14 April 2008

Artificial Intelligence Applications Institute



Artificial Intelligence Applications Institute Nice looking organization, with some interesting links. Associated with Edinburgh University, so it is reasonable to assume legitimacy of the organization (rather than a group using AI as just a buzzword to generate business).


Tags:

Labels:

18 March 2008

The Knowledge Representation Requirements Model

Laszlo's Pyramid of Meaning is certainly an interesting way to view the presentation of meaning. It is an upside down triangle, with the narrowest point being "data", which then ascends through "information", "knowledge", and on up through "comprehension", "understanding", "wisdom", and finally "enlightenment". It looks like this (sort of)...

. . . . . enlightenment . . . . . .
. . . . . . wisdom. . . . . . .
. . . understanding . . . .
. . . comprehension . .
. . . knowledge . .
. information .
. . data. .

The objection that I have to the Laszlo pyramid, for my work at least, is that it seems to blur knowledge representation (for automata) at the lower levels, and then knowledge, epistemology, mereology, and a bunch of natural language concepts at the upper levels.

Because of this, I think that my own Knowledge Representation Requirements Model will serve as a better benchmark for what I want to show. It is more in line with the ontological representation method I am proposing, and also in line with the Levels of Conceptual Interoperability Model.

The lowest level of the KRRM is data. Data with semantic identification becomes information. Information in context becomes knowledge. Knowledge in a time-senstive context becomes awareness. Awareness subject to comprehending context changes becomes understanding.

Understanding
Awareness
Knowledge
Information
Data

More later...

Tags: , , , , , , ,

05 March 2008

Does Artificial Intelligence worm its way into Studies of Cognition

Artificial Intelligence and Study of Cognition I
Artificial Intelligence and Study of Cognition II

Interesting article (two parts) discussig AI and cognition studies. Often, in the literature, the second is seen as (partially) a rebranding of the first, but I have felt that there is more difference than overlap. Nice to see this author agrees.

Tags: ,

Labels: ,

Comparing

Cognitive Distortion

Interesting to see how these sorts of "leaps in thinking" compare to the techniques of analogical reasoning

Tags:

22 March 2007

Why the Semantic Web will Succeed

Half an Hour: Why the Semantic Web Will Fail

The above link is to an article that gives some basic reasons (according to the author) as to why the semantic web will fail. In summary, his argument is that the semantic web will fail not because of any technology shortcomings, but because businesses will not want to cooperate for reasons of profit, proprietary domain, competition, etc.

This is ridiculous. The first (our current) version of the web works because there is a reasonable stable standard way for people to exhibit information, and for people to retrieve that information. The semantic web is the same thing, only for systems to do the exhibition and retrieval. The first time any company has a number of services that can be accessed by consumers in a system-to-system manner and this shows even an inkling of profit, the corporate world will beat a path to the doorstep of some self-synchronized standard.

This has happened in history - think of the adoption of comon means of trade, the adoption of common formats of record keeping, using standardized banking transfer methods, the adoption of the fax machine, the adoption of supercalc, the adoption of microsoft office as a defacto standard. The adoption of the pdf. In all of these cases, collaboration proved to be more profitable and more valuable than choosing a (perhaps better) different path to the same goal.

Once people saw that there was benefit to producing information in a semi-standardized format (html?), then everyone felt the need to produce their information in such a format. Sure it was bumpy and crunchy fighting through the evolution of the "standard" (and the evolution continues) and there were corporate sticks-in-the-mud that purposefully deviated in order to capture some market advantage, but in the end it is all still pretty much standardized.

Almost any browser can read almost any webpage, and if you have a business case for presenting information, then you are a Fool (notice the capital F) if you rely on a presentation method that locks out part of the audience by not adopting to the defacto standard. Sure, appealing to an elite part of the audience by using a cutting edge method for presenting content is cool and stylish, but you wouldn't want to base a business plan that needs to reach the masses on such an approach.

This same self-synchronization will occur in the system-to-system web world (the semantic web, as envisioned by Sir Tim Berners-Lee). Yeah, it will be rough, and it might take a couple of years to get to such a standard. But if we apply a little perspective here, we are talking about maybe a decade. And then think about what a difference there will be in the world when this starts to take hold. For comparison, think of the difference just a decade made between 1990 (no WWW) and 2000 (ubiquitous WWW).

Tags:

Labels:

22 November 2006

Raganathan's Facets

Interesting idea - Raganathan's ideas of facets are similar to my idea of concepts. To be explored further . . .

Ranganathan

Tags: , ,

Labels: , ,

Information Architecture

Interesting site, I found while looking up information concerning the formulation of a thesaurus.

Information Architecture

Labels:

Ontopia Article Captures the Ontology Spectrum

Leo Obrst has published (as slides from talks) a couple of times something that he calls the ontology spectrum (or, in an upcoming book, the "Logic Spectrum").

The same ideas appear in this article.


Tags: controlled vocabularies, taxonomy, thesauri, ontology, topic map

Labels: , , , ,

14 September 2006

Why Johnny Can't Code

Why Johnny Can't Code A Salon Article written by David Brin.

This is a great article talking about the sad reality that these days, many operating systems don't have a neat, easy built in command line programming language. Like BASIC, which was in all the OS's from the micro, mini, and home computer revolution in the 70's, 80's, and early 90's.

