Artificial Intelligence as a subject of computer science

Posts 1-9 of 9

Alexander Judin Premium Member
The company name is only visible to registered members.

Prolog/lisp - commercial use
Does anybody have any experience in successful applying Prolog for commercial projects?
Probably as a rule engine or core of an expert system?
- 01 Mar 2008, 12:43 pm
Alexander Judin wrote: > Does anybody have any experience in successful applying Prolog for > commercial projects? > Probably as a rule engine or core of an expert system?
Matthias Schwarze Premium Member
The company name is only visible to registered members.

Re: Prolog/lisp - commercial use
Hi Alexander,

I used prolog as a rule engine in a medical archiving system a few years ago.
It ran super smooth and fast - we never had a single problem with it.

The prolog interpreter was swi-prolog (http://www.swi-prolog.org/) first on windows and later also on linux.
- 01 Mar 2008, 10:04 pm
Matthias Schwarze wrote: > Hi Alexander, > > I used prolog as a rule engine in a medical archiving system a few > years ago. > It ran super smooth and fast - we never had a single problem with it. > > The prolog interpreter was swi-prolog (http://www.swi-prolog.org/) > first on windows and later also on linux.
Alexander Judin Premium Member
The company name is only visible to registered members.

Re^2: Prolog/lisp - commercial use
Hi Matthias,
Thanks for replaying :)
Did you have some kind of web-based application to business users may change rules without changing code itself? Or you just hard-coded all rules in Prolog?

Thanks
Alex
- 02 Mar 2008, 09:54 am
Alexander Judin wrote: > Hi Matthias, > Thanks for replaying :) > Did you have some kind of web-based application to business users may > change rules without changing code itself? Or you just hard-coded all > rules in Prolog? > > Thanks > Alex
Matthias Schwarze Premium Member
The company name is only visible to registered members.

Re^3: Prolog/lisp - commercial use
Hi Alexander,

it was not web based and not all user were allowed to change the rules (only some trained admins and mainly our own support guys). It was simply too dangerous to let the customer "play" with the ruleset.
The rule engine reads and parses a file with simple IF-THEN rules - it's important for the admin / support to focus on the ruleset and not on learning prolog.

On the technical side, the rule engine was a standalone program that gets the rules from a rules file and the data (to apply the rules on) via a TCP/IP socket connection.
- 02 Mar 2008, 3:37 pm
Matthias Schwarze wrote: > Hi Alexander, > > it was not web based and not all user were allowed to change the > rules (only some trained admins and mainly our own support guys). It > was simply too dangerous to let the customer "play" with the ruleset. > The rule engine reads and parses a file with simple IF-THEN rules - > it's important for the admin / support to focus on the ruleset and > not on learning prolog. > > On the technical side, the rule engine was a standalone program that > gets the rules from a rules file and the data (to apply the rules on) > via a TCP/IP socket connection.
Alexander Judin Premium Member
The company name is only visible to registered members.

Re^4: Prolog/lisp - commercial use
well... I see... so you just implemented your own Rule Engine based on Prolog...
Do you have an experience with ILOG? do you have an idea how expensive it is?

Sorry for asking so many questions... I have 10 years IT experiance but unfortunatelly never was faced with AI .. but looks like it is very interesting area and not so many products have been built here...
- 03 Mar 2008, 2:00 pm
Alexander Judin wrote: > well... I see... so you just implemented your own Rule Engine based > on Prolog... > Do you have an experience with ILOG? do you have an idea how > expensive it is? > > Sorry for asking so many questions... I have 10 years IT experiance > but unfortunatelly never was faced with AI .. but looks like it is > very interesting area and not so many products have been built here...
Matthias Schwarze Premium Member
The company name is only visible to registered members.

Re^5: Prolog/lisp - commercial use
well... I see... so you just implemented your own Rule Engine based on Prolog... Kind of - I simply defined the statements ("if", "then", ...) as prolog operators and loaded these "human readable" rules as a prolog source file. So most of the magic is done by the prolog interpreter, i just helped with the "translation" ;-)

Do you have an experience with ILOG? do you have an idea how expensive it is? Sorry, never used it or even saw it in action.

Sorry for asking so many questions... You're welcome - feel free to contact me if you like to discuss this in more detail!
- 04 Mar 2008, 12:04 am
Matthias Schwarze wrote: > > well... I see... so you just implemented your own Rule Engine based > > on Prolog... > Kind of - I simply defined the statements ("if", "then", ...) as > prolog operators and loaded these "human readable" rules as a prolog > source file. So most of the magic is done by the prolog interpreter, > i just helped with the "translation" ;-) > > > Do you have an experience with ILOG? do you have an idea how > > expensive it is? > Sorry, never used it or even saw it in action. > > > Sorry for asking so many questions... > You're welcome - feel free to contact me if you like to discuss this > in more detail!
Dr. Alexander K. Seewald Premium Member
The company name is only visible to registered members.

Re^5: Prolog/lisp - commercial use
Alexander Judin schrieb:
Sorry for asking so many questions... I have 10 years IT experiance but unfortunatelly never was faced with AI .. but looks like it is very interesting area and not so many products have been built here... Uh, well the reason for this is probably that it is extremely hard to build consistent, error-free rule bases for non-trivial tasks. People underestimated the effort greatly, even Doug Lenat with his CyC project ultimately failed after almost two decades of effort to teach the computer background knowledge to understand 100 entries of an encyclopedia...

This is why nowadays stand-alone Lisp/Prolog system are not bult, but embedded into "normal" IT Software. In fact, AI has become more like a programming method than a stand-alone system, and aspects from AI like search space complexity, optimal search algorithms and so on have become main stream and are no longer called AI.

