Lars Ludwig
SemanticWeb.com Contributor

I have been developing, testing, and researching an innovative semantic
knowledge management system called ArtificialMemory
for more than 5 years.

ArtificialMemory is a personal and enterprise knowledge management system integrating wikis, personal information, blogs, and document management systems.

Here are (the
first) 5 of the important lessons learned. These lessons often contradict common
opinions on Semantic Web technology and usage stemming from the
dominant artificial intelligence and knowledge representation
research tradition.

1. Simple ontologies are as important as complex ones

The foundation of any Semantic Web application is at least one
ontology scheme. An ontology scheme can be defined more or less
rigidly. A
thesaurus
would be a simple form of ontology, an OWL scheme would be a
more refined
kind allowing for some logical inference of new knowledge. A
well-structured
/ rigid ontology definition forces or tempts one to more often
work on the
ontology scheme instead of inserting data thus preventing one
from getting
things done fast. In my experience, combining simple ontologies
such as
thesauri with more complex ones such as RDFS or OWL is the way
to go. It
is
important to be able to just enter a keyword / tag to denote or
relate to
another word or object instance as well as to be able to create and
instantiate a well-defined object class.






More From Jupitermedia

Military, Universities Team Up on Big CALO Project

Radar’s Twine Ties the Semantic Web Together

A Snapshot of Semantic Web Trends

Semantic Web as Competitive Advantage

Smartening Up Your Links

If you want to comment on these or any other articles you see on Intranet Journal, we’d like to hear from you in our IT Management Forum. Thanks for reading.

- Tom Dunlap, Managing Editor.

2. Full-text search is far less, entity search is far more
important in Semantic Webs than in Document Webs

When one starts storing one’s knowledge in highly interrelated
sub-document
information chunks like semantic webs, full text search becomes
obsolete. Knowledge refers to object instances.

In semantic
knowledge webs,
information is related to object instances and thus can be
easily found.
Searching most times merely consists of jumping to an instance
of interest
and identifying the relevant statements / information. What is
important in
Semantic Webs is a sophisticated auto-completion (that is fast)
search box
for instance names (relevant entities).

3. Adding information to Semantic Webs has to be as easy
as in
natural
language notes

Examining the current Semantic Web applications, one will find
that they
tend to either make it very simple to add information by more or
less automatic data imports, or very difficult by forcing users
to annotate
manually using a
multi-step process or complex meta-language (RDF, OWL). The
first solution
will soon create an information pool not reflecting the user’s
memory and
thus adding irrelevant information that endangers use and usability of the
ontology data store (it’s like getting a huge dictionary instead of the
vocabulary you already know and want to relate to).

The second
approach slows down the process of adding information to such a
degree that less
statements and instances will be created and more information is
stored in
a
traditional documents invalidating the Semantic
Web. The information added to a (personal) knowledge store has
to be
inserted consciously using a limited set of natural language
expressions creating structured information fast and easy.

4. Information in Semantic Webs has to be normalized

Adding information to a Semantic Web has to take place in a
controlled manner. Duplication could happen in at least three
ways: entity
duplication,
statement / triple duplication, and text duplication. Entity
duplication can
many times be avoided by instance name checks. Nonetheless, entity
duplication will occur. Therefore, being able to easily merge
entities is
of
great use. Triple duplication has to be avoided by consistency
checks that
are not part and parcel of standard Semantic Web ontology data
insertion [an
inference engine or SPARQL statements would have to be used].
Avoiding text
duplication is one of the toughest parts: It does imply that the
user is
able
to reuse natural language text chunks as can be found in value
properties as part of Semantic Webs.

In order to enable reusing
NL-text chunks, it
must
be possible to integrate them in an indefinite number of text
sequences simulating traditional documents. Thus documents become
sequences of
highly
normalized text entities. For example, if each section of a book
or each slide of a presentation becomes a singular entity, all sections /
slides
can
be reused in other presentations or books. Changing a text
section or
slide
will change it in any book / presentation using it. Traditional
documents are just sequences of NL-text entities that can be
replaced by
semantically
enriched entities in Semantic Webs.

5. Any Semantic Web has to be a personal, integral
information store

That’s something not yet widely understood. Any knowledge is
personal. Adding semantics to data is giving meaning to data. The
Semantic Web is
not
the same as the data web. Searching in other people’s Semantic
Webs can
result in data meaning nothing to you. What is meaningful to you
has to
somehow be
related to your (personal) knowledge (Semantic Web). Meaning is
knowledge, is personal. An ontology is personal, even if it
happens to be
shared by
many people. In this case, it’s just several times personal. The
Semantic Web thus is personal in character, made of individual
memory reflections.
It
is in the nature of Semantic Webs that they represent personal!
knowledge, and the knowledge of one person is integral by nature.
That is, in
principal, each and every one of us has only one category
representation (however different it may be) for, say, ‘dog’, and
whenever we say ‘dog’
we
actually mean our concept of ‘dog’.

What Semantic Webs can help
us achieve
is to integrate personal utterances into a single personal
Semantic Web
reflecting and extending an individual’s memory / knowledge, and
usingthis
to communicate, to relate to and learn from other people’s
knowledge, to
automatically match and select people by their knowledge, to
pass on and
distribute changes of knowledge, and so forth. A Web of interrelated
Semantic Web is a Web of Data. I guess we will proceed from a
Web of
depersonalized documents to personal Semantic Webs, and from
there to a
Web of personal data
that can be acted upon intelligently.

Lars Ludwig is a cognitive psychologist with a professional and scientific IT
background living in Germany.

RELATED: