Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Shirt Without Stripes

github.com

Post reply on HN
Print view View original

311–320 of 639 posts

Re: Shirt Without Stripes

#311

Post by DenisM »

js2 wrote: post #299
derivativethrow wrote: post #288

Earlier quoted context omitted.

Consider the query, "non-glass skyscrapers", which suffers from the same problem. What do you call a skyscraper like that if you want to refer to it? They exist, but you can't find them using that search term on Google.

I'd call it windowless: https://www.google.com/search?q=windowless+skyscraper&tbm=is...

You missed the Seattle Tower Building. It has windows, but very little in the way of visible glass.

https://www.emporis.com/buildings/119453/seattle-tower-seatt...

Windowless is a superset of glassless.

Re: Shirt Without Stripes

#312

Post by Miraste »

jjnoakes wrote: post #285
londt8 wrote: post #272

Earlier quoted context omitted.

I tried to search for "cheese without holes" on Google and it yielded good results. I think the problem here is that the query is something people would rarely search.

I just searched google images for "cheese" and "cheese without holes" and I got roughly the same results (about 1/3 of the images had holes in both cases).

"pictures" and "pictures without color" show that it does get some of these, although not the way I expected.

Re: Shirt Without Stripes

#313

Post by holler »

nikanj wrote: post #305
seiferteric wrote: post #137

I have noticed in the past few years google results have become noticeable worse for similar reasons. Google used to _surprise_ me with how good it was able to understand what I was really looking for even when I put in vague terms. I remember being shocked on several occasions when putting in half remembered sentences, lyrics, expressions from something I had heard years ago and it being the first! result. I almost…

Your search for "skiing Norway" mostly returns results for skiing in the French Alps, because those pages have much higher visit rates. Google is a dumbass nowadays, and regularly ignores half your search terms to present you with absolutely irrelevant results, that have gotten lots of visits in the past.

I've noticed this too, and frequently wonder why there aren't new and better search startups...

Re: Shirt Without Stripes

#314

Post by animalCrax0rz »

atq2119 wrote: post #309
animalCrax0rz wrote: post #259

Earlier quoted context omitted.

"there is no evidence of cancer" and "there is evidence of no cancer" are two different statements with different meaning, so it's more complex a task than just understanding the importance of "no" in a sentence. It's involves semantic analysis of the sentence. The paper I linked to below describes a technique they call "deep parsing." Check it out for more context.

This is true, but isn't really that relevant to the parent's point about statistical methods. Statistical methods (and "deep learning" is such a method) could certainly take the order of words in a sentence into account, for example.

Which part of "The paper I linked to below describes a technique they call "deep parsing." Check it out for more context." could you not parse?

I mean if us humans have a difficulty parsing each other's statements, then why should machines do any better?

Re: Shirt Without Stripes

#315

Post by DenisM »

animalCrax0rz wrote: post #225
DenisM wrote: post #197

This problem is known as "attribution" - you have a "no" or "without" in the sentence, but you don't know where it belongs. One could (and one does) argue that the problem cannot be solved with statistical methods (ML), especially not in any domain where accuracy is required, such as medical recored analysis: "no evidence of cancer" and "evidence of no cancer" are very different things. Zooming out, the language fiel…

There is this, too, from 4 years ago, which seems reasonable in my not-very-well-informed opinion (speaking of which, I'm not sure the work referenced here can deal with hyphenated negation, but it should be simple to include) https://www.aclweb.org/anthology/P14-1007.pdf code: https://github.com/ffancellu/NegNN

Thank you, I knew my effort was not for naught!

Re: Shirt Without Stripes

#317

Post by hiram112 »

floatingatoll wrote: post #273
seiferteric wrote: post #137

I have noticed in the past few years google results have become noticeable worse for similar reasons. Google used to _surprise_ me with how good it was able to understand what I was really looking for even when I put in vague terms. I remember being shocked on several occasions when putting in half remembered sentences, lyrics, expressions from something I had heard years ago and it being the first! result. I almost…

Google was good at launch because it was harvesting data from webrings and directories to provide it "high quality" link ranking data. However, they didn't thank or credit or share any of their revenue with the sites whose human curation helped their results become so impressive. Seeing that Google search was effective, most human curators stopped curating directories and webrings. The SEO industry picked up the slac…

> Google has not yet discovered how to automate "is this a quality link?"

