Journal of Knowledge Management Practice, August 2005

Notes Re: Personal Knowledge Formalization And Modeling

Lev Romanov, Independent Contractor

ABSTRACT:

Knowledge Management is usually applied to Organizational Knowledge System, and the fact that Personal Knowledge is the building block of such system is sometimes overlooked. The value of the System of Personal Formal Knowledge is not limited by organizational use, but it is a valuable personal intangible asset. This article introduces definitions and principles for formal model of Personal Knowledge; in particular, classification, type and context are defined based on relations between knowledge cells. Practical solution is realized as a demo system that describes someone's personal knowledge about the system itself. Some prospective directions for further research and development are introduced.


Discourse And Definitions

Why ‘Knowledge’? It is personal; I have a very hard time remembering things, so some media is always needed to set them aside for future use.

Why ‘Formalization’? This is followed from previous statement – 'putting a piece of knowledge aside' means it should be presented in Formal way, i.e. formalized.

Much confusion comes from the mixed use of terms information and knowledge. We will understand the word 'knowledge' as superior level of information (Godbout, 1996) in contrast to 'the process of knowing'. Information becomes Knowledge if it is actually operated by human being (Godbout, 1999). From this prospective, any Knowledge that resides on media other than human brain is actually information. The Knowledge always resides on media (assuming that the human brain is type of media) and depends on Context. So we will understand Knowledge as 'Information in Context'.

On other hand, aiming Knowledge Formalization, we need to differentiate the information that came from human brain as the result of the process of expressing Knowledge. Let’s also not forget that process of expressing Knowledge is preceded by acquiring Knowledge - picking the knowledge from the outside world and putting it into our heads. Either process: expressing or acquiring knowledge could be viewed as picking the cell of Information, separating from context and putting it into the different context. From a formal point of view, let’s assume that a cell of Knowledge could be extracted and evaluated (including true-false evaluation if appropriate). Picking a piece of Knowledge, separating it with or without limiting Context, place it in another Context - is a process that we will call Knowledge Formalization. This process is started and completed in our heads even if the result is placed outside. The result of Knowledge Formalization will be creation or modification the System of Formal Knowledge. Each of us maintains a personal Formal Knowledge System that is based on the Model called the brain.

While studying or just getting into a new subject we are acquiring knowledge. The personal Knowledge that is ‘sitting’ in our heads is not perfect: whenever we are looking for a specific piece of information it takes time to find it not to mention that this search is not always successful. To compensate this weakness we are using different ways to store pieces of knowledge from notching to writing notes that means we are using different Models of System of Formal Knowledge. Actually from modeling prospective, it doesn’t matter if the System of Formal Knowledge resides in someone’ head or on any other media.

The Model of the System of Formal Knowledge is understood as a set of rules, symbols etc that makes the System unique. It may be as simple as a To Do list or as formal as one described by Maurice Gittens (2003). His research is the most completed among a few publications in Knowledge Management area and concurrent fields that are devoted to theoretical foundations. However, while presented conception is more or less solid, the apparatus to be yet developed.

The Model heavily depends on media. For instance, a book became a revolutionary type of media and the Textbook accompanied it as The Model of the System of Formal Knowledge. Each type of media has its own life cycle. For instance, a lector in class verbally expresses Knowledge: the air in the class would be the media, and life of such media is quite short. Needless to say, that computers stimulate activities in creating Models more suitable for digital media, and this article illustrates one of many such attempts.

Natural language is also an example of such a Model. Historically, the natural language was developed as related to verbal media (brain + air) and later transformed into written language. The idea to use a combination of relatively small set of sounds (letters) to express potentially unlimited number of objects (concepts) seems as came from nature. People developed it to alphabet and added rules to create the new Model that we are using to communicate (including communication with ourselves). The word in the model represents the concept in first approximation. The same word but in the sentence better represents the concept, so relation between concepts does matter to express the concept itself. Similar conclusion is valid if applied to the sentence, paragraph, book, etc. Therefore, nothing in this model represents the concept completely, but up to some level of approximation based on Context.

