What is a knowledge representation? Although knowledge representation is one of the central and in some ways most familiar concepts in AI, the most fundamental question about it–What is it?–has rarely been answered directly. Numerous papers have lobbied for one or another variety of representation, other papers have argued for various properties a representation should have, while still others have focused on properties that are important to the notion of representation in general. [continue reading]

Other than that, the scope of the workshop will remain deliberately unspecified so that we are allowed to decide session after session what topics should be discussed. I’ll be posting the slides and research produced in the context of the workshop on this blog, so maybe also others will be interested in taking part in this (either physically or electronically!). if you do, please get in touch :-)

Here’re the slides from our first meeting:

>>>

Among the TOPICS that emerged as needing more reflection:

Finally, some BIBLIOGRAPHY:

Doug. Ontologies: State of the Art, Business Potential, and Grand Challenges. Ontology Management: Semantic Web, Semantic Web Services, and Business Applications (2007) pp. 1-20

Sowa. Knowledge Representation: Logical, Philosophical and Computational Foundations. Course Technology (1999)

Niles and Pease. Towards a Standard Upper Ontology. FOIS’01 (2001)

Doerr. The CIDOC conceptual reference module: an ontological approach to semantic interoperability of metadata. AI Magazine archive (2003) vol. 24 (3) pp. 75-92

Gangemi et al. Sweetening Ontologies with DOLCE. 13th International Conference on Knowledge Engineering and Knowledge Management (EKAW02) (2002)

Smith. Beyond Concepts: Ontology as Reality Representation. Proceedings of FOIS 2004. International Conference on Formal Ontology and Information Systems (2004)

Guha and Lenat. Cyc: A Midterm Report. AI Magazine (1990) pp. 1-28

Gruber. It Is What It Does: The Pragmatics of Ontology. Invited presentation to the meeting of the CIDOC Conceptual Reference Model committee (2003)

Guarino and Welty. Evaluating ontological decisions with OntoClean. Commun. ACM (2002) vol. 45 (2) pp. 61-65

That’s it for now, cheers!

The first excerpt is about the ‘ever-changing meaning of ontology’. The second one instead is a graph depicting the role of ontologies in semantic systems.

And the good news is: you can get a pdf pre-print for free!

Abstract
Several technologies are emerging that provide new ways to capture, store, present and use knowledge. This book is the first to provide a comprehensive introduction to five of the most important of these technologies: Knowledge Engineering, Knowledge Based Engineering, Knowledge Webs, Ontologies and Semantic Webs. For each of these, answers are given to a number of key questions (What is it? How does it operate? How is a system developed? What can it be used for? What tools are available? What are the main issues?). The book is aimed at students, researchers and practitioners interested in Knowledge Management, Artificial Intelligence, Design Engineering and Web Technologies.

During the 1990s, Nick worked at the University of Nottingham on the application of AI techniques to knowledge management and on various knowledge acquisition projects to develop expert systems for military applications. In 1999, he joined Epistemics where he worked on numerous knowledge projects and helped establish knowledge management programmes at large organisations in the engineering, technology and legal sectors. He is author of the book “Knowledge Acquisition in Practice”, which describes a step-by-step procedure for acquiring and implementing expertise. He maintains strong links with leading research organisations working on knowledge technologies, such as knowledge-based engineering, ontologies and semantic technologies.

The seminar’s title is “reasoning about knowledge‘. Fagin and others, lead authors in the area, define its scope as follows (get the PDF here):

As its title suggests, this book investigates reasoning about knowledge, in particular, reasoning about the knowledge of agents who reason about the world and each otherâ€s knowledge. This is the type of reasoning one often sees in puzzles or Sherlock Holmes mysteries, where we might have reasoning such as this:
If Alice knew that Bob knew that Charlie was wearing a red shirt, then Alice would have known that Bob would have known that Charlie couldnâ€t have been in the pantry at midnight. But Alice didnâ€t know this . . .
As we shall see, this type of reasoning is also important in a surprising number of other contexts. Researchers in a wide variety of disciplines, from philosophy to economics to cryptography, have all found that issues involving agents reasoning about other agents†knowledge are of great relevance to them. We attempt to provide here a framework for understanding and analyzing reasoning about knowledge that is intuitive, mathematically well founded, useful in practice, and widely applicable.

For the moment we’ve just been clarifying our language and the conceptual tools we need to move on to the core issues. But the discussion’s been really lively, so I guess I’ll keep posting about this. A simple map of the recent meeting is online for public consumption :-)

The article dates back to 2004, but the insights you’ll find there are (unfortunately) still valid today. For example, the problems the SW inherits from AI, but hardly recognizes as such in this newest community(here i just outline them – have a look at the paper for their specifications):

• knowledge representation problem
• higher order problem
• abstraction problem
• frame problem
• symbol-grounding problem
• problem of trust

None of this has been solved yet – although apparently the ontologies in the SW are increasing in both number and size… how come? Of course, that’s what the research is about, tryin to solve unsolved problems, but what the heck, shouldn’t we already be aware of their status as “VERY OLD PROBLEMS”?
Think of an ideal world where, as you polish up your SW-paper, on the side of the ACM category descriptor, you should also explicitly mention what problem you are tackling. Mmmmmm too dangerous.. don’t know how many papers would be classified as “novel” or “interesting”, then…
As Halpin says (quoting Santayana) “those who do not remember the past are condemned to repeat it”.
I agree. And I also agree on this conclusion, which I entirely report:

Engineering or Epistemology?

