Why declarative programming didn't succeed?

I'd like to write something like this (simplification):

Diagram is a set of circles with radius 10 representing my notions
(from a given set of notions), connected by lines denoting
relationships between these notions (from another given set of
relationships). Circles on the diagrams do not have common areas.
Lines are crossing in as little number of places as possible. Diagram
is as little as possible. I give you 300ms to find the best diagram
for my data you can.

Why I cant just put this text into my computer's console and get
beautiful diagram of my data? What is so difficult? Why after so many
years of research it is still not possible?

Thank you very much for all interesting responses :-)

Krzysztof

Krzysztof wrote:
> I'd like to write something like this (simplification):
>
> Diagram is a set of circles with radius 10 representing my notions
> (from a given set of notions), connected by lines denoting
> relationships between these notions (from another given set of
> relationships). Circles on the diagrams do not have common areas.
> Lines are crossing in as little number of places as possible. Diagram
> is as little as possible. I give you 300ms to find the best diagram
> for my data you can.
>
> Why I cant just put this text into my computer's console and get
> beautiful diagram of my data? What is so difficult? Why after so many
> years of research it is still not possible?
>
> Thank you very much for all interesting responses :-)

Looks like some job for CHR or other contraint programming techniques,
but I'm not familiar enough to sketch a solution, although I read the
book on CHR ...

On Aug 18, 9:20*pm, Krzysztof <Krzysztof.Ch...@gmail.com> wrote:
> I'd like to write something like this (simplification):
>
> Diagram is a set of circles with radius 10 representing my notions
> (from a given set of notions), connected by lines denoting
> relationships between these notions (from another given set of
> relationships). Circles on the diagrams do not have common areas.
> Lines are crossing in as little number of places as possible. Diagram
> is as little as possible. I give you 300ms to find the best diagram
> for my data you can.
>
> Why I cant just put this text into my computer's console and get
> beautiful diagram of my data? What is so difficult? Why after so many
> years of research it is still not possible?
>
> Thank you very much for all interesting responses :-)
>
> Krzysztof

Krzysztof

Your problem could be broken down to two sub-problems

(1) having your natural language text parsed by a computer
(2) convert the parser output into diagrams

Have a look at Attempto Comntrolled English (attempto.ifi.uzh.ch) that
can help you with the first step – translating a variant of your text
into logic.

--- nef

Hi Norbert,

Thanks for link to ACE, it is a very interesting language! Still,
there are more steps missing that a PC would need to answer my query I
think. Let me summarize all the steps:
(1) parse my query with ACE parser
(2) retrieve notions from my database (that should be specified in the
query text) and link data from database with facts presented in the
query
(3) resolve the query with some constraint solver?
(4) map the results to some form that can be displayed, SVG for
example?

Well, it might be true then, that all the bits and pieces needed to
perform the tasks are ready, just nobody ever tried to put them
together. However I have some doubts:
(3) - could query similar to the one presented by me be really
efficiently resolved? From what I remember from school, most of first
order logic problems will be too difficult for computer, not to
mention that for some problems first order logic will be not enough.
Is that true?
(1) and (4) - there would have to exist a database with predefined
notions ready for use, so the user wouldn't have to define everything
from scratch and wouldn't have to define how to use SVG format etc.

It would be very interesting to see how my query would look like in
ACE (complete version that would give computer all the necessary
information) and then to analyse what kind of solver could potentially
solve it. I'll try to learn more about ACE and prepare the query
later.

Also, from the usability point of view, how could the user be informed
of problems with query he has created? Example problem: "unable to
resolve query because searching space is too broad." In such case the
computer could ask the user to provide some more facts that could
narrow the search space.

I am not a theoretical computer scientist really, just a programmer,
so I guess that in this post I'm touching lots of different areas, but
it is very interesting how all of it could work together.

Best Regards,
Krzysztof Choma

On Aug 20, 8:18*am, "norbert.e.fu...@gmail.com"
<norbert.e.fu...@gmail.com> wrote:
> Your problem could be broken down to two sub-problems
>
> * (1) having your natural language text parsed by a computer
> * (2) convert the parser output into diagrams
>
> Have a look at Attempto Comntrolled English (attempto.ifi.uzh.ch) that
> can help you with the first step – translating a variant of your text
> into logic.

On Sun, 24 Aug 2008 06:55:00 -0700 (PDT), Krzysztof
<Krzysztof.Choma@gmail.com> wrote:

>Hi Norbert,
>
>
>I am not a theoretical computer scientist really, just a programmer,
>so I guess that in this post I'm touching lots of different areas, but
>it is very interesting how all of it could work together.
>

Study the history of Artificial Intelligence - a big part of this what
you are writing about was implemented in late 70 - early 80. Such as
program for solving mathematical problems formulated as text, planner
for "block works" controleld by natural language and other...

Unfortunately, for various reasons, nobody wanted to invest more money
to continue this research.


A.L.

Krzysztof <Krzysztof.Choma@gmail.com> writes:
>I'd like to write something like this (simplification):
>
>Diagram is a set of circles with radius 10 representing my notions
>(from a given set of notions), connected by lines denoting
>relationships between these notions (from another given set of
>relationships). Circles on the diagrams do not have common areas.
>Lines are crossing in as little number of places as possible. Diagram
>is as little as possible. I give you 300ms to find the best diagram
>for my data you can.

dot/neato. Takes more than 300ms to type, though.