Re: [tlug] Japanese input method

[Raymond Wan <rwan.kyoto@example.com>]

On Mon, Jul 13, 2015 at 10:29 AM, Stephen J. Turnbull
<stephen@example.com> wrote:
> Raymond Wan writes:
>  > I'm not too sure if circular linking has anything to do with it.
> I'm pretty sure it does by now.  Just watch an unfortunately prominent
> subset of the Chinese academic community "circle jerk" their citation
> counts into the stratosphere.[1]  My students aren't good enough to
> tell the good scholars from the bad, so I just have to tell them to
> basically ignore papers in their native language because citation
> indicies and impact factors are almost completely unreliable, at least
> in the social sciences.  This is unfortunate for the many excellent,
> honest scholars who can't get the exposure they deserve.


There are surely problems in the sciences as well.  If a reliable
method ever gets developed for measuring research impact, surely
someone will be spending time to try to cheat it...

Your last statement is surely applicable to science and engineering...
I often read the RetractionWatch web site for entertainment.  :-)


> The exact same principles apply to PageRank (impact factors are
> basically a variation on the theme of the PageRank algorithm).


Yes -- I think prior to PageRank and Google, many search engines was
just looking at index searching and not the link structure between
documents.  So, in the IR field, it came at the right time, too.

Mind you, the whole idea of a highly reputed person knowing someone
else and, thus, raising their reputation existed long before computers
on primary / high school playgrounds.  :-)


>  > Of course, this was the original 1998 pagerank algorithm (up to my
>  > level of understanding) -- over the years, no doubt Google has
>  > updated it.
>
> One of my students investigated last year.  As far as he could find,
> Google never published their actual algorithm, and said of several
> algorithms "that's not it, we substantial have improvements".  So, as
> far as I know "Google PageRank" is a trademark, not an algorithm. :-)


In 1998, there was a flurry of papers from Brin and Page.  Like this
technical report [I just found it and it does seem to have a figure
about circular linking...I'm not going to bother reading it though ;-)
]:

http://ilpubs.stanford.edu:8090/422/1/1999-66.pdf

I remember they had a paper in WWW 1998 or something.  But I think all
that did was give them seed funding to start Google.  Once that
started, they could improve on the algorithms without publishing.
Google staff publish in academic conferences, but I *guess* they
publish it years after it's developed...


>  > The part I don't know is if a poorly ranked web page (i.e., mine) is
>  > A, does it contribute very little to page B or does it penalize page
>  > B.  Of course, if it contributes a little, that's no big deal.
>  > Penalizing wouldn't be any good.
>
> A low rank won't cause a penalty in any version of PageRank I've seen
> (all coefficients are positive, so more is better).  But it's possible
> that there are "negative ranks" in Google's algorithm, I suppose.  I
> don't see why you'd get one of those.


Yes, I guess so, too.

Ray