Google has so much more data than just the keywords and searches people make, it seems like this should be a problem they could solve.

Through tracking cookies (e.g. Google Analytics) they should be able to follow a single user's session from start to finish, and they also should be able to 'rank' users in some vague way where they'd learn which users very rarely fall for ads or spend time on the sites that they know are BS. Those sites that are showing up on page 5 or 6 of the search results, but still get far more attention than others on the first few pages, could get ranked higher.

But I don't think many of Google's problems these days are technical in nature. They're caused by the MBAs now having more power at Google than the techies, and thus increasing revenue is more important than accuracy.

Re: Shirt Without Stripes

#318

Post by cowmoo728 »

bartread wrote: post #286
tines wrote: post #261

Earlier quoted context omitted.

I think the skewing of results lessening your impressed-ness is the wrong takeaway. If anything, the AI is a more perfect mirror of the society it learned from than you expected. Perhaps the right way to look at it is that we are capable of producing things that we don't understand, that are more sophisticated than we realize.

You may be right. It's been bugging me since I posted earlier on so I fired up a VPN with an endpoint in Japan, along with a private browsing session in Firefox, to see if I got different results. As it happens the results were interesting : - If I entered "person" I'd see a mix of images substantially similar to what I saw using google.co.uk up to and including Terry Crews, which was frankly a little weird, and othe…

This seems like you're attributing motive to google here, but I don't believe that's right. For example, Terry Crews appears in the query "person" because his "TIME Person of the Year 2017 Interview" article was very popular online. I get a lot of Greta Thunberg because she was TIME Person of the Year 2019 and received similar online attention because of Donald Trump.

The TL;DR of it is that google crawls the internet for photos, associates those photos with text content pulled from the caption or from the surrounding page, and gives them a popularity score based on the popularity of the page/image. There are some cleverer bits trying to label objects in the images, but it's primarily a reflection of how frequently that image is accessed and how well the text content on the page matches your query. There's some additional localization, anti-spam, and freshness rating that influences the results too.

The majority of pages with "人" and a photo on it that has a machine labeled person image would be a photo of a japanese/chinese person, and if you're being localized to japan with a vpn, that would be even more true.

Google doesn't "know" what you're trying to search. It's a giant pattern matching game that slices and dices and rearranges text to find the closest match.

Re: Shirt Without Stripes

#319

Post by BaronSamedi »

Tade0 wrote: post #81

My Operating Systems professor (Tomasz Jordan Kruk, PhD) in college had an appropriate anecdote for this: "Humans usually don't intuitively understand the word 'no'. Please imagine a non-pink elephant."

The funny thing is that this search is pretty easy to put into first order logic (shirt(x) & ~striped(x)). I guess we now have computers that are bad at logic.

Re: Shirt Without Stripes

#320

Post by mcbits »

smoyer wrote: post #266
DenisM wrote: post #197

This problem is known as "attribution" - you have a "no" or "without" in the sentence, but you don't know where it belongs. One could (and one does) argue that the problem cannot be solved with statistical methods (ML), especially not in any domain where accuracy is required, such as medical recored analysis: "no evidence of cancer" and "evidence of no cancer" are very different things. Zooming out, the language fiel…

While this is indeed an example of the attribution problem, I'd argue that this particular query will never be solved. I don't search for a "shirt without stripes", I search for a "solid shirt, or a " hawai'ian shirt". I'd be curious to see how many sentences with attribution problems actually have other structural issues. If I want to write clearly and without ambiguity, I rewrite sentences that have these problems.…

I don't think this particular problem is related to the language model. "[item] without [attribute]" is trivial to understand even without a sophisticated language model.

The bad results are because they're not positively indexing the absense of the feature by deeply analyzing the images or products beyond the descriptions. "Shirt with stripes" yields almost exclusively striped shirts. Exclude those results from all "shirts" and there are still a lot of striped shirts that the search algorithm doesn't know enough to exclude.

Post reply on HN