WO2000020964A1

WO2000020964A1 - Fractal network n-ter order for handling complex structures

Info

Publication number: WO2000020964A1
Application number: PCT/EP1999/007137
Authority: WO
Inventors: Günter Schmidt; Maria Athelogou; Martin Baatz; Andrej Kharadi; Jürgen Klenk; Peter Blöchl; Gerd Binnig
Original assignee: Definiens Ag; Ibm International Business Machines Corporation
Priority date: 1998-10-02
Filing date: 1999-09-24
Publication date: 2000-04-13
Also published as: JP2002526856A; EP1116108A1

Abstract

The invention relates to a fractal network which is provided for handling complex structures and which is comprised of a multitude of units. The fractal network contains semantic units which respectively comprise information contents, and has logic units which describe a relationship content. The relationship content respectively links two semantic units in such a way that the mutual relationship of both linked semantic units is determined by the relationship content.

Description

description

Fractal n-th order network for handling complex structures

The invention relates to a n-th order fractal network for handling complex structures and in particular to a fractal or fractal-hierarchical network with a large number of semantic units, by means of which semantically structured information can be analyzed and processed.

With the advancing transformation of the industrial society towards the information society, there is an increasing need for a tool for processing the growing flood of information. In particular in the field of image recognition, speech recognition and simulation, in-depth investigations were carried out in order to simplify the recognition, modification and utilization of complex structures, such as speech and images.

Such systems in the prior art, however, suffer from low flexibility and an extraordinarily complicated provision and processing of the data or information used. The data to be processed is also essentially static.

Processing of such data is extremely difficult or even impossible, particularly in the case of dynamic complex structures or chaotic technical systems.

It is also known in the prior art to use the XML or eXtended Markup data description language

Language (a derivative of SGML, IS08879) information treat content in a structured manner. The structuring can be semantic. Semantic means here that references of one information content to another information content can have a meaning. It is possible to formulate metadata, that is, data that describes data. In the data description language XML, however, it is not possible to store information about processes in such a way that they can flow into a data analysis and into an "intelligent" behavior of a semantic network itself.

The current state of the art in the field of knowledge about processes is reflected in methods and processes for pattern recognition and simulation. Although the methods currently used are very mature, there is no knowledge of objects in their semantic context. Accordingly, in a simple observation for clarification, it can be said that, for example, a pattern recognition currently used does not know that "a coniferous forest is generally a forest" and "a bridge often crosses a river".

The present invention is therefore based on the object of creating a fractal n-th order network for handling complex structures, which enables information or knowledge to be stored in a structured form and to be analyzed and linked on the basis of this data.

This object of the present invention is achieved by the features specified in claim 1.

More specifically, according to the present invention, a fractal network for treating complex structures is created, which consists of a multiplicity of units.

The fractal network contains both semantic units, that each have information content, as well as linking units that describe a relationship content. The relationship content links two semantic units in such a way that the mutual relationship of the two linked semantic units is determined by the relationship content.

The central element here is the semantic unit, which represents an "object" or a "process of the world" as a data structure. An essential feature of the semantic unit is the ability to store information content in a structured manner and to network with other semantic units. So that two semantic units can be linked in such a way that the link has meaning or is semantic, these semantic units are connected to one another with the special linking units. Such a linking unit can, for example, also be provided implicitly in a structured information content of a semantic unit.

These linking units can be a special form of semantic units that can have information content and relationship content.

In order to be able to carry out clear operations in the "world knowledge" present in the fractal network, each semantic unit can be assigned a label which is unique within this "world knowledge".

Furthermore, there is also the possibility of creating a data structure which makes it possible to change information or knowledge already existing in the fractal network at any time and to add new parts at any time. In that knowledge not only contains information about objects, but also knowledge about information processing processes, the content and structure of the knowledge can be changed in a dynamic process.

Complex structures can represent language, images, networks or chaotic systems, such as technical, cultural, economic or ecological relationships.

Further advantageous embodiments of the present invention are the subject of the dependent claims.

The present invention is explained in more detail below on the basis of exemplary embodiments with reference to the accompanying drawing.

It shows:

Fig. La to le different types of link units used in the embodiments of the present invention;

2 shows an illustration of a fractal network of the n-th order according to a first exemplary embodiment of the present invention;

3 structured information contents and relationship contents in semantic units or link units according to the first exemplary embodiment of the present invention; 4a and 4b representations of further fractal networks according to the first embodiment of the present invention;

5 structured information content in semantic units with attributes according to the first embodiment of the present invention;

6 shows an illustration of a fractal network of the n-th order according to a second exemplary embodiment of the present invention;

7a and 7b representations of a semantic network according to a third exemplary embodiment of the present invention;

8a and 8b representations of a semantic network according to a fourth embodiment of the present invention; and

9a to 9c representations of a semantic network according to a fifth embodiment of the present invention.

The following is a description of exemplary embodiments of the present invention.

Before the embodiments of the present invention are described in detail, the following should be said. A nth order fractal network for handling complex structures is generally expressed as a multitude of units. The fractal network contains both semantic units, which each have information content, and linking units, which describe a relationship content. The relationship content links two semantic units in such a way that the mutual relationship of the two linked semantic units is determined by the relationship content. The term "semantic" used here means "to have a meaning".

Linking units of this type can represent a special form of semantic units which can have information content and relationship content.

In addition to linking semantic units via linking units, there is also the possibility that one or more linking units in turn via one or more linking units with one or more semantic units and / or one or more linking units in turn via one or more linking units with one or more Linking units are linked, as can be seen from the following description.

Such relationship contents of linking units can generally be freely selected by a user. However, it makes sense to define some elementary relationship contents of linking units in a basic library in advance. Exchange relationships and relations can be viewed as elementary relationship contents of linking units. Exchange relationships are defined as relationships that describe an abstract, material and / or communicative exchange between semantic units. Relations, on the other hand, are those relationship contents of relationships linking units that describe any relationships between semantic units.

Figures la to le show some such elementary linking units that describe a relationship content.

In the case of hierarchically structured knowledge, such as in the fractal network, link units of the type exchange relationship can be further divided into two groups.

In FIG. 1 a, a link unit 1 of the type exchange relationship is shown, which connects semantic units in different hierarchical levels of the fractal network of the nth order. The nature of the relationship from a larger, i.e. higher-level, semantic unit to a smaller, i.e., lower-level, semantic unit and vice versa is described. In other words, this means that the scale is changed. Linking units with relationships that have the two characteristics mentioned, that is, an exchange and a change of scale, are referred to below as linking units of the type VA / VS. In the case of the expression "VA / VS", the expression "VA" stands for "exchange" and the expression "VS" for "scale change". Simply put, such a linkage unit 1 of the VA / VS type can be viewed in the direction of the arrow from A to B shown in FIG. 1a as "A contains B" and in the opposite direction as "B is part of A". This corresponds to the definition of an embedding hierarchy.

In Fig. 1b, link units 2, 2a and 2b of the exchange relationship type are shown, the semantic units in the same hierarchical levels of the fractal network Connect nth order with each other. In other words, this means that no scale change is carried out. Linking units with relationships that have the two characteristics mentioned, ie an exchange and no scale change, are referred to below as linking units of the type VA / VH. In the case of the expression "VA / VH", the expression "VA" stands for "exchange" and the expression "VH" for "no change of scale". Simply put, such a VA / VH-type link unit 2a in the direction from A to B as "A is the input variable of B" and in the opposite direction as "B is the output variable of A" and such a VA / VH-type link unit 2b in Direction from A to B as "A is described by B" and in the opposite direction as "B is attribute of A".

Likewise, in the case of hierarchically structured knowledge, as in the fractal network, link units of the relation type can be further divided into two groups.

FIG. 1c shows a link unit 3 of the relation type, which connects semantic units in hierarchical levels of the fractal network of the nth order that are different from one another. The nature of the relationship from a more general semantic unit to a more specific semantic unit and vice versa is described. In other words, this means that the scale is changed. Linking units with relationships that have the two characteristics mentioned, ie a relation and a change of scale, are referred to below as linking units of the type VR / VS. In the expression "VR / VS", the expression "VR" stands for "relation" and the expression "VS" for "change of scale". Simply put, such a linking unit 1 of the VR / VS type can be in the direction of the arrow shown in FIG. 1c from A to B as "A" in particular B "and vice versa as" B is generally A ". This corresponds to the definition of a similarity hierarchy.

FIG. 1d shows link units 4, 4a, 4b and 4c of the relation type, which connect semantic units to one another in the same hierarchical levels of the fractal network of the nth order. In other words, this means that no scale change is carried out. Linking units with relationships that have the two characteristics mentioned, ie a relation and no scale change, are referred to below as linking units of the type VR / VH. Accordingly, in the expression "VR / VH" the expression "VR" stands for "relation" and the expression "VH" for "no change of scale". Simply put, such a VR / VH type link unit 4a may be "A is (location-related) adjacent to B", such a VR / VH type link unit 4b may be "A is similar to B", and such a link unit 4c may be of type VR / VH in the direction from A to B can be viewed as "B follows A" and in the opposite direction as "A will be followed by B".

Fig. Le also shows a further link unit 5, which can be viewed in the direction from A to B as A has Janus / function B "and in the opposite direction as" B is Janus / function of A ". For a more detailed description of this link unit 5 reference is made to the following description of the exemplary embodiments.

Finally, it should be noted that, as can be seen, link units can be both directional, that is, directional, and bidirectional, that is, non-directional. A description will now be given of a first embodiment of the present invention.

2 shows a simple fractal network, the interaction of the above-described link units with other semantic units that are present in the fractal network being illustrated.

In Fig. 2, reference numeral 3 denotes a VR / VS type link unit, reference numeral 4b denotes a VR / VH type link unit, and reference numeral 6 denotes respective semantic units.

If the sentence "Human is generally mammal" is to be represented in the "world knowledge" available as a fractal network, the semantic units 6 labeled "Human" and "Mammal" with the directional, that is, directed, linking unit 3 of Type VR / VS, more precisely of the type "is / is in particular", is linked to one another. If the statement is added that "monkey and human are 95% similar in the context of genetic analysis", the semantic unit 6 labeled "monkey" with the semantic unit 6 labeled human is replaced by a bidirectional linking unit 4b of the VR type / VH, more precisely of the type "is similar to", linked together. The linking unit 4b has a weighting of 95% in its information content. The linking unit 4b is also linked via a linking unit of the type “in context” (not previously explained) to the semantic unit 6 labeled “gene analysis”. FIG. 3 shows structured information contents and relationship contents of the semantic units or linking units shown in FIG. 2.

In the upper part of FIG. 3, the information contents of the respective semantic units of FIG. 2 are shown, which contain an identifier, a name and identifiers of the link units connected to them. Thus, the semantic unit 6 labeled "human" in FIG. 2 has an identifier "1" and the name "human" and is linked to linking units which have identifiers "12" and "13". The semantic unit 6 designated "mammal" in FIG. 2 has an identifier "2" and the name "mammal" and is linked to the linking unit which has the identifier "12". The semantic unit 6 denoted by "monkey" in FIG. 2 has an identifier "3" and the name "monkey" and is linked to the linking unit which has the identifier "13". Finally, the semantic unit 6 labeled "gene analysis" in FIG. 2 has an identifier "4" and the name "gene analysis" and is linked to a linking unit which has the identifier "134".

In the lower part of FIG. 3, the relationship contents of the respective link units of FIG. 2 are shown, which contain an identifier, a name, identifiers of the link units possibly connected to them, identifiers, the semantic units or link units linked by them, and the type of this link included. The link unit 3 shown in FIG. 2 has the identifier "12" and the name "is in general", it is not connected to any other link unit and links the semantic unit of the identifier "1" directionally with the semantic unit of the Identifier "2". The link unit shown in Fig. 2 4b has the identifier "13" and the name "is similar to", it is connected to the link unit with the identifier "134" and bi-directionally links the semantic unit with the identifier "1" to the semantic unit with the identifier "3" , with a weight of 95%. Finally, the linking unit shown in FIG. 2 "in context" has the identifier "134" and the name "in context", it links the linking unit 13 directionally to the semantic unit 4.

If the relationships shown in FIG. 3 are represented graphically, the fractal network is accordingly shown in FIG. 2.

Generally speaking, it should be noted that the information content described by a semantic unit represents an identifier and / or a list of the linking units which connect this semantic unit with other semantic units, the identifier preferably being a name or a serial number and the information content likewise being preferred is in a structured form.

The link units describe relationship content which, in addition to information content, also contains a link identifier. This linkage identification describes the respective identification of the semantic units and / or linkage units that are linked by them, one or more directions in relation to these linked semantic units and / or linkage units and / or weightings that describe one or two directional information.

As can be seen from the first embodiment, there is also the possibility that a link unit is linked to a semantic unit via another linking unit. In addition, the relationship content of the linking unit can optionally contain information about the respective type of linking the related semantic units, this type of linking possibly providing additional information about a relation, ie, a comparison of the respectively linked units, and / or about one Exchange relationship, ie, contains a one-way or two-way interaction of the linked units, and the type of link also contains additional information as to whether or not a scale change is taking place. In the case of an exchange relationship, this information regarding a change in scale can be the type of relationship to a larger, ie, superordinate, or to a smaller, ie, subordinate, semantic unit or vice versa, or the type of relationship to a more general or to a more specific semantic unit , describe.

