EP1301872A2

EP1301872A2 - Method for comparing search profiles

Info

Publication number: EP1301872A2
Application number: EP01956301A
Authority: EP
Inventors: Daniel Veit
Original assignee: Siemens AG
Current assignee: Siemens AG
Priority date: 2000-07-17
Filing date: 2001-06-29
Publication date: 2003-04-16
Also published as: WO2002006974A3; CN1304991C; CN1455902A; DE10034694B4; US7831602B2; JP4868484B2; JP2004515837A; WO2002006974A2; US20040030680A1; DE10034694A1

Abstract

The invention relates to a method for automatically comparing search profiles. According to said method, a search profile is compared with one or several offered profiles that are stored in a database. The individual profiles are subdivided into data fields, in which the information to be compared is saved. Different types of data fields are provided within each profile and the individual profiles have the same structure. During the comparison of a search profile with an offered profile, at least two different types of data fields comprising different comparison functions are compared. A final comparison value is calculated from the interim comparison values that are obtained in this way. According to the inventive method, different comparison functions are thus used for the different types of data fields. This allows a substantially more individual comparison of information than can be obtained using conventional methods.

Description

ω ω Ni t H> h- ¹ cn o (Jl o Cn OC i

0 rt ω P o fD 0 P>

0 3 fD rt cn CO co H-

• o O

CΛ o 0 0 o P>

Hl n cn H rt EU fD 0 s: ^ O P <!

P 0 P> H-

H j t P H

Φ cn rt O

CD H- 3 t → t- ¹ Ω 3 sQ 0 fD

Φ * _* 0

0 0 rt rt s rt ω cn fD ^

CΛ ι-i

OP ^* OC |

0 r- ¹ fD OO

0 rt fD 0 rr cn! -R K H

0 ^J Ω P> 0

Φ 0 ^* , ~. he P ⁾ w H- p

0 0 cn Ht rt • rt o

Φ fD • <p

H K ^

0 O H- CΛ fD ^• ö PH rt fD U3

H <£> Q Ω £

P υs O

O er ^ o ≥! s H- he σ

C_j l- ^J P

1 H- P! Λ

^ rt fD

C | • M Ω

1 0 O ι-3 P Hl H> 3 o

1 o H • _^

<£> 0 P

CO sQ. P> (Λ

1 rt Ό t ffi H- fD

M D O Ω

* _* rt 0 H- fD P ⁾

HH CΛ iT O><fD ιQ cn H fD 1 ω 1 1

ω to tv> H ¹ π O Cπ O Cπ O cπ

O CO ro MP ¹ P ¹

Cπ o Cπ O Cπ o cπ

3 P ιP g NP ¹ cn P ö P ιq rt 2 0

P- P> Φ p 0 Φ φ 3 Cd Ω Hi

P- Φ P ^J P- σ? D rt

P- Φ 0 Ω α rt 0 <! <53 n

P- 0 Φ 0 P ¹ Φ Φ P- 0 P- Φ 0 φ Φ Φ P Φ rr Φ φ cn Hi p cn rt P- Φ p ^j Φ Ω P 0- rt H Ω φ 0 H Ω Hi Φ P " HH P- rt H

P- P P- l_l. P Φ P ^J ιq PP P- P 0 ^J P 0 sQ Hi rt

0 N Ω N cn P Φ φ 3 Φ P Φ Hi 3 Cd Ω <! H rt H - P Φ P P

0 0- 2 cn P. H 2 P- P- ιq 0 PP Hi PHPOP 3 Φ P- Φ ^* H 0 P rt Φ Φ Φ Φ cn Φ 0 P Φ 3 Hi P 0 Φ rt Hi 0 cn Λ r P P - P ü ö Φ P- H t cn ιp

Φ P- P P- P C r P- P. φ P- ι 0 H 0 P- P- H) 1 s: 3 r? Φ Ω Φ P- ιq Φ

H 0 rt rt Φ P ¹ φ 3 P: rt P Hi Ω cn P Ξ H Φ P- p: H l_l. tf 0 P φ cn er

P P. Φ P cn cn cn er Ξ P ^J Ω α φ 0 P P- φ O g P φ Φ s: O

Ω φ φ P 2 N rt <rt Φ P 3 0 ^J 0 0 0 Hi rt H 0 3 H 2 P cn Φ rt y cn cn Φ 2 CΛ 3 Φ φ • Φ P φ P- Φ cx 0 Φ Hi Φ M rt Φ P- P- n Φ

P- rt CΛ H φ 0 φ 0 H P s: P- rr 0 ι P- H H P 0 P- P- P- 0 p- P rt P

Φ φ 0 Ω P- Ω - cn Hi P- P ι Cd 3 Φ H Ω ι P "0 φ Φ Φ

PNH Ω P "t HPS ι-i Φ Φ P <P ¹ Φ φ P P- y Φ P- rt P

2 cn 0 ^J Φ α i Φ Φ P ^* Φ HP Φ Φ P- 2. rt p: Φ tr ¹ Φ sQ 3 cn

P- Φ rt Xi HPHH P- ι-i HN 3 H Φ CΛ Φ P- ⁴ H φ 0 g Φ Cd φ P 2 P)