Great reading, and certainly gives some insight into why a generation of computer geeks (like myself) have it built into our system (crikey - I've been programming since 1977, when the father of a friend of mine taught his son and me how to program in basic on a TRS-80 Model 1). These days, kids learn it all in school. Many get into it because it will be a "good career". Well, you can certainly see the difference between someone who is self taught and has it in their soul, compared to someone who learns it as career development. The difference is between a practitioner (the latter) and a geek (the former). Geeks breathe the stuff, practitioners merely do it.


Tags:

01 May 2006

Concept based ontology system described using set theory

If my system (based on concepts) is right, then the universe of discourse can be described something like:

1. There exist a finite number of properties, the set of which we will refer to as P. There is, likewise, a set of property values, which we shall call Pv. Each property in P can be given quantification and qualification via 1 or more members of Pv which are mapped the member of P. Together this makes a P-PV mapped, which we shall describe as Pr{P, Pv}. When discussed in abstract, it is possible to acknowledge that a particular {P, Pv} pairing will have a range for its property values, rather than a specific value. Then when the property is applied with specificity, a value within that range reveals itself.

2. There exist a finite number of concepts, each of which have some subset of Pr that it exhibits. A concept, as described here, is an agreed to mental idea of a trait common to everything in the universe of discourse said to exhibit the concept (a somewhat circular definition, but it is being worked on). The set of all concepts, we shall call C. The idea of a propertied concept, that is the concept considered with all of it's exhibitted properties, exists, and the set of all of these propertied concepts we will call Cp. The properties in question come from Pr (and may, therefor, have the Pv expressed as a range, when discussed in the abstract, or an actual value, when discussed with instance specificity). Each member of Cp contains one or more members from Pr, which we signify by writing Cp(Pr*).

3. Within the universe of discourse, there are a number of entities. Entities are "things" within the universe of discourse. By "things", it is meant to consider both mental and physical things, as well as processes and events. For each of these, there are both abstract types as well as specific instances. The complete set of entities we shall call E.

4. There is a power set of Cp, which we shall refer to as power(Cp).

5. Each entity can definited by one subset from power(Cp). It is highly likely that within the universe of discousrse that members of power(Cp) will not be applicable to any of the entities within that universe of discourse, due to the conflicting nature of the concepts involved.

6. There is a universal entity type which is the most general of all entity types, which has no defining concepts other than those required to establish the existence of an entity. We shall call this universal entity U. U will have several children entities, each of which are disjoint from each other and also disjoint from their common parent by having a different subset of power(Cp) as its defining list of concepts, properties, and property values.

7. Each entity within the universe of discourse evolves as a child entity of some more general (i.e. - closer to U in heritage) entity. Each child entity inherits all of power(Cp) that its parent, yet is disjoint from its parent, and all its siblings, by having a different subset from power(Cp) that defines this. Note by 2 above, Pr is definied as a pairing of P and Pv, and that Pv can be a range or can be a value. If the only difference between siblings, or between a child and a parent, is a different specific value for Pv, then this required disjointness is satisfied.

8. Starting with U and tracing the "heritage" (path from parent to child to grandchild, etc) will give a series of propertied concepts, all of which accumulate with each new generation in the heritage.

9. Depending on the nature of the universe of discourse, there will be other semantic links between entities that are not parents or children or siblings of each other, depending on the entities in question and their relationship to each other within the universe of discourse. These semantic links we will call relations, and the set of all possible relations we will call Re. Each Re can have any number of satisfying strings, a string consisting of two (or more) entities that are linked together by the relation. We shall write this as Re(E1, E2), where E1 and E2 are entities (types, instances, objects, events, or processes).

10. It is possible that a number of entities that have semantic links (relations) can be combined to make a statement within the universe of discourse. In order to ensure that related entities are combined in the correct manner, there should be a number of guidelines ensuring that any declarative statements made within the universe of discourse create analytic statements (statements that have a priori truth) that are internally correct and make good semantic sense. These will be called the Internal Rules, and all of the possible such rules make up a set, which we shall call RuI. Such a set of rules could be considered a grammar. Each rule will have a condition describing when it could be applied, and the condition can have a conditional value. Condition values can exist as a specific value, or can exist as an allowed for range (i.e. - the Condition Value CoV(x) for Condition C(X) must be within the range of y to z in order for the RuI(a) to be true). Each member of RuI will then have a number of strings containing a relation, and a condition-condition_value pairing. We shall call conditions Co, and condition values CoV. The RuI string will be written RuI(Re,(Co,CoV)).

11. It is possible to make statements within the universe of discourse that have a number of contextual affects on when it is proper to make such a statement. There are a set of rules describing such statements, and these are referred to as the external rules. The set of all such external rules we will refer to as RuE. Rather than being analytic statements, which have a priori truth, these are synthetic statements, which are only true within a certain context. It is therefore important to capture the context state, and any values that the context state may have in order to describe when these rules may be applied. Note that as with conditions above, contexts values may exist as specific values, or defined as a satisfying range. A context will be writen as Ct, and a context value will be written as CtV. Therefore the total set of external rules will contain, for each RuE, a number of strings with the value RuE(Re,(Ct,CtV)).







Tags: