POV-Ray : Newsgroups : povray.off-topic : Why semantic search reeks Server Time
5 Sep 2024 17:17:31 EDT (-0400)
  Why semantic search reeks (Message 1 to 10 of 21)  
Goto Latest 10 Messages Next 10 Messages >>>
From: gregjohn
Subject: Why semantic search reeks
Date: 26 May 2009 18:40:00
Message: <web.4a1c6f626b94f33f34d207310@news.povray.org>
Can anyone figure out how to google up the example files for shapes.inc, which
will tell me the syntax for using Rounded_Box_Things?

Of course, there probably isn't a Rounded_Box_Thing, but I want the syntax for
whatever there IS.  I have instead found a whole lot of incorporated companies
with the word "Shape" in their name.

I think that this shows why I hate semantic search.  The computer assumes all
the other things you could possibly want besides what you typed in verbatim,
and goes to dig them up.  Once I had a "buzzing" sound in my PS/3.  Searching
for "buzzing" sound and PS/3 brought up hundreds of entries for some service or
website called "Sony Buzz".

I think I need searches to be more literal, and less semantic.  Anyone know of a
way to make google more literalistic, or have a better search engine?


Post a reply to this message

From: Gilles Tran
Subject: Re: Why semantic search reeks
Date: 26 May 2009 19:28:01
Message: <4a1c7b01@news.povray.org>

news:web.4a1c6f626b94f33f34d207310@news.povray.org...
> I think I need searches to be more literal, and less semantic.  Anyone 
> know of a
> way to make google more literalistic, or have a better search engine?

First, use double quotes when you don't want Google to be too smart. 
"buzzing" ps/3 returns what you want.

In any case, think about specific keywords closely associated to your topic 
that are likely to be in your target pages and add them to your search. When 
I was giving lectures about internet search in the early 2000s I was always 
giving the following example: at that time, looking for swine feeding (I'm a 
specialist in farm animal nutrition) always returned pages with Matthew 8:30 
("Now there was a herd of many swine feeding at a distance from them") 
instead of technical documentation about pig farming. The trick: add 
"protein" to the search because this word is extremely likely to be used in 
the target documents and completely absent from the Bible. The example no 
longer works now but the trick never fails.

Note that Google semantic search seems pretty much in its infancy and more a 
buzzword than anything else: right now, the bulk of the search seems to be 
done after lemmatisation of the input (i.e. it goes back to the canonical 
form of the words) with some phonetic algorithm to find probable homophones.

G.


Post a reply to this message

From: Nicolas Alvarez
Subject: Re: Why semantic search reeks
Date: 26 May 2009 19:50:59
Message: <4a1c8062@news.povray.org>
gregjohn wrote:
> Can anyone figure out how to google up the example files for shapes.inc,
> which will tell me the syntax for using Rounded_Box_Things?
> 
> Of course, there probably isn't a Rounded_Box_Thing, but I want the syntax
> for
> whatever there IS.  I have instead found a whole lot of incorporated
> companies with the word "Shape" in their name.
> 
> I think that this shows why I hate semantic search.  The computer assumes
> all the other things you could possibly want besides what you typed in
> verbatim,
> and goes to dig them up.  Once I had a "buzzing" sound in my PS/3. 
> Searching for "buzzing" sound and PS/3 brought up hundreds of entries for
> some service or website called "Sony Buzz".
> 
> I think I need searches to be more literal, and less semantic.  Anyone
> know of a way to make google more literalistic, or have a better search
> engine?

Google tries to be smart and shows me results in my local language first.
Which, for programming stuff, are usually less useful results.

Another interesting one:
I challenge anyone to find useful information about COM, using Google.


Post a reply to this message

From: gregjohn
Subject: Re: Why semantic search reeks
Date: 26 May 2009 21:00:01
Message: <web.4a1c8f9a910979c334d207310@news.povray.org>
"Gilles Tran" <gil### [at] gmailcom> wrote:

> First, use double quotes when you don't want Google to be too smart.
> "buzzing" ps/3 returns what you want.
>

Hmmm.  That problem was actually a frustration from my memory of several months
ago.  Your idea works (I already *knew* that)   ;-) , but interestingly,
**today**, even plain ol'
    Buzzing sound ps/3
works sufficiently well without quotes. Could folks clicking on those X boxes
actually make google smarter?