Ω P- Φ H rt P Φ 0 Φ f? P- s sQ ιP 0 HPPM 0 0 Φ y ιq φ P = r PP er Φ H) φ P 0 Φ φ φ P ¹ rt Ω rt PP "P- P- 2 H φ H Hi φ 0 Hi Φ P P- ü 3 PHP "P P- Φ 0 ^J 0 φ P- Ω ιq Φ rt Cd φ Φ PP

P 0 CΛ P- P- Hi P ¹ PN Ω rt cn 0 P- P- P 0 P Ω tf H P- H li HP Φ H rt 0 P ¹ Φ cn rt N rj Φ ^J Ω φ Ω P- 0 ιq ty Φ P- Hi 1 ιq et Φ PP 3

<Φ Ω <! Φ 2 p P 2, 0 'φ Hi cn P Φ 0 Ω P g φ Φ P • P φ H 0- Φ Φ 0 P φ <! P- Φ Cd Φ H <! 0 0 ^J P ¹ P er 0 cn rt rt ι-i cn xi V 3 Φ 0 Hi P- <! ιq Φ HP φ 0 PP Hi <! tsi H rt 0 Φ Φ P- ιq Ω H rt H Φ rt φ φ H £ 0 (H ιq φ φ s: Φ Ω P- <! XI P

H y Ω P l_l. P ¹ Φ H 3 Hi ω φ 0 Hl Φ ιq φ H P- Φ PJ ¹ cn PP

P- Hi S P φ H φ 3 CL 0 vQ p: P Φ H 0 ιq 0 P H Hi rt 0 cn 0 H 0 cn

P- φ P- cn 2 P- Φ Φ So "0 'rt α P 0 0 PH 0 NP φ PP ¹ Φ

Ω P <Φ rt cn H CΛ Φ HH 0 er N ιq Hl P ⁴ PH t rt rt P-

P 'H cn ι P- 0 P- α Φ 0 Cd Φ 0: φ Φ H ö' Φ p Φ N 0 Φ P- P cn P- P ¹ P ¹ φ NP Ω Ω φ 0: ιq φ ιP cn 0 Φ p P- H 0 Hi 1 ^*

Hl Ω 3 φ P- p- 2 0 0- P ^J PP ⁴ s: N P- P Φ 0 <P <£ Φ

0 t P- P- iP P φ Hi XI Φ N P 3 Φ 0 ιq 0: 0 φ Φ P Φ H Φ P P-

0 Φ rt Ω Φ φ P- Ξ HP 0 j P ¹ P: H ιq e H

Φr g N er H l-i P-

P- 3 rt rt ιr 0 P ^J P 3 rt Φ PP p P- ιq rt HP <ιq er PP Hi Ω Φ rt P. Φ P- Hi Φ P Ω P "Φ 0 σ H φ p Φ PP ¹ Φ Φ rt P f_T ^J P

P- PJ Φ P- α 0 0 cn P- P- er <y P- 0 0 vq 0 Ξ H φ P- 3 P- p

P 3 3 cn P 0 φ r-> P p- φ Ω α φ φ φ <p: er P- rt P H P er

0 rt rt rt CΛ P φ p: H P- 0- <! Φ P w P- φ Φ Ω Φ O 0 φ Φ

Φ Φ N • φ Φ 0 cn P ιq cn P H p P H rt cn p- 0 Φ φ P φ P-

0 P 2 PH Ω 0 2 ιq P ¹ rt P- 0 Φ er Φ φ ιq P- rt Φ cn 0 li

Φ 2 Hi cn '0 P φ P- Φ cn P- Φ 0 P P "ι P- P ι P N l_l.

<<! P- Φ φ Ω Xi P er H ιq P- 0 rt H P P- φ rt g l-i 3 P H er φ

Φ P rt HP ¹ P ^* H Φ cn Φ Ω 0 0 rt m P- p 0 0 rt 0 Φ

H P-

P Φ P- P- P- P P- rt P y P Hi <! Φ rt 0 0 Φ cn; _* r φ

P- Φ φ m Φ Ω s: ιP P φ φ Hi P- P φ rt P- P- Φ Φ ö P α P- H Hl P P 3

P. P- φ PT) 0 3 P- HH Φ P ¹ φ H cn ιq 0: 0 0

P- P CΛ Φ P ¹ P "P- P p- Φ Hi 0 'rt CΛ

Ω rt 0 er cn P- cn σ φ CΛ rt 3 s: H P P- CΛ Cd p: Φ

<rt P P Φ P Φ 0 φ rt he H P rt Ω

0 'Φ Ω Φ Ω P 0 Φ o P rt P ⁴ H 0 Φ 0 0 Ω 2 NP Φ 3 _">

Φ P p- P- φ P ^J l_l- rt CΛ Ω PP ι Φ P- H P. H cn 0 ^J Φ P ¹ P- 0 PH

0 Hl xi 3 H Φ Φ φ 0 P ^J rt Φ HP Φ φ Ω α WPH P- cn rt P-

• Φ ι-i cn P 0 Ω Ό P CΛ n 0 P 0 y <P 0 rt Ω t P- P rt r-> P <rt Hi y HP 0 P- Cd Φ • φ Cd 0 Hl Φ m 0 ^J

P- φ Xi P cn Ω 3 P: cn cn rt

Hi Φ Φ 0 φ 0 N? H φ 0 H 0 P- 0 Ω

Φ P- HP r->-> H Hi Xi P ^J ι 2 g O: 1 2 P 2 rt 0 cn cn

H P "1% cn P P P- H • ö φ φ N p 0 φ ι <! P- 1 Φ Φ P- φ

P 0 φ Hi P "0: H 3 H 0 0 H φ Ω φ 0

P H P- cn 1 P 1 P 0 Φ 1

CL 1 1 0 0

^p a- Distributive embodiments of the invention are specified in the subclaims.

