Representation of meaning is just one “dimension” of development and management of knowledge as explicit expression of meaning. An appropriate model for these processes requires at least three “dimensions” (or, perhaps, “aspects”):
- The Dimension of Meaning itself — Insights, Facts, Concepts, and the relationships among them.
- The Dimension of Participation — Participants and their actions relative to the explicit knowledgebase being constructed, including creation, evaluation, actions, and implementation.
- The Dimension of Information in which instances of meaning occur — documents, email, etc. — in natural language and other forms of communication.
- [Optional] The Dimension of Administration — information about construction, organization, and management of sample knowledgebases and why they are organized in the way shown. In the best of all worlds, this Dimension would not be required (or needed only rarely); the organization and underlying model of the knowledgebase should be evident.
Other “dimensions” may be possible. (In Conceptual Data Structures for Personal Knowledge Management, Völkel and Haller speak of the five axes of knowledge: identity, order, hierarchy, annotation and linking. My notion of Dimensions is different, but I do recommend their paper.)
This facet of ICKMOD — deconstruction of knowledgebases into orthogonal “dimensions” — is particularly helpful in graphic representations of practical knowledge because you can work with objects and relationships in each aspect (each dimension) separately. For example, you can organize and edit Insights and their supporting Facts and Observations without touching information about Participants who have evaluated those objects and relationships … or the relationships among those Participants.
The single greatest challenge to useful semantic implementation [of the types I propose] may be that such goals as “interpreting natural language” and “[being] the latest hope for conferring interoperability on incompatible systems” superimpose frameworks/viewpoints that are inappropriate for addressing the problems we need to address most. These goals are not “wrong”; they should be viewed as complementary to representation of meaning as a resource for humans. But they often obscure what we need to do first.
The Dimension of Meaning
The dimension of Meaning is described in other sections of this publication. However, I want to emphasize points made there and elsewhere:
- The “natural” level of communication of meaning among people is Insights.
- When you discuss structured representions of meaning, you must be able to point to a set objects (and the relationships among them) and assert, “I am talking about about that set of things.”
- Parsing natural-language sentences — especially sentences in isolation — does not get you to meaning … at least, not reliably. Natural language can be used as a foundation for structured representation of meaning — for example, Attempto Controlled English — but such controlled natural languages are themselves new languages whose vocabulary and rules must be learned, clearly understood, and applied rigorously.
The Dimension of Participation — creation, evaluation, and impact
A substantial segment of the knowledge-representation community devotes most of its time to machine interpretation of natural language. Such narratives frequently include statements that repeat or interpret what others have said. As John Sowa observes:
Although the paradoxes of logic are expressible in natural language, the most common use of language about language is to talk about the beliefs, desires, and intentions of the speaker and other people. Many versions of logic and knowledge representation languages, including conceptual graphs, have been used to express such language. As an example, the sentence Tom believes that Mary wants to marry a sailor, contains three clauses, whose nesting may be marked by brackets.Tom believes that [Mary wants to marry a sailor]].
The outer clause asserts that Tom has a belief, which is expressed by the object of the verb believe.
Source: John F. Sowa, “Conceptual Graphs” p. 17 at http://www.jfsowa.com/cg/cg_hbook.pdf
The desire to describe assertions about assertions often leads to complex, nested representations, including Sowa’s own example expressed in a pictorial representation of his Conceptual Graphs on that page.
I bring in Sowa’s explanation and example to underscore a key difference between the goals of the knowledge-representation community and the purpose of ICKMOD: Sowa is representing part of a narrative — a description of reality that has been “flattened” into a single stream in a natural-language text and also into a 2-D representation in the graphic. The flattened description treats participants’ opinions about a topic as the same kind of thing as the topic under discussion (Mary’s desire to marry a sailor).
If the task at hand is to enable computer applications to “understand” natural language by parsing it as a single stream, Sowa’s representation is a good solution. But that’s not what I want to do with ICKMOD. I am concerned with representation and management of useful knowledge in work environments. In ICKMOD, the task at hand is to enable people to understand assertions about reality more reliably and put that understanding to work, so in this example only Mary wants to marry a sailor. is “meaningful.” That’s the part of the original sentence that is of interest when representing meaning. If Tom is a member of the community that shares such interests, his belief about that meaningful statement belongs on a different plane — one that is reserved for commentary on what Mary wants to do.
This is the Dimension of Participants and their activities — their commentary on meaning. In this dimension, we represent Participant activities, including the creation, evaluation and impact of Participants on the knowledgebase. The act of creation is not a part of meaning. Evaluation, too, belongs in this separate plane and must be resolved. These activities are important (vital, even) but they are orthogonal to meaning, belonging with assessment of importance and relevance. Such a separation into multiple planes is a more appropriate model for the practical knowledge of an organization. And placing Participants and their activities in a separate dimension has the additional benefit of dramatically simplifying the representation of meaning [in the dimension of Meaning].
