[EXMO] glossary

The present glossary aims at providing the meaning that we assign here to a restricted set of unusual (and, for some of them, buzz) words. It also provides some references to relevant web pages but this is not its main goal. It is not pretended that you will find here the utimate or real true definition of a term, but a moderately precise definition of what it means in the Exmo pages.

We do not pretend to define the term "agent". But in the context of the Exmo web site, this term is used in both a broad and precise way.
An agent will be nothing more than a computer program with two characteristics:

Hence, the minimal need for a well defined Agent Communication Language*. It is noteworthy that the definition of agents can be applied to human beings (if the reference to computer program is withdrawn). When human beings must be included in the meaning of agent, this is explicitly mentioned.

Agent Communication Language (ACL)
An agent communication language aims at allowing agents* to communicate*. Most current languages, e.g., KQML, FIPA ACL, are based on speech act theory*, because it is supposed to provide an idea of the meaning* of an assertion without knowing its content*.

An alignment is a set of correspondences* between entities of two ontologies*. These correspondences* are usually made of the pair of entities (classes, properties, formulas, etc.) and the relation that is supposed to hold between these entities. An alignment is the output of the ontology matching* process.

Alignment server
An alignment server is a networked application providing the services of matching ontologies*, storing, manipulating and retrieving alignments*. They are part of the semantic web* infrastructure.

In a broad sense, an annotation is some element attached to a primary document but not constitutive of it. This can be data about the content (meta-data: author, year of publication), about its encoding (media data: number of pages, encoding), comments from other sources (errata, note) or a representation* of the content* itself (including a summary or a formal reformulation of the content). In these pages, we usually consider this latter, narrower sense.

A representation* is an approximation of another iff the latter contains all the information* availlable in the former.

Argumentation is a process in which parties exchange arguments and counter-arguments for or against positions on a specific issue (arguments being themselves positions). A set of arguments which counters any attacking argument is a said admissible and corresponds to a coherent set of adopted arguments. However, there may be several admissible sets out of a particular set of arguments leading to different conclusions (adopted positions). Hence, parties may resort to preferences in order to select admissible sets or aggregate all admissible sets. Argumentation may be used for selecting alignments* or correspondences* to be used by agents* in order to communicate.

Belief could be defined as endorsed information*. While knowledge* would be true beliefs (though several authors would consider that knowledge is more than true belief, e.g., justified true belief, because it could be accidental true belief).

Collaborative work
Collaborative work is intended here in the sense of "computer-based collaborative work" or CSCW. It aims at providing tools and methodologies for collaborating through computers.
CSCW is traditionnaly divided into three important functions:

Of course, each layer takes advantage of the preceding ones, but each one raises its own problems. The former are heavily system-related although the farther we go the closer we are to the object of the collaboration.
We can make two further distinctions between:

Collaboratory is a buzword for a computer environment for collaborative scientific research. It usually includes tools for collaborating, e.g., shared editors, electronic communication, and tools for domain-independent, e.g., text-editing, bibliography maintenance, and domain-dependant research tasks, e.g., genetic sequence alignment or socio-economic simulation. In fact, it is an environnement for computer-supported collaborative* intellectual work.

The fact that one entity has some direct influence on another (through a perceptive channel and not because of resource availability).

The content, generally speaking, is the meaning* of a document. In the present context, this word will denote the meaning* that we want to represent*. Often, content is opposed to structure, but structure itself can be seen as a rough type of content.

The part of a situation in which an assertion has been issued which is not part of the assertion. Understanding the meaning* of an assertion may require the understanding of its context. The context impact on communication is studied by pragmatics*.

A correspondence expresses a relation supposed to hold between entities of two ontologies*. It essentially identifies the relation and the two entities. Relations come from a set of relations which may be some arbitrary set or a relation algebra. Entities may either be element of the ontologies* or expressions built from these ontologies, e.g., queries.

Correspondence pattern
A correspondence pattern generalises a set of correspondences* by setting some of its elements as places or variables. They may be used for matching* ontologies by trying to instantiate such patterns in the ontologies*.