In the method according to the invention for the automatic comparison and evaluation of information, a search profile given to a user is compared with a reference profile stored in a database. The profiles are each divided into a certain number of data fields in which the information to be compared is stored. Each profile has at least two different types of data fields. The profiles to be compared each include the same types of data fields.

When comparing a search profile with an offer profile, the at least two different types of data fields are compared with different comparison functions and the respective comparison is evaluated with a preliminary comparison value. A final comparison value is calculated from the preliminary comparison values.

The method according to the invention thus compares profiles structured into individual data fields. With the invention, different comparison functions are used for the different types of data fields, with which preliminary comparison values are calculated. As a result, the contents of the individual data fields can be compared and evaluated on a type-specific basis. A final comparison value is calculated from the preliminary comparison values.

The invention thus compares the individual data fields in a type-specific manner and the results of the individual comparisons, that is to say the preliminary comparison values, are combined to form a final comparison value.

With the method according to the invention, the comparison of the individual data fields is carried out in a type-specific manner, which results in much more realistic results than in the previously known. co CO IV)) P ¹ P ¹

Cπ O Cπ o cπ o Cπ

second vector plus the distance between the second and third vector.

The method according to the invention can very advantageously be integrated into an agent system. This agent system comprises at least three types of agents, namely a search agent, an offer agent and a comparison agent which, when requested by the search agent, compares and evaluates the profiles stored in the search agents and offer agents. The agent system is preferably an open agent system, which means that further agents, in particular offer agents, can be added. The agents are preferably mobile agents, that is to say that they can be active in different places in a computer network and can change their location in the computer network.

The invention is explained in more detail below on the basis of the exemplary embodiments illustrated in the drawings. In which show:

1 is a table showing different basic data fields,

2 is a profile description in tabular form,

3 shows a profile structure in a block diagram,

4 shows the method for automatically comparing and evaluating information in a flow chart,

5a two free texts to be compared,

5b shows two data records which are derived from the free texts shown in FIG. 4a,

5c evaluation results for the individual words of the data records in tabular form, 6 shows an example of an offer description for a cooperation exchange,

7 shows an agent system in a block diagram, and

FIG. 8 shows a network for connecting computers on which the agent system from FIG. 6 is installed in a block diagram.

In the inventive method for automatically comparing and evaluating information, a search profile is compared with an offer profile stored in a database. 2 shows a profile description of an exemplary embodiment of the invention. This profile description comprises eight data fields, of which the name of the respective data field is given in the left column in FIG. 2, the variable character of the data field in the middle column and a brief description of the data field in the right column.

Basically, an automatic comparison process differentiates between an offer profile and a search profile. The profile description of the offer profile and the search profile match in structure. They differ only in the content of their profile type data field, in which the information on whether it is an offer profile or a search profile is stored. The data field profile type t is a boolean data field, the content of which can be either 0 or 1. The other data fields are the title, the keywords, the exact description, the costs, dates, duration and the participants. The title data field contains a brief description of the service offered or sought in the form of a so-called verb-noun expression. The use of such verb-noun expressions is known from VS Subrahmanian (editor), Piero Bonatti, Jürgen Dix, Thomas Eiter, “Heterogeneous Active Agents *, Cit Press; ISBN: 0262194368 known. The Keywords field contains a set of keywords. For the purposes of the present description, a sentence is an unordered collection of elements of the same type, such as words, real numbers, integers or the like. The variable of a sentence is shown between two curly braces.

The exact description data field includes a free text that describes the service offered or sought.

The data field Cost contains information about the minimum or maximum expected costs. The cost data field therefore represents an interval.

In the Duration data field, the length of time that is required to carry out the service offered is specified.

The Participants data field contains a list of the names of the participants who offer or are to offer the service. A list is shown in the form of a superscript plus sign. The expression in brackets [1: 2] means that each list element is composed of two individual elements, namely the first and last name. The data field% β [1: 2] ⁺ and the data field (τi) are complex variables which are explained in more detail below.

FIG. 3 shows the structure of the profile description from FIG. 2. The profile description is divided into three levels (level 0, level 1 and level 2). Level 2 is the highest level in which the data fields shown in FIG. 2 are arranged. The complex data fields τ and τ ₈ [1: 2] ⁺ each refer to further data fields, which are represented by corresponding variables in the level below. Thus, several data fields τ _x are arranged in the level, in each of which a keyword is stored. The complex variable τi thus relates to the list of keywords stored in level 1. The complex data field x ₈ [l: 2] ⁺ the participant refers to a list of further data fields. The elements of this list are field arrangements [1: 2], each comprising two names, the first and last name. Basically, a field arrangement comprises a certain number of elements of the same type. The field arrangements x ₈ [1: 2] thus refer to further data fields, which are arranged in level 0 and each have a word entry, namely the first name or the last name. Two such data fields x ₈ are combined to form such a field arrangement.

The data fields that refer to other data fields in a subordinate level are called complex data fields. The remaining data fields are basic data fields.