FIGS. 4a and 4b show further fractal networks according to the first exemplary embodiment of the present invention, which serve to facilitate understanding.

4a shows a fractal network in which a semantic unit 6 designated "forest" via a link unit 3 of the VR / VS type, more precisely of the type "is in general / is in particular", with a segment designated semantic unit 6, wherein the linking unit 3 of the VR / VS type also contains a weighting of 70%, which results in the statement “segment classified as 70% forest.” The linking unit of the VR / VS type can be used more precisely here VR / VS (+) are referred to, since there is obviously a change of scale to a smaller scale from the semantic unit 6 labeled "Wald" to the semantic unit 6 labeled "Segment" results, whereby the smaller scale in the present application example results from a smaller fuzziness in the attributes of "forest" and "segment" not described in more detail. In the above example, a similarity hierarchy is formulated, whereby if the weighting (here 70%) in the information content of the linking unit is not clear, the role of a measure for belonging to a corresponding class (here "forest") is assigned. If the linkage unit 1 of type VA / VS, more precisely "consists of / is part of", is considered, the statement "forest consists of trees" is created, which implicitly expresses that a tree is significantly smaller than a forest is on a lower or finer scale.

4b shows a fractal network in which a semantic unit 6 designated "Peter" is linked to a semantic unit 6 designated "Paul" via a link unit 4 of the VR / VH type. Furthermore, the linking unit 4 of the VR / VH type is linked to a semantic unit 6 labeled "friendship" via a linking unit 2b of the VA / VH type, more precisely of the type "is described by / is attribute of". In the end, the statement "Peter and Paul are friends" is obtained, since the linking unit 2b, with the help of the semantic unit 6 designated "friendship", describes an abstract exchange ("friendship") in more detail.

Finally, it should be noted that with the help of linking units of the VR / VH type, that is to say relations without changing scales, associations and comparisons can be defined. Here it is often useful to understand the weighting in the information content of the linking unit as a measure of the similarity of the linked semantic units. The statements are for this "Humans are 95% similar to monkeys" and "winter is followed by spring" examples.

5 shows structured information contents of semantic units with attributes according to the first exemplary embodiment of the present invention.

Each semantic unit can store data and functions of any form in its information content. According to the first exemplary embodiment of the present invention, the name of the semantic unit and its identifier have already been described. In addition to or instead of static data, information contents of the semantic units and / or linking units can also contain algorithms, functions and / or mathematical formulas.

Furthermore, there is also the possibility that semantic units have information content that represents attributes, these attributes describing other semantic units or linking units in more detail (see, for example, the semantic unit 6 labeled “friendship” in FIG. 4b). The fractal network has special link units that serve to link semantic units that represent attributes with those semantic and / or link units to which these attributes are assigned (see, for example, link unit 2b in FIG. 4b). These special linking units 2b are designated by "is described by / is attribute of".

For example, these attributes can contain values that are elements from a set, range, list, or other ordered or unordered structure. This ordered or disordered structure can through one or more numbers, vectors in n-dimensional spaces, text data, image data, video data, audio data, calendar data, tables, geometry data, geographic data, fuzzy logic quantities, Internet content or bundled data or a combination of these be formed in order to be able to store "world knowledge" in an advantageous manner. An example of this is shown in FIG. 5, but because of the self-describing character of this figure, a more detailed description thereof is omitted.

A description will now be given of a second embodiment of the present invention.

An essential feature of the second exemplary embodiment of the present invention is the possibility of including special semantic units in the fractal network which are able to perform certain operations on other semantic units. These special semantic units will be referred to as semantic Janus units.

In this context, a semantic Janus unit 6 (see FIG. 6) denotes a special semantic unit that has an algorithm or a collection of algorithms that change the information content of semantic units and / or generate new semantic units or existing semantic units can destroy. A semantic Janus unit is connected to one or more semantic units, in the vicinity of which the semantic Janus unit is to operate, via a special linking unit 5 (see FIG. Le) of the type "has Janus / function / is Janus / function of" . This means that the functionality of the semantic Janus unit is restricted in such a way that it is only able to carry out the specific operations on those semantic units which are in a predetermined neighborhood area of a semantic unit linked to it. Furthermore, a semantic Janus unit can be linked to other semantic Janus units and / or to attributes via one or more linking units.

In detail, a Janus semantic unit can perform one or more of the following operations: creating new semantic units; the bundling of already existing semantic units into a single semantic unit, which may have to be newly created; changing and / or deleting existing semantic units; comparing existing semantic units; acquiring and changing values of the attributes of semantic units; executing an algorithm and / or calculating a function; the detection of a janus or a part of a janus, that is to say the classification of an algorithm or a part of an algorithm.

