Site Meter

April 06, 2009

Let's Stop Speaking Like Machines and Start Speaking Like People

The path from engineering to marketing is usually not a straight line. Often there can be many stops along the way, as a product goes from idea to a spec to prototype release build, through the quality assurance process, and eventually general availability to the marketplace. As the product develops, so does the way it gets named, branded and described - starting with what's typically a straight forward problem/solution issue from engineering, and morphing into a more refined, even if not always as accurate, pitch from marketing. But in the Valley, often we skip those steps, and it's our users who end up paying the price - by being taught to think and talk like machines.

Are you a big fan of hash tags? Or are you wild about boolean searches? Do you find yourself reverting back to "Run DOS Run" instead of just typing and talking like a human being often online? Despite billions of dollars of investment and a plethora of companies trying to develop natural language (especially in search), we still have a long way to go.


Twitter, the hot tech topic of the month and many others preceding it, largely relies on two specific machine language symbols to connect users. The first is the basic @ sign which signals a "reply". The second, a # mark, or "hashtag", tries to connect people talking about the same event, location or idea. But what we're doing by using these symbols is work the machines should do for us. Instead of posting a hashtag about our location or event, Twitter should pick that up based on our profile data or GPS from the phone, or even group people's topics based on the content contained in the tweet and those immediately preceding it.

Web search engines have similarly expected high levels of machine like language from all their users. For example, a search on Google that shows results that mention "dog" or "cat" or "fish" but don't have the word "bird" in them necessitates a search string of "dog OR cat OR fish -bird". If I wanted to demand it have both dog and cat in it, but not fish or bird, I'd be changing things up a bit, typing: "dog AND cat -bird -fish". We're talking like this because we're trying to make nice with a database who thinks this way.


Even in this morning's FriendFeed beta site do you see the same kind of expectation that pushes users away from being people and further along the path of being cyborgs. While the company has some helpful pull-down menus on its advanced search page, it doesn't let you search by specific services, such as YouTube or LinkedIn (while the old version did). Instead, you're expected to type in "service:youtube" in the search field. To search all my friends' posts from YouTube that contained the word pizza in the title, I'd have to set up the advanced search to look for pizza in the title from my friends, and then add "service:youtube" to the query.


I expect the FriendFeed team can fix that query fairly quick with the addition of a pulldown menu, but that will be knocking down one mole before another pops up somewhere else - and many other services are less responsive, expecting you to talk in a way a machine would. Jeremiah Owyang and Loic Le Meur exchanged tweets about a month ago, calling the Web "primitive". But the Web just turned 20 years old. If this is how far we've come in 20 years, do we have to wait another 20 before we can just type or speak in what we want to know and get the right result?