The information of the respective profile is stored in the basic data fields. Using the complex data fields, several basic data fields in the form of records, lists, field arrangements or registers (record) are combined into one

Field arrangement projected at the highest level. Similar to field arrangements, registers are formed from successive elements of a predetermined number, which in turn can consist of different types.

Due to the tree structure described above, using the complex data fields branching from a higher level to a lower level, only one data field is provided in the top level (here: level 2) for each conceptual unit.

1 shows a list of the basic data fields. Column 1 shows the variable names of the basic data fields Xi to x ₈ . The middle column contains the names of the corresponding basic data fields and the right column contains a brief description of the content.

x ₈ [l: 2] the value 0 entered as a preliminary comparison value. If this field arrangement (= first and last name) could not be found, the value 1 is entered in the corresponding data field in level 1. After completion of step S2, all data fields x ₈ [1: 2] are provided with a preliminary comparison value.

In step S3, the preliminary comparison values assigned to the names are evaluated. This is usually done by weighted averaging. Since the elements to be compared are each of the same type, they are equivalent and are therefore all weighted with 1. An average of the values entered in the complex data fields x ₈ [1: 2] is thus formed. This mean value is a preliminary comparison value of the second order, which is entered in level 2 in the complex data field of the list of names t ₈ [1: 2] ⁺ .

In the subsequent step S4, the data fields Xi of the search profile containing the keywords are compared with the corresponding data fields of the offer profile. The comparison function for comparing the keywords is designed such that each keyword of the search profile is compared with each keyword of the offer profile and if a keyword of the search profile is not among the

Keywords of the offer profile is included, the value 1 is saved. Otherwise the value 1 is saved. As a preliminary comparison value, the mean value of these values is calculated and entered in the data field of the list of key words {Xi}.

Method steps S3 and S4 are carried out in level 1.

In the subsequent method step S5, the contents of the data fields title x ₂ , exact description x ₃ , costs x ₅ , dates x ₆ , and duration x ₇ are compared with one another. The comparison function for comparing titles x ₂ is a common comparison function for comparing verb-noun expressions.

The comparison function for comparing the data field exact description x ₃ is a comparison function for comparing free texts. 5a shows two examples of free texts di, d ₂ , each of which contains a text in the English language u. These free texts are first transformed into data records DSi and DS ₂ , in which all words from the free texts that are not stop words are adopted. Stop words are words that have little information content. There are lists with the usual stop words. In the present case, the following words are assessed as stop words:

this, is, in, a, the, and, off, can, be, are, with, we, for, to, an, able, wich, our, not, shout, already, make.

In the data records DSi and DS ₂ , their frequency in the corresponding free texts is given behind the individual words. The individual words are sorted alphabetically in the data records.

To compare the free texts, the words of the data records must have weighting factors. To calculate the weighting factors, a so-called inverse document frequency idfj is first calculated, which is defined as follows:

',: = l ° gj7

where N is the total number of documents and df _{j is} the number of documents that contain the word j. In the following exemplary embodiment, each free text represents a document. In addition to the two free texts shown in FIG. 5a, there are a total of two documents another 18 free texts of another 18 offer profiles. The total number of documents N is therefore 20.

With the inverse document frequency, those words that occur very frequently are weighted with a value going towards 0 and the words that occur only in a few documents are weighted with a value going towards 1. As a result, rare words are weighted more heavily than common words in the inverse document frequency. Rare words usually have a higher information content than common words.

In addition to the inverse document frequency, the frequency tfι, j of the words j in the documents i is also taken into account. Thus, the weighting factor ι, j is the product of the frequency tfι, j and the inverse document frequency idfj

For the words of the data records shown in FIG. 5b, their inverse document frequency is dfj and the weighting factors w _x , j and w _{2; j are listed} in the table from FIG. 5c.

The weighting factors Wχ, j and w _2fj each form elements of document vectors DVi and DV ₂ .

When comparing two free texts, the distance between the corresponding document vectors DV _L and DV _{2 is} calculated. According to the invention, the distance between the two vectors is calculated as the Euclidean distance according to the following formula:

The Euclidean norm fulfills all requirements for a metric distance: - The distance between two identical vectors is 0. The distance from a first vector to a second vector is equal to the distance from the second vector to the first vector. That means the distance calculation is symmetrical. - The distance from a first vector to a third

Vector is less than the sum of the distances from the first vector to a second vector and from the second vector to the third vector.

Only if the distance calculation fulfills this requirement can it be ensured that a reasonable distance is always determined.

Instead of calculating the distance between the two document vectors using a Euclidean distance, it is also possible, as is carried out in conventional comparison methods, to calculate the distances between the two vectors using the cosine between the two vectors.

The comparison function for comparing the data fields containing the costs is a comparison function for comparing intervals. The distance between two intervals ii, i ₂ , which are given by real numbers iι = [lι, ri] and i ₂ = [l ₂ , r ₂ ], are calculated using the following formula:

Known comparison functions are used to calculate the distance between the data fields, dates and duration.

In the present exemplary embodiment, no numbers are compared, which is why no corresponding comparison function is used for the comparison. Such a comparison function can be implemented very simply, for example, by determining the absolute value of the difference between the numbers to be compared.

The preliminary comparison values determined when comparing the data fields x ₂ , x ₃ , x ₅ , Xβ and x are stored. This concludes step S5.

