Re: [RAS] test on ras.openlib.org
On Fri, Nov 18, 2011 at 7:40 PM, Ivan Kurmanov <duraley@gmail.com> wrote:
On Fri, Nov 18, 2011 at 4:57 PM, Christian Zimmermann <zimmermann@stlouisfed.org> wrote:
I still get no suggestion. But adding another name gives me more, all exact name matches, all with a 'no'
So it looks like the system does not offer suggestions with the last name any more, only exact names, and offers a 'no' for all of them.
Christian, please, test & review -- fixed now. -ivan
I will fix that.
-ivan
Ivan Kurmanov writes
On Fri, Nov 18, 2011 at 7:40 PM, Ivan Kurmanov <duraley@gmail.com> wrote:
On Fri, Nov 18, 2011 at 4:57 PM, Christian Zimmermann <zimmermann@stlouisfed.org> wrote:
I still get no suggestion. But adding another name gives me more, all exact name matches, all with a 'no'
So it looks like the system does not offer suggestions with the last name any more, only exact names, and offers a 'no' for all of them.
Christian, please, test & review -- fixed now.
How is this being fixed? Cheers, Thomas Krichel http://openlib.org/home/krichel http://authorprofile.org/pkr1 skype: thomaskrichel
On Wed, Nov 23, 2011 at 12:29 AM, Thomas Krichel <krichel@openlib.org> wrote:
Ivan Kurmanov writes
On Fri, Nov 18, 2011 at 7:40 PM, Ivan Kurmanov <duraley@gmail.com> wrote:
On Fri, Nov 18, 2011 at 4:57 PM, Christian Zimmermann <zimmermann@stlouisfed.org> wrote:
I still get no suggestion. But adding another name gives me more, all exact name matches, all with a 'no'
So it looks like the system does not offer suggestions with the last name any more, only exact names, and offers a 'no' for all of them.
Christian, please, test & review -- fixed now.
How is this being fixed?
Now -- by making Yes the default choice for exact matches. Not based on the relevance column, which is always zero in an environment without learning. This change is not compatible with your ACIS version. And there's a whole other bunch of incompatible changes, like switching back to HTML output in XSLT versus XHTML. And the backwards namespaces simplification: I have mostly eliminated the XHTML namespace use. I'm sorry about that, but I couldn't get some javascript processed otherwise. (Not sure if /adm/events javascript works on your installations, for example, and if it does -- how.) -ivan
Ivan Kurmanov writes
Now -- by making Yes the default choice for exact matches. Not based on the relevance column, which is always zero in an environment without learning. This change is not compatible with your ACIS version.
I am not impressed. I trust that this will only affect the presentation? Or is this is the perl code? Could this not be fixed by introducing a configuration paremented?
And there's a whole other bunch of incompatible changes, like switching back to HTML output in XSLT versus XHTML. And the backwards namespaces simplification: I have mostly eliminated the XHTML namespace use. I'm sorry about that, but I couldn't get some javascript processed otherwise. (Not sure if /adm/events javascript works on your installations, for example, and if it does -- how.)
I have never used /adm/events. But I am surprised it should be difficult to get this to work with a proper DOM approach. Cheers, Thomas Krichel http://openlib.org/home/krichel http://authorprofile.org/pkr1 skype: thomaskrichel
On Thu, Nov 24, 2011 at 12:45 AM, Thomas Krichel <krichel@openlib.org> wrote:
Ivan Kurmanov writes
Now -- by making Yes the default choice for exact matches. Not based on the relevance column, which is always zero in an environment without learning. This change is not compatible with your ACIS version.
I am not impressed. I trust that this will only affect the presentation? Or is this is the perl code? Could this not be fixed by introducing a configuration paremented?
To be honest, I'm not impressed either. Yes, the change is in presentation, two xslt files. Yes, it could be done more elegantly via a configuration parameter. But we a very short on time. aras@holda:~/dev/acis.r$ darcs diff -p "research/suggestions: back to showing the default choice based on status, not on relevance. incompatible with the learning, etc." | less would show you the diff of that change.