Context Free Grammar (CFG) was designed in attempt to represent concept fully. And it does, but only for concepts that are based on predefined set of rules. So it implements different type of Model of the System of Formal Knowledge. Advantage of CFG is that it is relatively simple model, and disadvantage – formalization limits (some thing that are obvious for little child couldn’t be represented in the Model, or their formalization is complicated and unpractical). In contrast, Context is a required element of Natural Language that makes such Model more suitable in everyday life.

Research in natural language processing using computers (i.e. CFG) doesn’t promise any significant results. Finally we invented HTML – the language that ignores whatever it doesn’t understand. I would say that this is very intelligent behavior. Some researchers are getting ever further stating that knowledge is inseparable from language (McKinley, 2003), i.e. inseparable from context, and therefore couldn’t exist without a context. This could be rephrasing as knowledge formalization could be achieved only in conjunction with natural language formalization.

Auto-formalization of Knowledge as it defined in (Gromov, 1986) is a process of Formalization that occurred without involving human brain. This is possible after all participating cells of Knowledge are formalized and desired functionality is defined (could also be treated as the cell of Knowledge). As a result – the new cell gets created. While from practical point of view – this is clear a new cell, it is predefined as soon as all participating cells and functionality cell are created and linked. Because Formal Knowledge does involve human brain and tied to Context, and assuming that only tiny part of global knowledge is Formal, it is hard to expect creation of ‘practical cells’ based on formal knowledge only and informal knowledge gets also involved. In this case ‘practical cells’ are not predefined, but regular cells of formal knowledge. On other hand, not all ‘predefined practical cells’ are created – this would overload the system - but those that are useful, i.e. could serve as an Interface between the human brain and the System of Formal Knowledge.

People keep thinking that storing knowledge is for exchange information. We wouldn’t argue what came first, but it definitely serves as a needy additional to our heads. Sometimes we are trying to organize our knowledge and draw diagrams or write notes follow the same procedure that described above. And thereafter we are putting all this stuff back into our heads, so Knowledge Formalization helps us to rethink that we already have known, and exchange information is one of possible purposes. Despite the computer era, people widely use paper for notes to store personal knowledge. Even computer professionals keep notepad handy.

Computer based Model usually has following components (layers):

Ø      Data base for storing information

Ø      Functional logic

Ø      Interface.

Such systems implement user oriented approach in contrary to data oriented one that emphasizes data units linked in hyper net by isotropy connections. Activities in data oriented systems are triggered by data independent events.  And (Packin, 2003). We see three major problems that make difficult storing personal knowledge in computer driven media:

Ø      Psychological problems with knowledge formalization (i.e. understand, but can’t explain);

Ø      Models for Knowledge Formalization and its interfaces(GUI) are still not adequate for use by general public;

Ø      Knowledge Exchange protocol yet to be developed. Similar to translation between languages, there is a need for some common ground for different Models, like Knowledge Interchange Format (Genesereth, 1998)

While the usage of computers has advanced, and Internet become more and more integrated in the everyday life, adequate Model of the System of Formal Knowledge seems as became more critical, and the first step might be - Formalization of Personal Knowledge.

Modeling Knowledge Formalization

Let’s say, Q to be a ‘real’ knowledge that belongs to the space of possible knowledge M

Q Î M

Assuming that set X represents Q in the system of Formal Knowledge S,

q(Q,X) – will be a function that describes such relation.

Excluding trivial situation when knowledge Q doesn’t have any representation in S,

X = Æ

Will take a close look at the object x - formal peace of information,

x Î X.

By definition, x will also represent Q in the system of Formal Knowledge S, so

(Ñ x Î X ) q(Q,X) Þ q(Q,x).

This is not true as formal piece of information (object x) doesn’t express itself but in connections (relations) with other relevant pieces (objects). For instance, ‘John is the son of Peter” – is a peace of information that doesn’t make any good, unless in context with another informational object(s). Let’s say, this is an illustration of the relation between two objects John and Peter, and the original object becomes more valuable. So expression above need to be modified:

(Ñx Î X; Ñy Î X) (q(Q,X); f(x,y))Þ q(Q,x).