In step S6, the individual preliminary comparison values for the data fields Xi to x ₈ of level 2 are used to calculate a final comparison value. A weighted average is calculated here, the individual data fields being weighted differently depending on their importance. The result of this weighted averaging is a distance value that specifies the distance between the two profiles to be compared, the search profile and the offer profile.

Since, as a rule, a similarity value and no distance value is desired, the reciprocal of the distance value is formed (step S7). This similarity value represents the final comparison value. This comparison value is output in step S8. The method is ended in step S9.

The final comparison value can be used to sort the corresponding offer profile in a list of offer profiles according to the calculated similarity to the search profile.

If the user specifies when initiating a search process that he wants the most similar offer profiles, then the method according to the invention described above is carried out for each offer profile, the individual offer profiles are sorted with decreasing similarity with respect to the search profile and the most similar offer profiles are output to the user as a result.

The method according to the invention can be implemented as a computer program for the automatic comparison of profiles. A particularly advantageous implementation of the method according to the invention is in the form of an agent system.

Agents are autonomous, cooperative software units that consist of code and data. They are software units that function independently and do not require constant interaction with the user. There are both stationary and mobile agents.

Mobile agents are e.g. known from US 5,603,031. Mobile agents are programs that can be active in different places on a computer network and can change their location in the computer network.

7 schematically shows the sequence of the method according to the invention using three agents. A comparison agent, a search agent and an offer agent are used here. The comparison agent contains a database in which the offer agents known to him are stored with their respective offer profiles. The offer agents can enter their offer profile in the corresponding database or delete this offer profile if they no longer maintain the corresponding offer.

A search agent who is looking for a specific service turns to a comparison agent and sends a search query to the comparison agent. The search query contains a corresponding search profile. The comparison agent compares this search profile with the data stored in its database. bot profiles and evaluates them according to the procedure described above. It transmits a corresponding search response to the search agent, which contains a list with the names of the relevant offer agents, each offer agent being rated with a comparison value.

The search agent can either forward the search response to its original client or send a request for delivery of the corresponding service to the offer agent to whom the best comparison value is assigned. The service can then be provided by the offer agent to the search agent, who forwards it to his client.

Fig. 1 shows schematically simplified a network in which such an agent system is implemented. The network has a number of computers 1 which are connected to one another via data lines 2. An agent system AG is installed on each computer 1. In the network there are some mobile agents AG-I to AG-IV, which are either arranged on one of the computers 1 or move from one computer to another.

Each agent system has an agent platform that includes utilities that an agent needs in order to be able to be executed on the respective computer 1.

Agents AG-I are offer agents and agents AG-II are search agents. Agent AG-III is a comparison agent. The offer profiles of the offer agents AG-I are stored in the comparison agent AG-III. A search agent AG-II can make a search request to the comparison agent AG-III, which the latter answers with a corresponding search response.

The search agents can then process the search response in the correspondingly predetermined manner and, in particular, can forward it to the user who is operating a computer in the network. O CO>) P ¹ P>

Cπ o Cπ o Cn o Cπ

CL P <er X ⁴ <ö ιq Hd Ω PM P. CL Ω φ cn P φ φ O cn P Φ 0 P- ü P ⁴ 0 0 0 Φ Φ P ⁴ H Ω P li rr li ⁴ P i H Φ O Φ Ω P ⁴ 0 H φ P ⁴ P ⁴ cn

N 2 P Hi Ω P- Hi 0 P ⁴ P ιq Φ P p: P- cn

P- 2 0 PP ⁴ Ω P- P ¹ Cd li CL P ^< φ

P ⁴ Φ cn NP ⁴ P ⁴ P "2 3 rt P- Φ Q φ rt P Hi

P i Φ CL • H CL φ Φ O: Φ cn 2 P- PP ¹ 0:

Φ CL P ⁴ H Cd Φ P- PP ii ιq P rt φ - er 3 P- H

P Φ Φ Φ • P φ P er CL left • XI P Ω

PP P- cn φ ι φ P- P φ rt li PP ⁴ P

N • N 2 φ P Φ 0 Ω cn 0 Φ O Φ P-

0 tr £ P- er P ⁴ Cn P P- Hi Φ Φ ιP σ P P- Φ HN n P- • »N ι P P- P- <!

Φ P- 0 cn Ω CL 0 ii P "s: er • φ P" 0 Φ 0

P Φ Hi cn O CL O CL Φ Cd P. φ Φ H 0

H ω Hi CL p: Hi Φ ö ⁴ P P- • 0 cn. ιq rt

CL φ CL 0: p: φ rt P- rr φ ω H Φ 3 Φ

PPP & 3 HNP ¹ P- cn o 3 Ω P ⁴ D 0 Φ H

Φ H cn Hi P- cn 2 er P: P ⁴ li P- P "P- cn rt P ιq Cd P- P- φ π P- sQ φ rt Ω Ω

Φ Hi r 0 P- Ω 0: H P- H ιq Φ p- cn P- P ⁴ P "

PO φ 0 cn P 0 ⁴ er CL 0 O φ P- P- P Φ cn P-

MP "rt φ cn Φ φ φ Hi H Ω P CL P- Hi φ cn 3 P- Φ P- PH cn P- P ⁴ φ P- 3 3 0 CL rt P Φ HP rt g φ PP ¹ P 0 φ 0 P "