And there's a whole other bunch of incompatible changes, like switching back to HTML output in XSLT versus XHTML. And the backwards namespaces simplification: I have mostly eliminated the XHTML namespace use. I'm sorry about that, but I couldn't get some javascript processed otherwise. (Not sure if /adm/events javascript works on your installations, for example, and if it does -- how.)
I have never used /adm/events. But I am surprised it should be difficult to get this to work with a proper DOM approach.
You never used /adm/events, and that is fine. But Christian and I use it. And your change from HTML to XHTML broke it. I failed to find the way to make it work with xhtml output, without ugly constructs, that would render the javascript pieces unreadable. If you find it, i'm fine. (I just don't see any benefit in xhtml versus html to spend any more time on this.) -ivan
On Thu, Nov 24, 2011 at 10:52 AM, Ivan Kurmanov <duraley@gmail.com> wrote:
On Thu, Nov 24, 2011 at 12:45 AM, Thomas Krichel <krichel@openlib.org> wrote:
Ivan Kurmanov writes
Now -- by making Yes the default choice for exact matches. Not based on the relevance column, which is always zero in an environment without learning. This change is not compatible with your ACIS version.
I am not impressed. I trust that this will only affect the presentation? Or is this is the perl code? Could this not be fixed by introducing a configuration paremented?
To be honest, I'm not impressed either. Yes, the change is in presentation, two xslt files.
Actually, there are also javascript changes. The change is: Wed Nov 23 23:42:09 CET 2011 duraley@gmail.com * fix the green / red, i.e. yes / no highlighting, when working with suggestions -i
Ivan Kurmanov writes
Actually, there are also javascript changes. The change is:
Wed Nov 23 23:42:09 CET 2011 duraley@gmail.com * fix the green / red, i.e. yes / no highlighting, when working with suggestions
ok, these JS and XSLT changed are changes to the default presentation? I suggest to open a RAS presentation in acis.r as additional file sets, and then apply the changes to that presentation. Could you do that? Cheers, Thomas Krichel http://openlib.org/home/krichel http://authorprofile.org/pkr1 skype: thomaskrichel
Ivan Kurmanov writes
To be honest, I'm not impressed either. Yes, the change is in presentation, two xslt files.
Yes, it could be done more elegantly via a configuration parameter.
If you can't get this done, I will do it in the Summer.
But we a very short on time.
There are two issues I have with this. The lack of a parameter is one. The other one is Christian's disinterest in learning. He heaps praise on the learning in NEP, but he does not want to have it in RAS. The RAS data has mistakes that the learning will have find and/or prevent. The learning is a a better predictor of relevance that the old full/partial name. But I can't demonstrate these claims.
aras@holda:~/dev/acis.r$ darcs diff -p "research/suggestions: back to showing the default choice based on status, not on relevance. incompatible with the learning, etc." | less
would show you the diff of that change.
thank you!!
I failed to find the way to make it work with xhtml output, without ugly constructs, that would render the javascript pieces unreadable.
and you would never ever write a piece of unreadable code, not even under threat of torture ;-) Cheers, Thomas Krichel http://openlib.org/home/krichel http://authorprofile.org/pkr1 skype: thomaskrichel
On Thu, Nov 24, 2011 at 5:32 PM, Thomas Krichel <krichel@openlib.org> wrote:
There are two issues I have with this. The lack of a parameter is one. The other one is Christian's disinterest in learning. He heaps praise on the learning in NEP, but he does not want to have it in RAS. The RAS data has mistakes that the learning will have find and/or prevent. The learning is a a better predictor of relevance that the old full/partial name. But I can't demonstrate these claims.
My issue with learning in ACIS is lack of documentation describing it. If I were to enable it in RAS, I'd be at a loss on what's needed to make it run, and how it has to be configured. Maybe it is documented somewhere, but I missed it? Similarly, your use of sub-project code names, like pitman, cardiff, etc. in ACIS code is a pain for me. Are they explained somewhere? (They are still useful as they are, as they mark related pieces of your code, but they are annoying like signs in a foreign language i don't have a dictionary for.) (And I know you have been frustrated by lack of documentation for my code.) -i
participants (2)
-
Ivan Kurmanov -
Thomas Krichel