the tools available for how people search. For example, TinEye, from Canadian company Idée, is billed as a “reverse image search engine”: you provide the sample image and it will check its indexes for potential matches, using multiple digital signatures of any image so a user can find complete or even partial matches. That would seem idea for copyright owners looking to find infringers, but there are more powerful applications.

The company already is developing a service, to first be available on iPhones, in which a person can take an image of a CD cover and then get pricing and sites for the product. The company expects to expand this to books, movies, and games.

As Michael Hickins and I have reported here, Everyzing has an approach to searching video by doing a voice-to-text translation of the accompanying audio in real time, allowing people to perform a text search to find a particular part of the video. (Google is also working on an audio indexing mechanism.) That’s fine, but what about finding video when one might not know exactly what was said, or when there isn’t any recognizable dialog? A similar ability to search the video content of frames might come in handy.

The beta version of Spezify.com is another example of how companies are trying to break the assumed link between search and text. Specify a term and you get a mix of text and images. And, as is true for images, they often are worth a thousand words.

It must be clear that the old way of slapping text up onto a screen isn’t going to be enough to keep people loyal to a given search engine, particularly as the masses of links become impossible to sort through, making bulk a less impressive differentiating feature. The old purveyors of search are beginning to realize this:

• Microsoft’s new search engine may attempt to better understand what people are typing to improve user experiences.
• Google Squared is an attempt to find new ways to display text, to make results more useful.
• Yahoo wants to shift from a document view to structured objects as search results.

But all of these are refinements to the existing text model. Even the presumptions of how to create a viable business model may have to change. Look at Twitter, which, in one sense, you might call text search waiting to happen, as people want to move beyond the restrictions of real time streaming. The company is saying that it doesn’t plan to sell ads against the streams of content. If you think about it, trying to jam in an ad in and around every 140-character message would be overwhelming. (And can you imagine trying to deliver such information onto a phone?) What the company plans, among other things, is making analytic tools available to companies, and to charge for inclusion in a database of commercial accounts. Regardless what you think of the concepts themselves, they do show that management there is considering what might exist other than online ads.

Perhaps the underlying model of text search, and not Google specifically, may be what has to be swept away to allow a new generation of search, not just another round of infants born wearing grey flannel suits.

[My colleague, Michael Hickins, has an interesting take on whether Microsoft could beat Google in a new game involving natural search. He mentions a David and Goliath strategy. Just as a caution, I'd point out that by all accounts, David had significant help on his side.]

Magnifying glass image via stock.xchng user thiagofest, standard site license.