H rt he Cn N P- Φ CL φ 0 P- P ¹ P t * P-

0 P- P- Ω er er 0 cn Φ cn CL £ rt cn rt Ω f O sQ P ⁴ φ Φ P- cn cn i 0 Hi P φ 0 P- P- p- P ⁴ rt P Φ P- P- HP ¹ rt φ P 0: cn P Φ CL OO Φ

0 cn Φ P φ CL P- 0 er rt 0 ⁴ cn cn P- P 0 P

H Φ H CL P- 0 3 φ Φ li Ω P ¹ 3 P Φ

P- P- 0 P "H Ω 0 p CΛ P- li 0 CL P ⁴ Φ φ P ¹ 0 m φ P Hi P- Φ P ⁴ ιq rr rt P cn 0 P- Φ P φ ü P ⁴ P Ω • H φ Ω ι Φ P Cd cn P rt Φ li P ⁴ P φ Λ ⁴ 0 0 P ⁴ cn Φ p- Cd 0 s φ

Φ P- 3 Φ rt HP Λ ⁴ P- Hi Cd P- ^* SO φ ιq P

0 rt PP P- P ⁴ 0 rt πd φ OP cn φ 0 55 Φ

Φ rt <i Φ 0 0 p, CL H cn rt li P φ s: <! πd P P- H er H o 3 P- • rt HP o li PS fc ⁴ P "CL Hi P- cn 0 Φ rt 0 P

O 2 0 P> P- P 3 P- Ω er Hi H 0 0 CL

Hi φ cn Φ φ Ω cn p- ¹ P ⁴ Φ Φ 3 ιP Cd 0 rt ö

P- H Φ P P- P ⁴ rt φ Φ P ¹ P * Φ ιq P

P p- P φ H φ CL ^• Λ 2 s: rt

Φ Φ P <i φ φ H P P ιq Φ P CL φ P Φ Φ

PP ⁴ o cn H Hi Φ 0 CΛ rt ü P ⁴ P- ü Φ HP

<! Φ P ^•> p- li H, rt 3 Φ rt H Hi φ P P-> Φ 0 li cn 3 φ 0 Φ Φ

HP rt ö H P- CL CL φ 0 P- P- P cn 0 N 0 P ¹ ιq P Φ P o rt 0 Φ P-? TP rt φ ιq 0 CL

P ^> PP rt Hi Φ 0 cn 0 rt C CL 'Φ

P- Φ P- li Q 0 P φ H CL <li

Ω 3 PPP ¹ rt cn N - ^> ii P ü φ Φ P- P

P ⁴ P- er Hi • vq 0 φ Cd CL p: H ii φ

Φ rt ιq Φ 3 Φ P- <! cn Φ Cd P ⁴ ιq H 0

P rt Φ P- Cd 3 <! rt Φ H ü P d P ¹ CL 0

• CL er CL rt cn p: Φ φ ü P- Φ Hi P- H Φ 0 rt

P ^> P- Φ E ii li <q cn P P- P- O P- H Φ cn H Φ 1 φ rr P Ω Hi 1 Ω li

1 P cn p- 1 P ⁴ P- 0 ^* 1

1 1

Claims

CO CO> MP ¹ P ¹

Cπ o Cπ O Cπ o cn cn Φ P P CO ιq P Ξ 0 3 CL er CL ιq cn 0 0 0 P Pl M Hi Ω 3 er Hi cn xi CL CL LJ. 3 P> d

P- P- P. P- 0 φ 0 P- P- Φ Φ P "Ω Φ 0 0 φ P. 0 P ⁴ P- Φ φ Ω li P- φ P-. P

Ω P cn P φ cn P. CL rt φ P- H φ P ⁴ 0 rt CL H Hl 0 Φ rt P- P-- P ⁴ O φ P 2 0 rt - Φ cn <J P- rt 3 0 P- P- Φ o <ϊ r P CL P- H) Φ CL <Φ li CL φ Ω P CL 0 N Ω Φ ιP P α li Φ rt CL 2 φ P- α CL P- Φ Φ 0

P CL ü 0 ⁴ Φ er Φ 0 g <g 0 ⁴ P Φ rt P- P 3 H P- m Φ φ P CL P cn rt

0 CL P- 0 Hi cn 0 Φ H rt 3 Φ P Φ P ¹ cn Φ Φ rt P Hl O * <3 ¹ φ P ¹ cn rt 0 cn rt Hi P

Hi Φ φ P 2 Φ P- H Hi 0 P- XI P- φ rt P 0 XI Ω cn P- φ Φ P 0

H li P ⁴ Φ <φ Ll- HP sQ P Ω φ P "πd P P- ⁴ φ Φ N 0 ⁴ Ω ι_ι. 0 li 3 0 P ⁴ cn

3 π li H 0 H, φ cn PP ¹ cn Φ P ⁴ P-rt p; er 0 ti 0 P 2 Φ Φ P ⁴ Φ Hi Φ cn li • d

Φ Cd P, Ω Φ rt H rt 2 Ω φ Φ cn 0 φ Ω o PP Φ φ 3 H Φ ≤ φ Ω P ⁴ Φ ü ¹ er PP Φ φ 0 ⁴ ω P- φ 0 ⁴ ω Hi P Φ 0 < ! <P- cn P Φ P ¹ ü φ P 0: ü Φ Hi P ⁴ er p: rt p- P- rt Ω 0 πd H φ P- P- - P φ o rt <rt P- CL 0 ⁴ φ P - Ω