A different way to capture human knowledge, which offers a lot of benefits versus the classical AI programming approach, is machine learning. Depending on your application, this might be more likely what you need.

Best,
Alex
- 08 Mar 2008, 2:39 pm
Dr. Alexander K. Seewald wrote: > Alexander Judin schrieb: > > Sorry for asking so many questions... I have 10 years IT experiance > > but unfortunatelly never was faced with AI .. but looks like it is > > very interesting area and not so many products have been built here... > Uh, well the reason for this is probably that it is extremely hard to > build consistent, error-free rule bases for non-trivial tasks. People > underestimated the effort greatly, even Doug Lenat with his CyC > project ultimately failed after almost two decades of effort to teach > the computer background knowledge to understand 100 entries of an > encyclopedia... > > This is why nowadays stand-alone Lisp/Prolog system are not bult, but > embedded into "normal" IT Software. In fact, AI has become more like > a programming method than a stand-alone system, and aspects from AI > like search space complexity, optimal search algorithms and so on > have become main stream and are no longer called AI. > > A different way to capture human knowledge, which offers a lot of > benefits versus the classical AI programming approach, is machine > learning. Depending on your application, this might be more likely > what you need. > > Best, > Alex
Alexander Judin Premium Member
The company name is only visible to registered members.

Re^6: Prolog/lisp - commercial use
well... looks like you are right... it is very unproductive trying to catch even simple knowleges with predicate calculum and Prolog/list language... and it is very hard to change that rules on the fly. I cannot imagine how it might work for commercial systems.

Dr. Alexander K. Seewald wrote:
. In fact, AI has become more like
a programming method than a stand-alone system, and aspects from AI like search space complexity, optimal search algorithms and so on have become main stream and are no longer called AI. ------------- could you recommend anything to read about that?

A different way to capture human knowledge, which offers a lot of benefits versus the classical AI programming approach, is machine learning. Depending on your application, this might be more likely what you need. -------------- same for this :) I would be very greatful for couple of useful links...

well... the task I am working on now is more about intellectual search... My team is working on big scientific encyclopedia http://www.thermopedia.com and the big problem is ... it is not enought just to index all content and make search by a word... I would rather like to implement some kind of itellectual search ... let me explain what I mean... for example let the server catch information about sessions of high experianced users (PhD etc) and catch words they are looking for during a session, then put it in KB and next time when another user looks for that word give him that words as a result.... of course this is a rough idea, I am just a direction what I am thinking about... So it will require keeping rules about sessions in KB...
Could you point me in right directions here?
- 10 Mar 2008, 09:27 am
Alexander Judin wrote: > well... looks like you are right... it is very unproductive trying to > catch even simple knowleges with predicate calculum and Prolog/list > language... and it is very hard to change that rules on the fly. I > cannot imagine how it might work for commercial systems. > > Dr. Alexander K. Seewald wrote: > . In fact, AI has become more like > > a programming method than a stand-alone system, and aspects from AI > > like search space complexity, optimal search algorithms and so on > > have become main stream and are no longer called AI. > ------------- could you recommend anything to read about that? > > > > A different way to capture human knowledge, which offers a lot of > > benefits versus the classical AI programming approach, is machine > > learning. Depending on your application, this might be more likely > > what you need. > -------------- same for this :) I would be very greatful for couple > of useful links... > > well... the task I am working on now is more about intellectual > search... My team is working on big scientific encyclopedia > www.thermopedia.com and the big problem is ... it is not enought > just to index all content and make search by a word... I would > rather like to implement some kind of itellectual search ... let me > explain what I mean... for example let the server catch information > about sessions of high experianced users (PhD etc) and catch words > they are looking for during a session, then put it in KB and next > time when another user looks for that word give him that words as a > result.... of course this is a rough idea, I am just a direction what > I am thinking about... So it will require keeping rules about > sessions in KB... > Could you point me in right directions here?
Ulrich Keil Premium Member
The company name is only visible to registered members.

Re^7: Prolog/lisp - commercial use
Alexander Judin schrieb:
well... the task I am working on now is more about intellectual search... My team is working on big scientific encyclopedia http://www.thermopedia.com and the big problem is ... it is not enought just to index all content and make search by a word... I would rather like to implement some kind of itellectual search ... let me explain what I mean... for example let the server catch information about sessions of high experianced users (PhD etc) and catch words they are looking for during a session, then put it in KB and next time when another user looks for that word give him that words as a result.... of course this is a rough idea, I am just a direction what I am thinking about... So it will require keeping rules about sessions in KB... Could you point me in right directions here?
You might want to have a look at AllegroGraph (http://agraph.franz.com/allegrograph/).

It's a commercial product which allows you to create and query a large KB via SPARQL, RDFS++ and Prolog from Java and Lisp applications.

Ulrich
- 10 Mar 2008, 3:36 pm
Ulrich Keil wrote: > Alexander Judin schrieb: > > well... the task I am working on now is more about intellectual > > search... My team is working on big scientific encyclopedia > > www.thermopedia.com and the big problem is ... it is not enought > > just to index all content and make search by a word... I would > > rather like to implement some kind of itellectual search ... let me > > explain what I mean... for example let the server catch information > > about sessions of high experianced users (PhD etc) and catch words > > they are looking for during a session, then put it in KB and next > > time when another user looks for that word give him that words as a > > result.... of course this is a rough idea, I am just a direction what > > I am thinking about... So it will require keeping rules about > > sessions in KB... Could you point me in right directions here? > > You might want to have a look at AllegroGraph > (http://agraph.franz.com/allegrograph/). > > It's a commercial product which allows you to create and query a > large KB via SPARQL, RDFS++ and Prolog from Java and Lisp > applications. > > Ulrich