The essential task of a semantic Janus unit is the bundling and context of information content. Bundles are to be understood here as the calculation of information contents of a semantic unit serving as a center from the information contents of neighboring semantic units. Contexts are to be understood as the inverse process analogous to bundling, that is to say information contents of the neighboring semantic units are changed as a function of the information contents of the semantic unit serving as the center, which defines the neighborhood. In this way it is possible, for example, in a simple way, constantly to receive current statistics of a set of semantic units (bundles) or to forward changes of framework conditions to a set of semantic units (contexts).

FIG. 6 illustrates an nth order fractal network that is used to clarify the statements made previously regarding the second embodiment of the present invention.

The fractal network in FIG. 6 serves to correctly average an average income depending on the respective framework conditions.

6 shows a semantic unit 6 designated as "law firm MM", which is linked via a link unit 1 of the VA / VS type to the semantic units 6 designated "Müller" or "Maier", which links the Art "law firm MM contains Müller / Müller is part of law firm MM" and "law firm MM contains Maier / Maier is part of law firm MM". In this exemplary embodiment of the present invention, the semantic unit 6, which is referred to as "law firm MM", is via a linking unit 5, that is to say a linking unit of the type "has Janus / function / is Janus / function of", with a "Bundle" designated semantic unit 6, which accordingly acts in this exemplary embodiment of the present invention as a semantic Janus unit with respect to the semantic unit 6 designated "law firm MM". The input type of this semantic Janus unit is the attribute type to be bundled, that is, in the case of this exemplary embodiment, the income, which consists of the law firm's individual income. The semantic Janus unit uses an attribute as an output variable in which the middle income is written. A key advantage of this type of statistical survey is that when a lawyer is added or removed to or from the firm, no changes to the mean income calculation procedure are required.

A description will now be given of a third embodiment of the present invention.

A major advantage of the Janus unit described above is that it only acts locally in a defined neighborhood. Accordingly, it is important to define the concept of neighborhood more precisely. This is done below in this third embodiment of the present invention.

The concept of neighborhood is closely related to the concept of distance. A first semantic unit is then defined as being adjacent to a second semantic unit if a distance between them is less than a predetermined or calculated value, that is to say a limit value. A measure of the distance depends on the information and / or meaning content of the semantic units, via which the second semantic unit can be reached from the first semantic unit.

For example, it is possible to calculate the measure of the distance with weights in linking units, the type of linking unit also being included in this calculation.

FIGS. 7a and 7b show a simple example of such a use of a distance measure in accordance with the third exemplary embodiment of the present invention. According to the fractal network shown in FIG. 7a, the task is to be solved, the neighborhood circle of acquaintances to the semantic unit labeled "Paul"

6 to determine. This is achieved by running only via link units 7 of the type "is friends with", it being assumed here that the weighting of the link units is used as a measure of friendship

7 of the type "is friends with" is specified and friends of friends also count as friends.

The weighting of the linking units 7 of the type "is friends with" can be converted into a distance, for example, by means of a logarithm function. For example, the distance between the semantic unit 6 labeled "Paul" and the semantic unit 6 labeled "Peter" is:

d (Paul, Peter) = -log (0.8) = 0.10

If a limit value for a maximum distance of 0.2 is now established in the semantic Janus unit 6 denoted by “ascertain circle of acquaintances”, which determines the circle of acquaintances of semantic unit 6 denoted by “Paul”, this results in a circle of acquaintances in this exemplary embodiment the semantic unit 6 labeled "Paul", the semantic unit 6 labeled "Peter" with a distance of 0.1, the semantic unit 6 labeled "Mary" with a distance of 0.07 and the one labeled "Jakob" designated semantic unit 6 with a distance of 0.12. Not included in the circle of acquaintances, however, is the semantic unit 6 labeled "Anne" with a distance of 0.25. The distance between the semantic unit 6 labeled "Paul" and the semantic unit 6 labeled "Jakob" is calculated as follows:

d (Paul, Jakob) = d (Paul, Mary) + d (Mary, Jakob) = -log (0.85) - log (0.9) = -log (0.85 * 0.9) = 0, 12th

The above calculation applies analogously to the distance from the semantic unit 6 labeled "Paul" to the semantic unit 6 labeled "Anne". More specifically, the respective weights of the linking units 7 of the type "is friends with" are used to determine the distance. multiplied. The circle of acquaintances can change here without it being necessary to change the method for calculating the circle of acquaintances.