This means that following will be trivial situations:

x is the only object in X - (Ñx Î X; Ñy Î X) (x = y)

all objects in X are not related -  - (Ñx Î X; Ñy Î X) (Ø f(x,y))

And f(x,y) on X is required in order X to represent Q.

We will define Yx as context of x if

(Ñy Î X) (f(x,y)) Û (y Î Yx).

Let’s visualize the collection of knowledge items – objects as nodes of the graph. Each object would have a number of connections with other objects. If we introduce a new object, the information it contains itself will be deployed with a context – a set of connected objects. Even the object is identical to another one in the system, context will be different and object in general will differ from the original one.

We will call object content – Information, and object with collection of connections – Formal Knowledge. In other words, Formal Knowledge is a function that related informational object x to its context Y:

fx: x ® Y.

So far, we separated informational object x and its context Y, but it is reasonable to conclude that if

§         object x could be realize in and only in its context Y, and

§         context Y doesn’t have any meaning without existence of object x,

the sets of objects and its contexts are isomorphic, and the only purpose of separating them in our model – access to object x that may be implemented only using its links to outside word – context.

Further, any subset of objects that has a meaning will represent a new object (and correspondently, its context). Because the object formed this way doesn’t have any information, we assume that any information the object might have in additional to context is nothing more but approximation for unidentified context, that in general could be extracted and extended to context.

Reorganization of the system of Formal Knowledge without adding substantially new objects we will call Optimization. (Substantially new objects – are those that couldn’t be formed as subset of existing objects).

 The system of Formal Knowledge S as it described above (presenting all-personal knowledge as a collection of items, fairly independent, and putting them in connections with each other) represents a model of the ‘real’ knowledge. As the system gets filled by gathering evidences the number of knowledge objects tends to increase, so the next faced problem (and step in development) is classification.

In other words, we need object types. Object type seems to us not more abstract than Object itself, so Object Type IS an Object. At the starting point we don’t know if any object could be related to known type, so let’s introduce Unknown Object Type, and all objects in collection will be related to this type unless the specific type will be assigned. Classification sets relation between two Objects, one of which is an Object Type, so we will define the Object Type as:

(Ña Î X) f(a,a)

Saying “Object x is of type a” means

(f(x,a))  (f(a,a)).

Based on the level of complicity we may allow assigning the same object to different types, means:

(f(x,a))  (f(x,b)). (f(a,a))  (f(b,b))

Complex type will be introduced as

(f(a,a))  (f(b,b)). (f(a,b))

As mentioned above, any object is realized in the Context as a set of connections and may be accessed from another related object (that might be a simple object or the object implemented a type) or combination of related objects (types).

Practical Solution

The early realization of described model (Romanov, 2003) is a demo application that demonstrates Personal System of Formal Knowledge about this application (i.e. help file). All knowledge for formalization purposes is represented by separate items (objects) that are stored in the objects table. Each object identified by system generated ID, and contains some information (cell of formal knowledge). Every two object could be connected by directional link. Links are stored in the relations table. The self linked object is a Generic Object (implementation of classification type). The Generic Object could appear 'on the right side' of link only, except self linkage (only simple types are allowed) All other objects are related to one and only one Generic Object (simple classification). The only predefined Generic object is Image. All others could be defined by users. Objects could be Viewed, Modified, Deleted, Copied, and Linked.

Because object could participate in several relations, it might appear multiple times in the tree view (but it will be still the same object), so it is not traditional hierarchical tree. Objects are loaded ‘on demand’ and theoretically tree is endless. Any object could be viewed as starting point. For practical purposes application loads all Generic Objects at startup. The user can follow from object to object by links. Each object is viewed as some information (in current implementation – text or image only) and context – set of links to other objects.

Figure 1 demonstrates relations between main blocks of the application. Data Structure is defined in logical interface that is separated from Storage Area. Storage contains Data Cells – representation of Formal Knowledge objects, Linkage Bank – Context information and Distributed Objects (In current implementation – only image objects). Business Layer is responsible for retrieving and updating Formal Knowledge objects and its Contexts (including Distributed Objects) using Data Structure as a placeholder and Connection Manager. Event driven Client generates requests to Logical Interface based on user interaction.

