Udi Manber wrote a blog post introducting search quality at Google. As Udi says, we are “quite secretive about what we do” and this is a nice, simple summary of the sorts of thing we work on and what we think about. I’m on the ranking team but have spent more and more of my time over recent years thinking about the intersection between ranking and user interface.
Category Archives: Search
Language models
I don’t usually talk in public about the technical side of what I do at work, but I co-wrote a post for the company blog a few weeks ago on how we build and use language models which I should have linked to earlier.
An Editorial Voice
Adolph S. Ochs, New-York, Aug. 18, 1896:
To undertake the management of The New-York Times, with its great history for right doing, and to attempt to keep bright the lustre which Henry J. Raymond and George Jones have given it is an extraordinary task. But if a sincere desire to conduct a high-standard newspaper, clean, dignified, and trustworthy, requires honesty, watchfulness, earnestness, industry, and practical knowledge applied with common sense, I entertain the hope that I can succeed in maintaining the high estimate that thoughtful, pure-minded people have ever had of The New-York Times.
It will be my earnest aim that The New-York Times give the news, all the news, in concise and attractive form, in language that is parliamentary in good society, and give it as early, if not earlier, than it can be learned through any other reliable medium; to give the news impartially, without fear or favor, regardless of party, sect, or interests involved; to make of the columns of The New-York Times a forum for the consideration of all questions of public importance, and to that end to invite intelligent discussion from all shades of opinion.
There will be no radical changes in the personnel of the present efficient staff. Mr. Charles R. Miller, who has so ably for many years presided over the editorial pages, will continue to be the editor; nor will there be a departure from the general tone and character and policies pursued with relation to public questions that have distinguished The New-York Times as a non-partisan newspaper — unless it be, if possible, to intensify its devotion to the cause of sound money and tariff reform, opposition to wastefulness and peculation in administering public affairs, and in its advocacy of the lowest tax consistent with good government, and no more government than is absolutely necessary to protect society, maintain individual and vested rights, and assure the free exercise of a sound conscience.
See more in History of the New York Times, 1851-1921 and and Discovering the News: A Social History of American Newspapers.
As He Did Think
I finally read Vannevar Bush’s essay As We May Think for the first time this week. For something written in 1945, it’s amazing; for that matter, if it had been written in 1975, it would have been just as amazing for its uncanny predictive power. He outlines something very close to the modern digital era. If you haven’t read it, you should.
Of course, it’s off on many details, comically so in some cases. His discussion of “dry photography” and the process of distributing books in microfilm form remind me of the pneumatic tubes of Brazil. With transistors still two years out, I guess “thermionic tubes” were the right technology to talk about; that they’re now used only by die-hard audiofiles might surprise him.
Where Bush falls short about technology is in not predicting the pervasiveness and connectedness that we have. Yes, scientists and researchers use modern day “Memexes,” but so do people looking for people, movies, restaurants, travel, trivia, porn, and a million other topics. The information in our memexes is distributed among a wider array of machines, all connected, giving a much larger field of information available to everyone. We also use the internet for forms of communication — blogs might have been predictable but eBay probably wasn’t — that I don’t think Bush envisioned.
Bush was better on technology than social trends. He didn’t foresee shifts in gender roles; these days, scientists type for themselves and “a whole roomful of girls armed with simple key board punches” are not transcribing the thoughts of great men. He predicted that books would be the unit of transfer, where the web “page” model is much finer grained. That almost all of us are still typing, rather than speaking for human- or machine-transcription, is an artifact of something else that I think is hard to predict: when we’ll adapt machines to our behavior and when we’ll adapt our behavior to what machines do easily.
But all that’s incidental to the astounding accomplishments of prediction in this essay. Search engines are trying to deliver on the potential of the Memex and he described information retrieval better than most people can today. The combination is digital photography and “radio” (or, as we think of it, “wireless”) is probably ahead of where he predicted and book digitization is almost there. His description of browsing and navigation make an interface of windows, scrollbars, and a pointing device (though not a “lever”) seem almost obvious.
(Props to The Atlantic for being true to the spirit of the essay by making it easily accessible.)
At least the internet never turns into a pile of grey sludge on your doorstep
This is rainy season in San Francisco. It’s also, unfortunately, the time of year when Pacific Bell SBC at&t delivers new yellow pages.
I just about stopped using the yellow pages nearly a decade ago, long before I started working at a search engine and long before there was good integration of local information with general searches. Certainly by the time I had always-on internet access at home, I gave up using a printed yellow pages except in the rarest of cases. If the local business has a website — almost always true for a restaurant, for example — and you can find it, the web is great. If it doesn’t, the presence of online yellow pages means you’ll at least get the basic contact information and, in some categories, third party reviews and discussion.
On the other hand, I have at least one friend who swears by the physical yellow pages these days. He loves how easy it is to find the big, credible players, because they buy display ads. And those big ads contain lots of information, often including open hours, manufacturers whose products the store carries, a map, and details that might give you a feel for the business. Exactly what you’d hope to find on a website.
Many of those display ads are placed by local businesses that don’t have a website. For example, one of our local hardware stores (Tuggey’s on 24th) has no website and the other (Cliff’s Variety on Castro) appears to have added a website only in the past few weeks. (Go, Cliff’s!)
So, when this year’s yellow pages turned to a pile of liquidy grey sludge after a couple of hours of waiting for us on the front steps, I wasn’t particularly disappointed. What surprises me is the people who would still be disappointed. And more surprising are the merchants who spend a significant amount of money to reach those people, but don’t even attempt to reach people like me.