[ACT-R-users] Question on new visual objects and accessing in productions

Shawn Nicholson shawn.nicholson at dnamerican.com
Fri Jul 1 20:41:38 EDT 2005


Great! thanks for the pointers to those docs - exactly what I needed.  And as for the feature values being keywords, yes that was actually a typo on my part.  In the code they are keywords :friendly, :unfriendly and a couple others, but when I try to access them in productions I found that I could not use the keywords, I had to specify them as non-keywords (and they were found).  I will go back and get rid of the keyword specifications though.  
 
Thanks,
  Shawn

	-----Original Message----- 
	From: Dan Bothell [mailto:db30 at andrew.cmu.edu] 
	Sent: Fri 7/1/2005 4:52 PM 
	To: Shawn Nicholson; act-r-users at act-r.psy.cmu.edu 
	Cc: 
	Subject: Re: [ACT-R-users] Question on new visual objects and accessing in productions
	
	



	--On Friday, July 01, 2005 3:32 PM -0400 Shawn Nicholson
	<shawn.nicholson at dnamerican.com> wrote:
	
	>
	> Now, everything works.  But I am wondering if what I have done is the
	> proper method for manipulating specialized icon objects in the vision
	> module. My questions (in summary):
	>
	> 1)    Is this the proper way to be adding new visual object types to the
	> system for accessing special slots and such?  (ie adding that feat-to-dmo
	> method specialized on aircraft-icon??)
	>
	
	I think you mean build-features-for instead of build-group-features-for,
	but otherwise, yeah, that's basically how it's done:
	
	Write a build-features-for to return a custom icon-feature and then
	write a feat-to-dmo to convert that icon-feature into a dme/chunk.
	
	There are details in the docs at
	<http://chil.rice.edu/projects/RPM/docs/index.html>
	for others that are interested, in particular the device interface and
	vision module sections.
	
	
	> 2)    Those two lines that I placed in my model seem like they should be
	> somewhere in the code where the aircraft-icon feature is defined.
	> However, unless I put them in the model, anytime (clear-all) is called,
	> the chunk types go away and they no longer exist for use in the
	> productions.  I could put and advice function around the function
	> reset-vision-module (which is where all the basic vision chunks are
	> defined) but that doesn't seem to be a clean way to change things.
	>
	
	Right, clear-all (or a reset) deletes all chunk-type definitions.  The
	chunk-type should probably stay in the model. Generally, I would advise
	against modifying or extending the main functions of the vision module
	(or really any of the modules) unless it's really necessary.
	
	The definition of the chunk named aircraft-object isn't really
	necessary because it will get created automatically. [To those that
	don't recognize the define-chunks function, it's an ACT-R 6 command for
	creating chunks without putting them into DM.]  I'm assuming you got a
	warning that there wasn't such a chunk since you use it as the value
	for kind and then added that to avoid the warning.
	
	
	One general comment is that I would recommend against using keywords as
	slot values i.e. chunk names.  In previous versions of ACT-R it could
	lead to some really hard to debug code, for example, in ACT-R 5:
	
	CG-USER(3): (add-dm (g isa visual-location value foo kind :foo))
	(G)
	CG-USER(4): (dm g)
	 G    0.000
	    isa VISUAL-LOCATION
	    screen-x nil
	    screen-y nil
	    distance nil
	    attended nil
	    kind Foo
	    color nil
	    value Foo
	    size nil
	    nearest nil
	
	note that it doesn't show the difference in the output. While ACT-R 6
	will show that correctly in that case I can't guarantee that it will
	treat them correctly everywhere.
	
	Dan
	
	



More information about the ACT-R-users mailing list