If a semantic unit 6 designated "Paul's circle of acquaintances" is to be formed, which can be returned to the fractal network as a result of a query, for example, then, as shown in FIG. 7b, it must be referred to as "Paul's circle of friends""designated semantic unit 6 are generated by the semantic Janus unit 6 designated" determine circle of acquaintances "and are linked to the corresponding semantic units 6 designated by name. It should be noted here that the semantic units 6 designated by name, which are included in the circle of acquaintances, that is to say, according to this exemplary embodiment, the semantic units 6 labeled "Paul", "Mary" and "Jakob", automatically with logic units 1 of the VA / VS, more precisely of the type "contains / is part of", with which the semantic unit 6 referred to as "Paul's circle of acquaintances" is linked, as is shown by dashed lines in FIG. 7b. As described above, according to this exemplary embodiment, a distance function is used to indicate the distance between two semantic units. Although in this exemplary embodiment a certain mathematical function, that is to say the previously mentioned logarithmic function, has been used to determine the distance from the weight of the linking units, it should be noted here that other suitable mathematical functions of a variable parameter G as the distance function can be defined, this parameter G being present in each linking unit and expressing the strength of the linking of respective semantic units.

A fourth embodiment of the present invention will now be described.

In order to be able to expand existing knowledge in a fractal network, there is a need to link new input data, preferably automatically, with existing knowledge. For this reason, the input data must be in the form of semantic units, that is, semantic input units must exist. These must also have an identifier that distinguishes them from the semantic units of the knowledge already available in the fractal network. Linking units of the type VR / VS or VR / VH are generated with an iterative classification or identification process between the semantic input units and the assigned semantic units of knowledge. Classification / identification means that the information content of each input data is related to one or more corresponding semantic units of knowledge. The weighting of the relation is a measure of the belonging of the input unit to the corresponding semantic unit of knowledge. FIGS. 8a and 8b show a classification / identification process of a sentence in a semantic network according to the fourth exemplary embodiment of the present invention. 8a shows an initial situation and FIG. 8b shows a result situation.

The sentence "The key is in the lock" serves as an example, the meaning of which cannot be understood without background knowledge, since "lock" can be a locking mechanism on the one hand and a building on the other.

The task of the semantic Janus unit 6 shown in FIG. 8a labeled "Classification janus" is now to correctly link the semantic unit 6 labeled "Castle" on the left side of this figure with the world knowledge available in the fractal network. This is done in that, for example, by means of a syntactical preliminary analysis, it is recognized that the semantic units 6 labeled "key" and "lock" on the left side of FIG. 8a are connected to one another via the semantic unit 6 labeled "insert" . In the already existing fractal world knowledge, on the other hand, there is a semantic unit 6 labeled "key" on the right-hand side of FIG. 8a by a relation of the type VR / VH (not described in more detail) with the semantic unit 6 labeled "lock" on the right-hand side 8a, which represents a special locking mechanism. Furthermore, this semantic unit 6 labeled "key" on the right-hand side of FIG. 8a is not connected to the semantic unit 6 labeled "lock" on the far right in FIG. 8a, which represents a special building. If a neighborhood analysis of the semantic units 6 labeled "key" and "lock" and of their linking units in world knowledge is now carried out by the semantic unit 6 labeled "classification jus", it follows that the semantic input unit labeled "lock" 6 on the left in Fig. 8a is classified as a semantic unit "lock" which is a special locking mechanism. Correspondingly, from the neighborhood analysis, which is often also referred to as the context, the semantic unit 6 labeled "insert" is a special case of the relationship 2 between the semantic units 6 labeled "key" and "lock", which is not further defined in the world knowledge available in the fractal network classified. This clearly shows the advantages of the semantic unit 6 designated "classification jus". Not only can the semantic unit 6 labeled "lock" on the left in Fig. 8a be properly classified, but it can also be learned that "plugging" is a possible relationship between those labeled "key" and "lock" is semantic units 6, as is shown by the dashed lines in FIG. 8b, which represents the result situation. As can also be seen from this figure, the new knowledge learned can thus also be incorporated into the knowledge available in the fractal network.

In summary, it can be said that semantic units and / or parts of the fractal network can be classified. This classification is carried out in such a way that the measure is determined which indicates how well the relevant semantic units or the fractal subnetwork fits the current position, and / or the position to which the relevant semantic see units or the fractal subnet fits particularly well. The semantic units preferably contain a marking which indicates whether it is a new input unit or an already existing semantic unit, input units possibly being in the form of a fractal subnet and / or possibly not yet being connected to the fractal network via link units . A new semantic unit or a new subnetwork is included in the fractal network, taking into account the classification. These new semantic units can be linked to a start Janus unit. Furthermore, there is also the possibility of imposing restrictions on the semantic units and / or linking units with regard to those types of units with which they can be linked. Although not mentioned above, one or more input / output devices can also be provided, by means of which the fractal network or a part thereof can be input or output.