 

Figure 1: Implementation Of Model of Formal Knowledge

Perspective

Some base definitions are proposed here that are still far away from being finalized, and apparatus is still to be developed. The author is also not confident that knowledge may be formalized at all, so there is a possibility that formalization would be quite limited. The main direction is seen as enhancing the Model and its realization, but if results would be favorable, research may be expanded or integrated into neighboring areas. In the proposed Model two knowledge cells may be related or unrelated, but because knowledge is undivided in out heads and existence of cells is our assumption mainly for modeling purposes – seems as  might be perspective to add some ideas of Fuzzy Logic (Freksa, 1994).

We intentionally didn’t mention the use of Formal Knowledge in exchange processes, assuming this is more complicated issue, but this area seems promising with wide range of practical results particular in knowledge sharing, learning, organizational knowledge, including security issues. In general, we might think about of the System of Corporate Formal Knowledge and connections with intellectual capital, as a dynamic set of Individual Systems, or as a subject matter Personal Knowledge Management System (Smith & McLaughlin, 2004). Correlation between formal and informal knowledge is another aspect that seems to be promising for research. GUI for the System of Formal Knowledge should be attractive and natural for users, and this is yet another research path. A further interesting direction would be developing a measurement or any kind of evaluation applied to the System of Personal or Corporate Formal Knowledge. Knowledge Formalization is a base for any intelligent system.

Among a number of publications related to Knowledge Representation, our research is limited to Formalization of Personal Knowledge and for personal use only. Before introducing classification, the model was nothing more than a notebook, and even now is only like an address book.

References

Freksa, C. (1994), Fuzzy Systems in AI, Fuzzy Systems in Computer Science, Vieweg, Braunschweig/Wiesbaden; http://www.cosy.informatik.uni-bremen.de/staff/freksa/publications/Fuzzy94Freksa.pdf

Genesereth, M. (1998), Knowledge Interchange Format, NCITS.T2/98-004; http://logic.stanford.edu/kif/dpans.html

Gittens, M. (2003), An Anatomy of Knowledge Representation and Meaning; http://www.gits.nl/anatomyOfKnowledgeRepresentation.pdf

Godbout, A.J. (1996), Information vs. Knowledge; http://www.km-forum.org/ajg-002.htm

Godbout, A.J. (January, 1999), Filtering Knowledge: Changing Information into Knowledge Assets, Journal of Systemic Knowledge Management, Vol. 1; http://www.tlainc.com/jkmpv1.htm

Gromov, G.R. (1986),  Auto-formalization – Knowledge Acquisition of Professional Skills , Microprocessor Devices and Systems, Moscow, 1986, N 3, p.80 – 91; http://www.netvalley.com/library/autoformalization/index.html

Igonor, A. (June, 2002), Success Factors For Development Of Knowledge Management in e-Learning In Gulf Region Institutions, Journal of Knowledge Management Practice, Vol. 3: http://www.tlainc.com/jkmpv3.htm

McKinley, S. (December, 2003), Natural Language and the Problem of Modeling Knowledge, Journal of Knowledge Management Practice, Vol. 4; http://www.tlainc.com/jkmpv4.htm

Packin, A.I. (2003), Data Centric Systems, V International Conference “Administration and Modeling Problems of Sophisticated Systems” Samara, Russia; http://packin.ru/artic/datacent.htm

Romanov, L. (2003), SoftDoc – Knowledge Formalization,  Winsite.com; http://www.winsite.com/info/des_package_19000000036527.html

Smith, P.A.C., McLaughlin, M. (January, 2004), Knowledge Management: People Are Important!, Journal of Knowledge Management Practice, Vol. 5; http://www.tlainc.com/jkmpv5.htm


About the Author:

Lev Romanov holds degree in Physics, Mathematics and Education from Kaluga State University (Russia). His earlier works are dedicated to modeling physical processes in turbo machines using Monte-Carlo method. He is working as an independent contractor for various software projects.

Contact: Lev Romanov, 9677 Grandview Dr, St Louis MO, 63132, USA; Tel: 314-997-3968; Email: levrom@hotmail.com