In implementation, this approach also allows us to take full advantage of the capabilities of computers to represent/deconstruct knowledge in more appropriate, more logical ways. We have graphical computers on every desktop — and in almost everyone’s pockets. Why should paper continue to be the required lowest common denominator? With objects on separate planes, we can easily meet the requirements of “reification” — that is, pointing to a set of meaningful objects and relationships that we want to discuss.
Although there are no tools that can fully implement the ICKMOD approach, there are a few, including one commercial desktop product (TheBrain) into which I can shoehorn a flattened (that is, single-plane) version of such a multi-dimensional model. (I did the same in a more recent post, Representing practical knowledge with ICKMOD — an example , using Instrumind ThinkComposer.) ICKMOD provides a target data model for a wide range of complementary products, through which knowledge workers can easily produce useful information in forms to which people and tools can incrementally add consistent structure. (See examples Incremental Formalization and Representing knowledge graphically.)
Placing commentary into a separate, orthogonal dimension is not a new idea. You have probably encountered similar approaches in a variety of tools and models. It makes sense! (For me, Akscyn and McCracken’s Knowledge Management System [KMS] — one of the legendary pre-Web hypertext systems — was a noteworthy implementation of this principle from the 1980s.) Argumentation systems in general — for example, Compendium and DebateGraph — posit a qualitative difference between nodes/links that describe “whatness” and those that represent the positions of participants with regard to that whatness … even if they do not implement “dimensions.”
[Note that people in the Knowledge Representation community sometimes use the term evaluation, but they mean testing an expression for its conformance with a stated set of rules for automatic processing of such data.]
This deconstruction should be formalized and applied to all systems designed for creating, managing, and applying meaning. It’s too easy to succumb to the temptation to conflate the planes of Meaning, Participants, and Information.
The Dimension of Information — documents, email, etc.
As we move ever deeper into the Age of Information, we need to rely much more extensively on representations of meaning than on information. Yet even the strictest formalism and the most graphic representations of meaning have to include information. The labels of nodes and arcs in graphic representations of ontologies are … information expressed in natural language. That’s unavoidable, because the names of representations of things must be expressed in terms we understand — even when the representations purport to be about Concepts and Ideas rather than terms and sentences.
William Woods makes a very funny and rather snarky comment on this problem when he revisits “What’s in a link?”:
Ross Quillian at one point had a semantic network system implemented in LISP and he decided to remove the labels and the resulting network, when printed out, was just a giant nest of parentheses. So it’s right, the semantics was all in the labels.
Source: “What’s in a link?”
Even if it were possible to have a system of representation of meaning that was completely free of natural language, we would continue to rely on natural language and other forms of information (and data) as our primary means of communication and storage of knowledge. And, of course, formalized representations of knowledge are not complete and not as expressive as many natural-language communications. We are not going to live without books, poetry, documents, email, multimedia, etc.
Important points about the Dimension of Information in the Insight-Centric Knowledge Model (ICKMOD) :
- When we are citing information in resources (documents), what we are actually doing is pointing to occurrences (instances) of Insights in those resources. (Sometimes also to occurrences of Facts and Observations.)
- In order to get to meaning, we extract Insights from information — that is, we reverse-engineer from information the meaning that its author encoded in natural language.
- An Insight may not correspond to a single sentence in a chunk of information. In many cases, the text from which an Idea may be drawn may be scattered across non-continuous stretches of text. And parts of the same text may give rise to other Insights.
- Back-of-the-book indexes (and the Topic Maps they inspired) are finding aids. Like the BoB indexing model on which they were built, Topic Maps describe aboutness of documents — an abstraction that is very different from representation of meaning — and relationships among those abstractions (associations).
- A table of contents is also a high-level representation of the rhetoric of a document. It is both an abstracted representation of that rhetoric and a finding aid. The plane of Information is the right place for tables of contents.
The Dimension of Administration
A good graphic representation should need little explanation. A really good graphic representation needs no explanation at all. In my implementation examples, I haven’t achieved that Nirvana. So I have found it necessary to add a fourth “Dimension” — a Dimension of Administration, in which I provide explanations of the sample knowledgebase, notifications of updates, and other information that is not central to the representation of knowledge itself.
My excuse is that I’m doing something new. But the truth is that the ICKMOD model itself and my early examples are not sufficiently elegant that they are clear. Perhaps I will be able to omit this kind of information as my explanations of ICKMOD and the quality of my representations improve.
© Copyright 2018 Philip C. Murray