A fifth embodiment of the present invention will now be described.

It is common for a semantic unit to create an instance that is a special case of that semantic unit. In this case, the semantic unit can be called the parent and the special instance the child. A child should inherit part of the neighborhood of its parent. A fractal network dealing with this case is shown in FIGS. 9a to 9c. It is useful here if a semantic Janus unit 6, which is referred to as "inheritance janus" in FIGS. 9a to 9c and is connected to the parent, carries out the generation and inheritance process. As shown in Fig. 9c, the in- Formation contents of the newly created semantic units are overwritten with information contents that come from input data or other sources.

More specifically, the semantic unit 6 shown in FIGS. 9a to 9c, designated "inheritance janus", uses the following method, for example.

The "inheritance janus" selects a neighborhood around the parent with whom it is connected. A neighborhood can be defined in a variety of ways, for example, that only via link units of the type VA / VS (+), "is described by", and "has Janus / function of", and that only direct Neighbors can be chosen. In the specific application example, the neighborhood of the "person" is defined by the fact that it is only allowed to walk over link units of the type "is described by", i.e. that "eye color" is in the selected neighborhood of the "person", but "living being" is not in the selected neighborhood of the "person" (see Fig. 9a). However, it should be noted that other neighborhoods suitable for the respective application can also be defined.

Subsequently, a semantic unit "child"("newperson" in FIG. 9b) is generated, which is a special instance of the semantic unit "parent"("person" in FIG. 9b). The "child" is linked to the "parent" via a link 3 of the type VR / VS (+). Then children are also generated for all semantic units from the selected neighborhood. These children are also linked to their respective parents via VR / VS (+) links. In the application example, the child "eye color of the new person" is created and with the semantic unit "eye color" linked (see Fig. 9b). Finally, all children are linked to each other according to the link between their parents. In the application example, the children "new person" and "eye color of the new person" are linked to one another by the linking unit 2b (see FIG. 9b).

Finally, the information content of the children can be overwritten by information content from input objects or other sources. In the application example, the child "new person" is overwritten by "Mr. Otto Maier" and the child "eye color of the new person" by "green" (see FIG. 9c).

In general, it can be said that the invention explained in more detail above with the aid of illustrative exemplary embodiments offers, for example, special advantages in distributed computer systems (such as, for example, networks, INTRANET, INTERNET, etc.), the information and linking objects being provided via a large number of computer systems (processors). and storage systems can be distributed. This makes it possible, for example, for many users (worldwide) to access such a low-order fractal network or to set it up and use it. Typical application examples for this are (multimedia) document management systems, geographical information systems with heterogeneous structured data and meta data, ie data describing the content and structure of data blocks, as well as project management systems for structuring and monitoring of business processes.

Furthermore, the fractal network according to the invention described above is not only suitable for processing, for example, voice data, image data or network structures, but also for treating so-called chaotic ones Systems that describe technical, cultural, economic or ecological relationships, for example. The complex structures can also be both static and dynamic, and the analysis and / or processing of the complex structures can in particular include describing, searching, changing and / or simulating.

With regard to further features and advantages of the present invention, reference is expressly made to the disclosure of the drawing.

Claims

Expectations

1. Fractal network for treating complex structures, the fractal network consisting of a large number of units, characterized in that the fractal network is both

semantic units, each with information content, as well

Contains linking units that describe a relationship content that links two semantic units in such a way that the mutual relationship of the two linked semantic units is determined by the relationship content.

2. Fractal network according to claim 1, characterized in that the linking units are a special form of semantic units, which can have information content and relationship content.

3. Fractal network according to claim 1 or 2, characterized in that the information content described by a semantic unit represents an identification and / or a listing of those linking units that connect this semantic unit with other semantic units.

4. Fractal network according to claim 3, characterized in that the marking described by the information content is a name and / or a serial number.

5. Fractal network according to claim 3, characterized in that the listing described by the information content is available in a structured form.

6. Fractal network according to one of claims 1 to 5, characterized in that the relationship content described by a linking unit contains, in addition to the information content, a link identifier which contains the respective identification of the semantic units linked by it, one or two directions in relation to this linked semantic units and / or weightings G which describes one or two directions.

7. Fractal network according to one of claims 1 to 6, characterized in that likewise one or more linking units in turn - via in each case one or more linking units with one or more semantic units and / or one or more linking units in turn via one or more linking units with one or several linking units can be linked.

