Lambda the Ultimate

inactiveTopic The Visual Language of Experts in Graphic Design
started 9/4/2001; 4:47:09 AM - last post 9/9/2001; 3:38:16 AM
Ehud Lamm - The Visual Language of Experts in Graphic Design  blueArrow
9/4/2001; 4:47:09 AM (reads: 328, responses: 1)
The Visual Language of Experts in Graphic Design
(via Camworld)

To investigate how designers communicate their knowledge, I surveyed how design knowledge is communicated to students in introductory design books. The results of this informal investigation were quite surprising to me. The books generally present copious selections of graphic examples of exemplary designs, accompanied by explanatory text. I was astonished by the difference in effectiveness of content between the illustrations and the text.

...The approach we take, then, is that expert systems from graphic design must learn by example. We have been investigating the technique of programming by example, [Lieberman 92, 93] as a means for embedding a machine learning engine in a graphical interface framework.

Interesting discussion, more so if you are into graphic design. This is related to communicating information in general - which is what programming languages are all about.

The other theme is, of course, programming by demonstration. PBD in the graphic doamin, is closely related to our recent discussions on XSLT.

Posted to general by Ehud Lamm on 9/4/01; 4:47:47 AM

jon fernquest - Re: The Visual Language of Experts in Graphic Design  blueArrow
9/9/2001; 3:38:16 AM (reads: 320, responses: 0)
He mentions using machine learning from examples.

Does machine learning have a role in programming by example?

Machine learning does seem like the only way to write programs that have to deal with encyclopedic numbers of examples like programs that parse and make sense of natural language in a reasonable amount of time.

Transformation Based Learning has become very popular in NLP/Computational linguistics because it has proved to be very reliable with up to 95%+ accuracy rates for some tasks like tagging words for their part of speech. MU-TBL a Prolog meta-interpreter for doing tranformation based learning also mentions that it can be used to learn generalized constraints:

MU-TBL also follows an approach that I really appreciate, namely creating two versions of the software: a feature rich, slightly bloated full version, and a "lite" version that someone has a reasonable chance of reading (like a book) and understanding the basic ideas behind the software. (toy + real tool!)


Jon Fernquest