Φ P P- P Φ 0 φ 0 ⁴ HS ti 0 P ¹ NHP Φ Φ φ ι-3 φ ü 0 N 0 ⁴

H Φ P ¹ PH Hi 2 P- CL N Φ o rt P φ P g ιq P li 0

P- P- ιq P ¹ 2 0 2 Hi Φ er 2 P "ö Hi s cn li Φ Φ g φ

Φ P Φ Ω φ 3 p ιq P ⁴ Ω ιq ii φ P- φ o P- P cn LI. ιq o P- P CΛ P 0 N CL t ö ö Φ 3 0 ⁴ φ CL Ω P- Φ er p- P- φ Φ er P Ω rt 0 P ⁴ 0 gg φ PN <!

P P- φ φ g 0 0 0 ⁴ P- Φ Φ <! P 0 2 cn Φ 0 P ⁴ Φ Ω H CL pp cn Φ rt 2 Φ rt PP ⁴ Φ φ Φ 0 P P- O P. φ * d P- rt φ PP ⁴ Φ Hi P- Φ Φ li

Φ li; cn rt <! 0 li 0 0 φ l_l. P 2 P- φ O P Hi XI P CL P P φ 0 P P- ιq

0 f Φ • d φ P <q rt cn φ Φ P ¹ P- l_l. 3 Φ H Φ cn CL H Hi rt

H) P p, Φ li 2 H CL P "<Φ 2 σ 0 P ¹ cn Ω φ P s: P" o er cn cn φ cn P Φ Φ Φ

Φ Φ 0 p- iQ φ Φ H CΛ φ P 0 Ω P ⁴ 2 r Φ CL Hi Φ φ cn rt 0 P-

Xi P Ω HP ¹ P- H cn 0 P- rt CL 0 ⁴ P- φ Φ P- H Φ P- P- NP rt Φ N CL Ω

PP ¹ Cd 0 ⁴ P Φ Ω ιq Ω Ω P ¹ φ Φ 0 ii P- cn CL ii P "3 2 0 Φ CΛ P ⁴ φ φ er P P- 0 ⁴ 0 ⁴ 0 ⁴ cn P ι_ι. 0 rt Ω Φ P φ N Φ li 0 Φ ii X Φ N) Ω φ P- Xi Hi Φ φ Φ cn P ⁴ P x P- ^* s: CΛ Ω P φ PN 0 ⁴ 3 P- φ H (X φ CL CL P - P Φ • 3 P- Φ rt φ Φ 0 P- P 0 ⁴

Φ cn Φ cn P- Ω CL O P- P "Φ P- 0 Φ P P- φ H Φ P- Ω 0 i Φ

P- P Φ s: rt P ⁴ P "Hi Φ x cn Φ φ P- rt ιq. 0 Ω 0 ⁴ Ω Hi li P-

0 ö Φ cn P- P- φ g & P <! NP ^» P ⁴ α XI 2 PP

Φ P P- H φ Hi Ω P "sQ H πd N er ιq φ 0 g Φ CΛ Φ P li P ⁴ φ Hi Φ li rt rt P- 0 0 ⁴ cn P ¹ PH 0 φ φ CΛ li 0 p P- 0 H rt P P- P- cn

Φ Hi Ω Φ 0 0 Φ Φ P cn er 0 sQ rt P- Ω Ω φ Hi P cn P ¹

0 PP 0 φ Ϊ 0 3 P- er Hi <J rt o Ω Φ PP ⁴ P ⁴ er P P- Φ Φ

0 Hi cn P ⁴ 3 rt P- Ω Φ P- Φ P- rt 0 ⁴ Φ H CL Φ Xi Φ Hl P ¹ Φ 0 H rt φ cn Φ P- ι-3 rt 0 ⁴ cn H s cn XI P- cn φ P li P- Φ cn 2 cn

Φ P ¹ φ P P- <! O * Φ P- ιq p • d li Ω Ω cn OP ¹ rt o rt

HPP 0 P 0 • d Φ 0 rt NP ¹ rt HOP ⁴ 0 ⁴ rt CL Hi L CL 0 er Φ Q Φ ii Φ Φ P- N g φ Φ o Hi Φ P- φ P- Φ Φ P Φ P φ < Φ P "0 0 m rt P- Hi P- P φ N cn 3 HP P. P-

O o Ξ rt 0 p: Φ ^ P- Ω [Jd P- P ¹ CL s: <! CΛ

HHP - • (X 0 <<! 3 • d P 0 ⁴ P "0 P ¹ Φ φ l_J.