> In any case, think about specific keywords closely associated to
> your topic that are likely to be in your target pages and add
> them to your search.

Okay, how about finding the syntax for shapes.inc's rounded thingies using
google?


Post a reply to this message

From: Darren New
Subject: Re: Why semantic search reeks
Date: 26 May 2009 21:25:47
Message: <4a1c969b$1@news.povray.org>
gregjohn wrote:
> I think that this shows why I hate semantic search. 

It's usually more of a problem when the search results you're looking for 
aren't actually indexed.

And my favorite screw-up was naming it "C# .NET"  Hey MS, could you possibly 
have picked a name for a technology that would be harder to search for? 
Maybe something consisting only of hyphens and spaces?

-- 
   Darren New, San Diego CA, USA (PST)
   There's no CD like OCD, there's no CD I knoooow!


Post a reply to this message

From: Darren New
Subject: Re: Why semantic search reeks
Date: 26 May 2009 21:27:06
Message: <4a1c96ea@news.povray.org>
Gilles Tran wrote:
> In any case, think about specific keywords closely associated to your 
> topic that are likely to be in your target pages and add them to your 
> search.

Or, alternately stated, remember that Google indexes answers, not questions. 
If you don't know what the answer looks like already, you're unlikely to 
find it.

-- 
   Darren New, San Diego CA, USA (PST)
   There's no CD like OCD, there's no CD I knoooow!


Post a reply to this message

From: Darren New
Subject: Re: Why semantic search reeks
Date: 26 May 2009 21:28:25
Message: <4a1c9739$1@news.povray.org>
Nicolas Alvarez wrote:
> Another interesting one:
> I challenge anyone to find useful information about COM, using Google.

Heh. Hmmm....

com site:msdn.microsoft.com

Does that count? ;-)

-- 
   Darren New, San Diego CA, USA (PST)
   There's no CD like OCD, there's no CD I knoooow!


Post a reply to this message

From: Gilles Tran
Subject: Re: Why semantic search reeks
Date: 27 May 2009 04:20:08
Message: <4a1cf7b8@news.povray.org>

web.4a1c8f9a910979c334d207310@news.povray.org...
>
> Okay, how about finding the syntax for shapes.inc's rounded thingies using
> google?

shapes.inc povray rounded example

G.


Post a reply to this message

From: Invisible
Subject: Re: Why semantic search reeks
Date: 27 May 2009 04:22:32
Message: <4a1cf848$1@news.povray.org>
Gilles Tran wrote:

> Note that Google semantic search seems pretty much in its infancy and 
> more a buzzword than anything else: right now, the bulk of the search 
> seems to be done after lemmatisation of the input (i.e. it goes back to 
> the canonical form of the words) with some phonetic algorithm to find 
> probable homophones.

When I was at college, my classmates were so stupid that they thought 
that a "homophone" was somebody who approves of gay killings. :-P


Post a reply to this message

From: Gilles Tran
Subject: Re: Why semantic search reeks
Date: 27 May 2009 04:31:43
Message: <4a1cfa6f@news.povray.org>

4a1c96ea@news.povray.org...
> Gilles Tran wrote:
>> In any case, think about specific keywords closely associated to your 
>> topic that are likely to be in your target pages and add them to your 
>> search.
>
> Or, alternately stated, remember that Google indexes answers, not 
> questions. If you don't know what the answer looks like already, you're 
> unlikely to find it
.
Yes, people tend to use general search terms as if the search engine was a 
real person with actual semantic abilities (and a real person is able to 
deduce or ask for missing context). Instead, I told them to imagine what 
sort of document would contain the answer and derive search keywords from 
that. I nicknamed that "empathy search" since one had to put oneself in the 
shoes of the writers of the target documents. Sometimes this requires doing 
preliminary searches for identifiying the proper vocabulary. All of this is 
less useful now because Google has become extremely efficient but that's 
still handy for difficult searches with a lot of noise.

G.


Post a reply to this message

Goto Latest 10 Messages Next 10 Messages >>>

Copyright 2003-2023 Persistence of Vision Raytracer Pty. Ltd.