Goo Goo Googly

[nancylebov]

A discussion of weird word count behavior at Google.

It seems reasonable that we have to use deduction to deal with nature--nature doesn't talk. It's almost reasonble that we have to use deduction for human institutions--not only are people frequently too busy to explain what's going on, but when you've got lots of them, you have emergent effects. It does *not* seem reasonable to have to do experiments to figure out how a computer program reacts to its simplest commands.

Link snagged from stoutfellow

Comments

[rysmiel.livejournal.com]

It doesn't seem unreasonable to me - there are a lot of classes of machine learning algorithm, that one trains rather than programming at a level explicitly specifying their goals and how to achieve them, which remain black boxes in terms of how they are doing what they are doing. There's a fair amount of work being done with them in bioinformatics - I do not like them myself, but there are many problems where they work rather well.

[nancylebov.livejournal.com]

Maybe I've got an excessive sense of entitlement about this, but I feel as though Google is in a better position to know what their commands do, and should tell people instead of having them use the search engine as an ever-changing black box.

It's possible that the problem is much harder than I realize and/or that Google has better reasons for keeping even basic facts about its structure secret.

[madfilkentist]

There are two kinds of users -- those who have only the vaguest idea of how a computer works, and no concept of a logical search strategy (aka them), and those who approach computers logically and know what they're doing (aka us). There are a lot more them's than us's, and Google has doubtless tuned its heuristics based on observation of the strange search habits of the them's.

[nancylebov.livejournal.com]

I don't think that's it--I suspect it's a combination of google being such a large system that it's genuinely not predictable and google keeping a good bit secret to make it harder for businesses to game the searches.

I don't see how trying to accomodate weird user behavior could lead to the results from the orgininal article about "hat".

[thnidu.livejournal.com]

It's not just use by the knowledgeable or the non-knowledgeable. I happen to know something about the population of users that the Language Log is addressed to, because Mark Liberman, who runs it, is my boss, and many of the regular posters, like Geoff Nunberg and Arnold Zwicky, are prominent linguists whose work or postings I have known for years. Google's numbers are weird. I'm not going to try to summarize it now; if you are really interested, this topic has been running for awhile on Language Log.

-- Dr. Whom, Consulting Linguist, Grammarian,
Orthoepist, and Philological Busybody
a.k.a. Mark A. Mandel
[This text prepared with Dragon NaturallySpeaking.]

[suecochran.livejournal.com]

Hi, Mark - I didn't know that you were here :) I just friended you. I would love to come up for the housefilk - I will have to see what the details are. My e-mail is sue at fiawol dot org. I hope to hear from you! Hugs, Sue