P- 3 CL P CL 0

CL sQ er ιq Hi Φ P Φ 0 CL Φ N 3 0 P- H φ p- φ P- Ω

0 φ Φ 0: P- H 0 g P 0 φ 0 P <P- CL Ω cn rt 2 rt cn cn Φ 0 ⁴

Φ cn P- P ¹ ιP vP P cn CL P ⁴ φ rt 0 ⁴ 0 rt O φ Xi r Φ rt Φ P ¹ ü ιP <! rt Φ P "K Cd Φ 0 Φ H P- Φ N cn CΛ ü

Φ P ⁴ N P- 0 P- P Φ o CL 0 ιq Φ φ 0 rt P cn P- 2 0 O

P φ ιq O rt er P P- N ö P ¹ P- 2 Φ P- P Φ Ω Hi

P 3 φ <"Φ o φ 2 Pl P P- 0 Φ <J li CΛ ιq ιq Φ p- P ⁴ P-

Cd P- H Φ φ 0 rt σ Φ 0 rt Ω Φ Φ cn 0 P ¹ φ 3 rt XI er P- PH 0 Hi cn P t> 3 P- Hi Φ P ⁴ 3 r li Ω Ω Φ P Φ H cn

Φ cn P <J ιq Φ ^• d rt 0 OP Φ Φ ιq 0 ⁴ 0 ⁴ p- 0 P o

P rt φ Φ P "s: P" H φ 0 H Hi P p- 0 P ¹ P- XI Ω α Hi 3

Φ cn ü Φ Φ CL OP <J 0 3 Φ 0 φ φ HP ⁴ P rt CΛ P- P- r 1 P- li Φ Hi Hl Φ rt P s: <! P- CL O rt Φ 0 rt er CL Ω PH P- φ H Φ rt α P- Φ P Ω P ¹ Hi P- Φ H Ω Φ

Φ P P- 0 ⁴ Φ 0 P "P" 1 H P- Φ ii P- 0 P ⁴ P- P- cn P 1 0 ⁴

1 ω P cn 0 1 1 O li 1 cn 1. P ¹ rt 1 1 1 1 i cn

draws, the complex data field being a variable into which a complex comparison value is inserted during the comparison, which is calculated from the data fields subordinate to the complex data field.

4. The method according to claim 3, so that the complex data field relates to basic data fields, the information of the profiles being stored in the basic data fields.

5. The method according to claim 4, characterized in that the complex data field is arranged in the top level and that several levels are arranged under the top level, the relationship of the complex data field of the top level to the basic data fields not being in the top level immediately lower level arranged basic data fields is produced via further complex data fields, which are arranged in the levels between the top level and the levels in which the basic data fields are arranged.

6. The method according to any one of claims 2 to 5, so that the final comparison value is calculated with a weighted averaging from the provisional comparison values.

7. The method according to any one of claims 2 to 6, so that the preliminary comparison values each represent an information distance, the value of which is greater, the more the corresponding information differs.

8. The method according to claim 7, characterized in that a final information distance is calculated from the provisional comparison values to calculate the final comparison value and the reciprocal value is calculated from this information distance, which forms the final market value.

9. The method as claimed in one of claims 2 to 8, that the comparison functions compare and evaluate data fields containing either two, each containing a date, a numerical value, a free text, a keyword, an interval, a time or a name.

10. The method as claimed in claim 2, a comparison function for comparing two data fields each containing a free text, by

- the two free texts are broken down into individual words,

- A record is created for each free text, which includes all words of the respective free text that are not a stop word,

- Each word of the data records is weighted according to its relevance within the free text and according to its relevance within the database with a weighting factor (wi), the weighting factors of the two data records being elements of a document vector (DVi, DVj), and

- Calculate a distance between the two document vectors, the distance representing the preliminary comparison value.

11. The method of claim 10, d a d u r c h g e k e n n z e i c h n e t that as a distance a Euclidean distance between the two

Document vectors (DVi, DV _j ) is calculated.

12. The method according to claim 10, characterized in that that the cosine between the two document vectors (DVj _. , DVj) is calculated as the distance.

13. The method according to any one of claims 2 to 12, d a d u r c h g e k e n n z e i c h n e t that a K.O. Criterion is used that monitors a preliminary comparison result for a specific field of the profiles to be compared and sets the final comparison result at a predetermined value of the preliminary comparison result to a predetermined value, regardless of the other comparison results.

14. The method according to any one of claims 2 to 13, so that at least one comparison function comprises a threshold value criterion which sets the comparison result to a predetermined value when the threshold value is exceeded or undershot.

15. The method according to any one of claims 2 to 14, a comparison function for comparing two each one

Number-containing data fields by calculating the absolute value of the difference between the two numbers as a comparison value.

16. The method according to any one of claims 2 to 15, characterized by a comparison function for comparing two data fields each containing an interval, the limits of the two intervals (ii, i ₂ ) being represented by real numbers (iι = [lι, rι] and i ₂ = [l ₂ , r ₂ ]) are given, and the comparison value d (iι, i ₂ ) is calculated according to the following formula:

17. The method according to any one of claims 2 to 16, g e k e n n z e i c h n e t d u r c h a comparison function for comparing two data fields each containing a name by the provisional comparison value being set to zero if the names to be compared match or are set equal to infinity if the names differ.

A network for connecting a plurality of computers, an agent system being installed in the displayed computers and the network having a number of agents, characterized by a comparison agent which is designed to carry out the method according to one of claims 2 to 17 and a search agent which is used by the Comparison agents can direct a search query against which the comparison agent automatically compares and evaluates offer profiles stored in a database with a search profile transmitted with the search query.

19. The network according to claim 18, and that the database with the offer profiles contained therein is contained in the comparison agent.

20. Network according to claim 18 or 19, characterized in that several offer agents are provided that transmit an offer profile corresponding to their offer to the comparison agent for storage in its database or, if the offer is withdrawn, transmit a message to the comparison agent to delete the offer profile.