The study of the process of ascribing a meaning* to a text (interpretation*) with regards to (cultural, historical...) context*. It gave birth to interpretative semantics. Hermeneutics can be considered part of pragmatics* and covering rhetorics*.

Generally, a fact is something which characterises a particular situation, e.g., "the sky is grey". Such a fact can be more or less informative. It seems that the commonsense definition of information is "data in context*". This meets the cybernetic definition in which information is a knowledge* element brought by a message, e.g., if I know that the sky is grey, the statement above is not informative. The information is measured as a reduction of uncertainty by the sole fact that the message content has been chosen among a particular set of possible contents.
This shows that information is not characterised by a signification, but by a modification of the knowledge* of the receptor. The context* is made of the knowledge about the set of possible messages and the place of the present one in this set. In that vision, the measure of information is provided by the context*.

Information flow
There are two meanings for information flow. The first, general, one is synonym of knowledge flow*. The second one refers to a particular formalisation of the flow of information by Jon Barwise and Jerry Seligman.

The process of (re-)constructing the meaning* of an assertion. In semantics*, it is given by an interpretation function which provides a mapping from terms to their meaning*. In computers, it can be a real process of computing directly the instructions of a programming language in order to provide the result.

The term knowledge is not easy to specify. Many people consider that "knowledge" can be only carried by "intelligent beings". They place the opposition between "knowledge" and "information*" in the presence of such a being for assimilating this "knowledge". Hence, there cannot be disembodied meaning*.
One can define knowledge as a set of imaterial resources which can be mobilised (as cognitive or informational processes) to achieve a goal. So, indirectly, knowledge is tied to action. Competence is then the capacity to take advantage of knowledge in actions.
The term "knowledge" is used in these pages as it is in the "knowledge representation*" field of artificial intelligence. It is possible, that the aforementioned people consider that "knowledge" is represented as information*. But, the goal of these works being to carry and interpret knowledge, the discipline is not badly named.
Another difficulty in French comes from the fact that the English "knowledge" covers both "connaissance" and "savoir". In English, "knowledge" is opposed to "belief*" in that it can be considered as a "true belief". This concerns the "savoir" meaning of the word only.

Knowledge flow
The knowledge flow (sometimes called information flow*) is the diagram showing the flow of knowledge* through an organisation (see figure 1). It is decomposed into atomic knowledge flow from one knowledge repository to another. The knowledge flow looks like what is best known as workflow*. The main difference is that workflow is task-driven* although knowledge flow is content-driven*. In fact, a workflow is a coordination and control diagram* and knowledge flow is a communication diagram*. However, coordination* require information* exchange and thus most of the workflow* models contain the knowledge flow. The reverse does not generally hold.

Knowledge management
In our view, knowledge management aims at providing the knowledge laying in a firm (or that can be acquired) to those who needs it. It thus modify the information flow* and storage (for satisfying future needs). It must be a long term effort and cannot be uniquely driven by computer, social or organisational aspects.
Because we are interested in formal knowledge representation* in computers, we are willing to apply this technology to knowledge management.

Knowledge medium
The term "knowledge medium" has been coined by Mark Stefik. It describes the way people can communicate through a knowledge* enhanced device. This is what the Exmo project is about.
The term is also used as the name of a laboratory, viz. the knowledge media institute at the Open university, and a NetAcademy on Knowledge Media.

Knowledge representation
(see also representation*).

Knowledge server
We coined the term "knowledge server" (and pretend it should be put in the public domain) in order to describe a system that provide knowledge* (in the sense above) on a client-server basis. There is no mystery under that term. The idea is that knowledge is accessible from a remote place and on demand. It can be available as HTML, XML or RDF.
This can be achieved in a variety of ways among which the use of the Worldwide web is the favorite one.

