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

Elevating original reporting in Search

blog.google

Post reply on HN
Print view View original

31–40 of 163 posts

Re: Elevating original reporting in Search

#31

Post by dexen »

danso wrote: post #29
dexen wrote: post #20

Earlier quoted context omitted.

>Isn't the original PageRank (...) defined by what Google engineers believe to be a relevant and informative webpage? No, quite the opposite. The original PageRank was defined by what other internet publishers considered to be a relevant and informative webpage. Specifically, the original PageRank was weighting content by the links (and content around the links) it was getting. The trust was placed on the internet pu…

Right, I understand your point, and maybe I'm being too reductive here, but the idea heuristic that what people link to is authoritative is an opinionated decision that ended up being satisfactorily accurate (and efficient to implement). And evaluating the heuristic's accuracy ultimately boils down to human judgment; from their early Google proposal (though not the original Pagerank paper): http://infolab.stanford.ed…

>but the idea heuristic that what people link to is authoritative is an opinionated decision

There are two decision spaces here. One is the selection of search ranking algorithm (whether to use PageRank or any other), and that decision was taken by Google[1].

The second decision space is the decision whether to link (or not) to any given document, and what content to put around it. That is a long lasting iterative process rather than any singular decision. Arguably it's geared towards approximating the distributed consensus - which is a global mutable state; expressed with edges the graph rather than any singular node.

Can the process be gamed, perverted, or corrupted randomly or maliciously? Sure. Do we know any better one? Not yet, at least not in the public.

--

[1] practically speaking we all selected PageRank, by preferring Google's over multiple competing engines.

Re: Elevating original reporting in Search

#32

Post by enlyth »

Has anyone found that recently Google search is becoming more and more useless?

Google tends to ignore what I actually type in, and tries to search according to some weird NLP machine learning inference on what it thinks I'm actually trying to ask.

Top results will include maybe 50-75% of the words I actually typed in, and it will treat the rest as mere hints or related words.

My queries end up looking like this after several tries and fails:

"something" "another phrase" "also" "this"

If I type the whole phrase without quotes I just get a bunch of ads, blog spam, and irrelevant stuff that is pretending to be useful.

Hell, most dev-related queries will return shallow medium-style blog articles instead of SO / Github.

Re: Elevating original reporting in Search

#33

Post by daniel_reetz »

enlyth wrote: post #32

Has anyone found that recently Google search is becoming more and more useless? Google tends to ignore what I actually type in, and tries to search according to some weird NLP machine learning inference on what it thinks I'm actually trying to ask. Top results will include maybe 50-75% of the words I actually typed in, and it will treat the rest as mere hints or related words. My queries end up looking like this afte…

I see the same behavior. I think it's time to assemble a list of examples.

Re: Elevating original reporting in Search

#34

Post by oliveshell »

enlyth wrote: post #32

Has anyone found that recently Google search is becoming more and more useless? Google tends to ignore what I actually type in, and tries to search according to some weird NLP machine learning inference on what it thinks I'm actually trying to ask. Top results will include maybe 50-75% of the words I actually typed in, and it will treat the rest as mere hints or related words. My queries end up looking like this afte…

Absolutely-— for a year or two now.

These days, if I have success with even a mildly technical Google search, it’s because I’ve enclosed every single term in quotation marks and/or turned on verbatim search, which doesn’t even work that well.

Google is trying so hard to think for you that they’re ruining their core product.

Re: Elevating original reporting in Search

#35

Post by frogpelt »

enlyth wrote: post #32

Has anyone found that recently Google search is becoming more and more useless? Google tends to ignore what I actually type in, and tries to search according to some weird NLP machine learning inference on what it thinks I'm actually trying to ask. Top results will include maybe 50-75% of the words I actually typed in, and it will treat the rest as mere hints or related words. My queries end up looking like this afte…

Same here.

Often, I have to rephrase my search 5-10 times before I see a result worth clicking on.

Re: Elevating original reporting in Search

#36

Post by brlewis »

daniel_reetz wrote: post #33
enlyth wrote: post #32

Has anyone found that recently Google search is becoming more and more useless? Google tends to ignore what I actually type in, and tries to search according to some weird NLP machine learning inference on what it thinks I'm actually trying to ask. Top results will include maybe 50-75% of the words I actually typed in, and it will treat the rest as mere hints or related words. My queries end up looking like this afte…

I see the same behavior. I think it's time to assemble a list of examples.

I bet googlers on HN who work on ranking would appreciate that. Here's a recent one from me:

  typical tcp packet size

Re: Elevating original reporting in Search

#37

Post by sweeneyrod »

danso wrote: post #7
sverige wrote: post #3

Am I the only one who thinks it's a bad idea for a tech company with no experience or education in the field to have so much influence over what people perceive to be good journalism?

Isn't the original PageRank, and every iteration since, defined by what Google engineers believe to be a relevant and informative webpage? They have a 160+ page guide that tells their ratings team how to identify high-quality content: https://static.googleusercontent.com/media/guidelines.raterh...

Yes, PageRank makes assumptions about what good content is and furthermore those assumptions are probably inaccurate in a lot of cases (dishonest clickbait can have a lot of backlinks). But it is unbiased in a sense that using human raters following guidelines is not. Moving from a purely algorithmic approach is a step away from being a neutral delivery service towards being a publisher, which carries extra responsibilities if you have as much power as Google.

Re: Elevating original reporting in Search

#38

Post by CDSlice »

brlewis wrote: post #36
daniel_reetz wrote: post #33

Earlier quoted context omitted.

I see the same behavior. I think it's time to assemble a list of examples.

I bet googlers on HN who work on ranking would appreciate that. Here's a recent one from me: typical tcp packet size

What exactly do you mean by that? When I search it up I get a SO answer saying what the max packet size is and another (closed) SuperUser answer that says that their is no average tcp packet size because that depends on the specific application. Both of those seem pretty good to me, but I'm not a low level networking guy.

Re: Elevating original reporting in Search

#39

Post by petard »

enlyth wrote: post #32

Has anyone found that recently Google search is becoming more and more useless? Google tends to ignore what I actually type in, and tries to search according to some weird NLP machine learning inference on what it thinks I'm actually trying to ask. Top results will include maybe 50-75% of the words I actually typed in, and it will treat the rest as mere hints or related words. My queries end up looking like this afte…

Do you experience this for dev-related searches or all types? I often have trouble finding relevant content for dev-related searches due to SEO content farms that wrap Stack Overflow content in an ad-infested user interface or worse.

Re: Elevating original reporting in Search

#40

Post by huffmsa »

enlyth wrote: post #32

Has anyone found that recently Google search is becoming more and more useless? Google tends to ignore what I actually type in, and tries to search according to some weird NLP machine learning inference on what it thinks I'm actually trying to ask. Top results will include maybe 50-75% of the words I actually typed in, and it will treat the rest as mere hints or related words. My queries end up looking like this afte…

Those blog style articles are the "original reporting" they're trying to promote.
Post reply on HN