The Semantic Web may not be able to solve many of these problems. Many Semantic Web researchers pride themselves on being engineers as opposed to artificial intelligence researchers, logicians, or philosophers, and have been known to believe that many of these problems are engineering problems. While there may be suitable formalizations for time and ways of dealing with higher-order logic, the problems of knowledge representation and abstraction appear to be epistemological characteristics of the world that are ultimately resistant to any solution. It may be impossible to solve some of these problems satisfactorily, yet having awareness of these problems can only help the development of the Web.

In a related and *extremely* funny rant Drew McDermott, in a 1981 paper called Artificial intelligence meets natural stupidity, is certainly not aware of the forthcoming semantic web-wave of illusions, but certainly points out a few common mistakes we could recognize also nowadays… it’s a relaxing reading , but very competent.. I just report the final “benediction“, in which he describes the major “methodological and substantive issues over which we have stumbled”:

1. The insistence of AI people that an action is a change of state of the world or a world model, and that thinking about actions amounts to •stringing state changes together to accomplish a big state change. This seems to me not an oversimplification, but a false start. How many of your actions can be characterized as state changes, or are even performed to effect state changes? How many of a program’s actions in problem solving? (NOt the actions it strings together, but the actions it takes, like “trying short strings first”, or “assuming the block is where it’s supposed to be”.)
2. The notion that a semantic network is a network. In lucid moments, network hackers realize that lines drawn between nodes stand for pointers, that almost everything in an AI program is a pointer, end that any list structure could be drawn as a network, the choice of what to call node and what to call link being arbitrary. Their lucid moments are few.
3. The notion that a semantic network is semantic.
4. Any indulgence in the “procedural-declarative” controversy. Anyone who hasn’t figured this “controversy” out yet should be considered to have missed his chance, and be banned from talking about it. Notice that at Carnegie-Mellon they haven’t worried too much about this dispute, and haven’t suffered at all.
5. The idea that because you can see your way through a problem space, your program can: the “wishful control structure” problem.

………. I couldn’t resist from adding also a reference (suggested by KMi’s mate Laurian) to a paper by Peter Gardenfors written for FOIS2004, titled “How to make the Semantic Web more semantic” . He’s proposing a novel and less-symbolic approach to knowledge representation, and the overall spirit of the paper matches the the quote from Santayana mentioned above. The conclusion reads as follows:

It is slightly discomforting to read that the philosopher John Locke already in 1690 formulated the problem of describing the structure of our semantic knowledge in his Essay Concerning Human Understanding: “[M]en are far enough from having agreed on the precise number of simple ideas or qualities belonging to any sort of things, signified by its name. Nor is it a wonder; since it requires much time, pains, and skill, strict inquiry, and long examination to find out what, and how many, those simple ideas are, which are constantly and inseparably united in nature, and are always to be found together in the same subject.” ([25], book III, chapter VI, 30) Even though our knowledge has advanced a bit since then, we still face the same problems in the construction of the Semantic Web.

In this tutorial Mizoguchi talks about representation – represented-thing – representational-form etc.. No mention of information-objects, which are instead (only as a term, maybe) always present everywhere else (Dolce, Cidoc, Cyc). So what’s the proper mapping? Is an IO a representation? Moreover, I am trying to put also another model in the picture, FRBR. This bibliographic standard focuses around concepts such as work, expression and manifestation, mainly. So how do these come into the game?

Sometimes I end up in some sort of ontological relativism. Since the objects we study are essentially multidimensional, and since we humans can only rationally perceive a portion of such dimensions at a given time, it follows that whatever perspective we decide to take on our objects of investigation, it will be fundamentally arbitrary and partial. In other words, there is no chance of having a single unifying perspective on reality, one which can contain all the others. No chance. So every representation gives you one side of the story only – which can be related to other sides, but never in its entirety.
If this is the case, we better keep all the possible ‘sides’ of this story, and just use the one we need whenever we need it. A pragmatist approach? Well.. more on this to come soon! (i’d also like to go back to Peirce, and check how all of this relates to his work)

Back to us: this is the model used in Dolce (more precisely, in the DnS module of Dolce, as described in Gangemi, A., Borgo, S., & Catenacci, C. (2005). Metokis deliverable D07 – Task Taxonomies for Knowledge Content. Deliverable of the EU FP6 project Metokis):

And this is an example of its instantiation:

The theory of Mizoguchi, as I said, it’s quite different. Here’s an interesting excerpt from the article Mizoguchi, R. (2004). Tutorial on ontological engineering – Part 3: Advanced course of ontological engineering. New Generation Computing, 22(2), 198–220.

A representation is composed of two parts, form and content.

Representation
…….p/o “form”: Representational form
…….p/o”content”: Proposition

where p/o stands for part-of relation/slot, “form” slot name and “: Representation” is a class constraint the slot value has to satisfy. Its identity is inherited from the form which is usually what people sense its existence. On the other hand, the content is the hidden part and it is a proposition which the author of the representation would like to convey through the representation.
[…]
It is critical to distinguish among proposition(content), representation and form of representation. In fact, although a novel is written in terms of sentences, novel is not a subclass of representation. What exists as a subclass of representation are what have the form of representation as its intrinsic property, that is, sentence, musical score, painting, etc. The sentences of Tale of Genji are instance-of sentence. However, representation and form of representation are different. Concerning a novel, representation is “sentence” which is composed of its content(novel) and “natural language” which is the syntactic part of the sentence, as the form of representation.

This is a quite impressive visual rendering of this ontological theory:

How are these theories different from each other? What are their pros and cons, where is it that they could be used more successfully? There’s work to be done here..