Linked data
Linked data is an interoperable way to publish data on the web. The four principles of linked data are:

When using semantic web* technologies (HTTP for dereferencing, RDF/OWL for expressing data and links, SPARQL for querying it), linked data provide a distributed and interoperable web of data. This is still heterogeneous when various ontologies* are used for expressing data but ontology alignments* may help solving this last problem.

See ontology matching*.


Mediology has been coined by Régis Debray as the analysis of the superior social functions (religion, ideology, art, politics) in their relations to the means of transport and transmission.

Usually, a model is a representation* abstracting* a phenomenon, i.e., retaining the most salient feature for the observer.
In model theory*, a model is a particular phenomenon than can be modelled by the representation* . This is quite the opposite.

Model theory
Logicians have developed powerful ways to provide the semantics of a language by model theory. Roughly, they provide a structural way to interpret a language over a domain. A model, of a set of expressions in the language, is an interpretation* that is coherent with all the expressions. The meaning* of a term, is then the union of its interpretations* in all the models.
The interesting feature is that, if someone thinks of a particular interpretation* of a language (or rather a set of expression), it must be a model. Then every processing on the language, which is based on all the models, is coherent with this user's interpretation*.
If this does work for a user interpreting consistently the set of symbols (s)he manipulates, it is not particularly adapted to several users interpreting the symbols diversely.

Network of ontologies
A network of ontologies is made of a set of ontologies* and a set of alignments* between them.

In philosophy, ontology is the study of being in themselves. This word has been borrowed by the domain of knowledge-based systems, for denoting a knowledge base with a sufficient universality for being shared by many people. The official definition of an ontology is "the specification of a conceptualisation" (two steps away from the beings in themselves). The term has gained even more popularity and is now used for many different things. The main shift is now from the concepts to the nouns: in fact, a terminology, a dictionary, a vocabulary, a mere list of words are called ontologies.
Technically, an ontology is a set of assertions that specifies the concepts involved in the domain. A simple list of concepts can qualify as a first ontology. The next grade (where many stop) is a hierarchy of concepts. This hierarchy is even more usable if it is a taxonomy: a hierarchy in which the individuals denoted by a concept are also denoted by its ancestors. Then the concepts can be described through theirs properties and the relationships in which they are involved with other concepts. Ultimately, there are a few ontologies which are axiomatisation of the domain.
Due to its fuzzyness, we have for long tried to avoid using this word. Now we have given up. For us, an ontology is a logical theory (even if the logic may be very weak).

Ontology matching
The process of finding correspondences* between entities of two ontologies. The result of this process is an alignment*.

Here pragmatics is considered as the account of context* in the interpretation of signs* (and this context encompasses the respective knowledge the interlocutors have of themselves or the dialog going on). It is thus a wider perspective on semiotics*.

A protocol is a set of rules that govern a dialog and provides a precise interpretation frame for assertions. Protocols are used by agents* restricting the meaning of assertions when communicating (and enhancing understanding). The protocol used in a communication is part of the context* studied by pragmatics*, it can also be considered as part of the interpretation of one compound sign*.

Representation refers here to a set of sentences in a (formal) language that must account for a particular situation.

We consider rhetorics in the narrow sense of the study of techniques that improve the conveying of a meaning* when communicating. In particular, it covers argumentation*. Rhetorics is thus wider than semiotics* and narrower than pragmatics*.

Some people thinks that by using identifiers with meaningful (for them) names instead of id#356, id#873, they "put more semantics in the system". For the system, in fact, the result is the same. Moreover they did not do anything with semantics: all this is syntax. There is a good note on the XML semantics topic which starts well but ends in the same thinking it first criticised.
In fact, the semantics of a language has to be given in relation with the meaning* and this cannot be done by syntax.
In logic, this is usually done by the mean of model theory*.

Semantic social network
A semantic social network is made of a set of users related by social relations. These social relations as well as user profiles may be expressed with respect to different user-specific ontologies*.

Semantic peer-to-peer system
A semantic peer-to-peer system is made of a set of peers annotating some resources with the help of an ontology* and having other peers as aquaintantes. Peer ontologies may differ, hence they use alignments* between their ontologies in order to communicate. Peer may communicate by querying their aquaintances about their resources. The query is transformed according to available alignments before being evaluated.