8. Fractal network according to one of claims 1 to 7, characterized in that the relationship content of a linking unit may contain information about the respective type of linking of the related semantic units.

Fractal network according to Claim 8, characterized in that the type of link described by a link unit may additionally contain information about a relation VR, ie, about a comparison of the respectively linked units, and / or about an exchange relationship VA, ie, about a one or bilateral interaction of the linked units.

10. Fractal network according to claim 8 or 9, characterized in that the link type described by a link unit additionally contains information as to whether the link type is a scale change VS or no scale change VH.

11. Fractal network according to one of claims 8 to 10, characterized in that the relationship content of a linking unit contains information about the respective type of linking consisting of the pairings VS / VR, VS / VA, VH / VR or VH / VA.

12. Fractal network according to claim 10 or 11, characterized in that the scaling information VS serves to describe the type of relationship to a larger, i.e. higher-level, or to a smaller, i.e., lower-level semantic unit.

13. Fractal network according to claim 10 or 11, characterized in that the scaling information VS serves to describe the type of relationship to a more general or to a more specific semantic unit.

14. Fractal network according to one of claims 1 to 13, characterized by a distance function which specifies the semantic distance between two semantic units.

15. Fractal network according to claim 6 and 14, characterized in that b indicates that the distance function is determined via a suitable mathematical function of a changeable parameter G, which can be present in several linking units and expresses the strength of the mutual link.

15

16. Fractal network according to one of claims 1 to 15, characterized in that the information content of the semantic units and / or linking units

20 in addition to or instead of possibly static data also contain algorithms and / or functions and / or mathematical formulas.

25 17. Fractal network according to one of claims 1 to 16, characterized in that the information contents of at least some of the semantic units represent attributes that describe other semantic units or linking units in more detail.

30

18. Fractal network according to claim 17, characterized in that the network further contains special linking units which serve to link 35 semantic units representing attributes with those semantic units and / or to create linking units to which these attributes are assigned.

19. Fractal network according to claim 17 or 18, characterized in that the attributes optionally contain values that elements from a set, a range, a list or another ordered or unordered

Structure are.

20. Fractal network according to claim 19, characterized in that the ordered or unordered structure representing the respective attribute by numbers, calendar data, audio data, video data, text data, tables, image data, geometry data, fuzzy logic quantities or bundled data or a combination of these is formed.

21. Fractal network according to one of claims 1 to 20, characterized in that the network additionally contains special semantic units, Janus units, which are able to perform certain operations on other semantic units.

22. Fractal network according to claim 21, characterized in that each Janus unit is linked via one or more linking units to one or more other semantic units, the functionality of the Janus unit being restricted in such a way that it is only able to to carry out the specific operations on those semantic units which were linked in a predetermined neighborhood area of this one or more semantic units.

23. Fractal network according to claim 21, characterized in that a Janus unit is possibly linked to one or more other Janus units via one or more linking units.

24. Fractal network according to one of claims 21 to 23, characterized in that a Janus unit is capable of performing one or more of the following operations: generating new semantic units; the bundling of existing semantic units into a single, possibly new semantic unit; changing and / or deleting existing semantic units; comparing existing semantic units; capturing and changing the values of attributes; executing an algorithm and / or calculating a function; capturing and / or changing algorithms; the detection of a Janus or part of a Janus.

25. Fractal network according to one of claims 1 to 24, characterized in that semantic units and / or parts of the fractal network can be classified.

26. Fractal network according to claim 25, characterized in that the classification is carried out by determining the fit measure which indicates how well the semantic units in question or the fractal subnet fits to a given location, and / or by Determination of those points in the fractal network at which the semantic units or the fractal subnetwork in question fit particularly well, the respective fit dimensions being able to be specified.

27. Fractal network according to one of claims 1 to 26, characterized in that the semantic units contain a marking which indicates whether it is a new input unit or an already existing unit, input units possibly being a fractal subnet and / or possibly not yet connected to the network via link units.

28. Fractal network according to claim 26 and 27, characterized in that the inclusion of a new unit or a new subnet in the fractal network takes into account the classification.

29. Fractal network according to claim 27 or 28, characterized in that new semantic units can be linked to a start Janus unit.

30. Fractal network according to one of claims 1 to 29, characterized in that the semantic units and / or linking units can be placed restrictions on those types of units with which they can be linked.

31. Fractal network according to one of claims 1 to 30, characterized by one or more input / output pre- directions for inputting or outputting the fractal network or a part thereof.