Semantic web
The semantic web is a web whose content can be processed by computers. It can be thought of as an infrastructure for supplying the web with formalised knowledge* in addition to its current informal content*.

The study of sign systems* is named semiotics (Charles Pierce, in French semiotique) or semiology (Ferdinand De Saussure, in French semiologie). It aims at considering signs* in a wider framework than (natural would say the computer scientist) language in order to globally consider meaning* attribution. In fact, it has been developed (in the seventies) mainly for structured languages: texts and movies. If these matters are the occasion to develop theories in the language of human sciences, one can find more systematic studies that can support a computer treatment. This is mainly true in America where disciplinary isolationism can be weaker.

Algebraic semiotics
The idea of algebraic semiotics, has been put forth by Joseph Goguen in the continuity of his work on institutions and in order to consider representations* (or specifications) in relation with their use. To that extent the representations* are called sign systems*. They include the syntax and semantics* of the considered language and - as far as possible - information for interpreting the representation*. A semiotic morphism transforms such a sign system* (the source) into another (the target) supposed more suited to a particular use. Algebraic semiotics studies sign systems and morphisms in the framework of category theory.

Computational semiotics
Computational semiotics is used by a variety of people. The basic idea is that the computer (in fact any computing device) can be used to manipulate signs*.
For some authors (see the 1st IWCS), it is restricted to manipulate data and to display it in a meaningful way. For others (see Ricardo Gudwin), it can be equated with artificial intelligence as a whole. For yet others (see Wolfgang Mack), it aims at reproducing on computers the emergence of signs* in a society (semiosis). Computer semiotics is a part of computational semiotics that consider computer programs as sign systems [Andersen 1990]. The approach considered here (see Gerd Döben-Henisch) is that computational semiotics is defined by "the existence of an algorithm, a given formal semiotic structure and a mapping function between both".

A sign is analysed in two parts: a signifier which is the form taken by the sign and a signified which is what it represents (see Daniel Chandler). Linguists insist on the arbitrary character of the association (there is no a priori reason to tie signifier and signified together outside a sign system*). The form can be a word, a sound, a picture or a complex of sound and successive images in a motion picture for instance.
Although semioticists consider the existence of signs in a sign systems*, others consider that the existence of a sign is related to an interpreter (and its context*) which relates the signified to the signifier. In fact, it is enough for the interpreter to know (and to identify) the sign system* to which the sign belong in order to perform the interpretation*.

Sign system
A sign system, as known as semiotic* code (see Daniel Chandler), is a set of rules and conventions for interpreting signs* and messages made of these signs.

Speech act
Speech act theory deals with the interpretation* of natural language assertions (or rather "utterance") as acts rather than truth assertions. In a speech act, one distinguishes

For instance, the utterances "The conference is open", "I declare open the conference" and "I consider the conference as open" carry the same propositional content, i.e., that the conference is open, although their meaning* is not the same: the first one is an assertion (assertive), the second one an act which commits the speaker and the hearer (commisive) and the last one a statement about the speaker's feelings (expressive).
This theory has been widely used in agent communication language* in order to design agents* that can react to a message because they know its illocutionary function while they do not understand the propositional content. Doing so, the ACL* can be independent from the content language.

A transformation is a computational way of generating one or several representations* from one or several other representations* (not necessarily written in the same formal language). It is noteworthy that a transformation, taken generally, does not differ from a program. In general, here, the word is used in a narrower meaning.

CSCW* software able to deal with the coordination* of people in an organisation by managing the tasks to handle, their completion statuses, the people who must carry them out and the follow-up tasks. This term can also refer to the coordination schema (or workflow diagram) implemented by the organisation. See the Workflow management coalition.

© | ? | *


Feel free to comment to Jerome:Euzenat#inria:fr, $Id: gloss.html,v 1.15 2013/10/01 14:38:28 euzenat Exp $