WO2003023650A2

WO2003023650A2 - Method and device for determining relevant objects

Info

Publication number: WO2003023650A2
Application number: PCT/EP2002/010023
Authority: WO
Inventors: Peter Krug
Original assignee: Peter Krug
Priority date: 2001-09-07
Filing date: 2002-09-06
Publication date: 2003-03-20
Also published as: DE10143940B4; WO2003023650A3; DE10143940A1

Abstract

The invention relates to a method and a device for determining at least one relevant object from a plurality of objects for a search request. According to the invention, at least one user profile is assigned to each respective user, said profile describing characteristic features of the user and at least one respective object profile is assigned to the objects, said profile describing the features of the object and being formed by taking into consideration at least one user profile. The method comprises the following steps: determination of a user by means of a unique user identification (100); assignment of a user profile to the user (110); comparison of the determined user profile with object profiles and determination of a respective comparison value for objects based on the agreement between the user profile and the object profile (140); and evaluation of the comparison values of the objects for determining at least one relevant object (150).

Description

Method and device for determining relevant objects

The invention relates to a method and a device for determining relevant objects from a large number of objects for a user search request.

The amount of information available on the Internet is constantly increasing. In order to use the Internet sensibly as a source of information, the support of search engines for finding content on the Internet is necessary. In general, the search for certain contents is carried out by means of a search engine using a full-text method, which analyzes the texts of the available electronic documents by evaluating the position, format and number of searched terms in the documents and evaluates the documents with regard to their relevance for the search query. The analysis of the content of the documents makes it possible to determine a similarity between the search terms and the content of the documents on the basis of the match with the search terms. However, it is often not possible to make a sufficient distinction between worthless and valuable documents for the search query. All too often, based on this simple measure of similarity, the user is presented with a large number of "worthless" documents, in the content of which the search terms appear more or less randomly, but whose topic does not match the user's ideas.

An improvement in the relevance of hits can be achieved by extending the text analysis procedure to evaluate the linking structure on the Internet. In the so-called "page rank ner driving" each reference (link) to a page (electronic document) is considered a voice rated for this page. Here, the quality of the referring page can be taken into account for the evaluation of the page pointed to by the reference. The more pages refer to a potential hit page, the higher the relevance of the page is rated. However, the behavior and properties of the users are not taken into account in this method either. The knowledge of the people who use a document can be highly informative about the quality of the document and its relevance to other users.

The object of the present invention is to create a method and a device for determining relevant objects for a user search request, in which information about users who have already accessed individual objects is taken into account for determining the relevance of the objects for later search requests become.

The task is solved by the independent claims. The dependent claims relate to advantageous refinements of the invention.

The method according to the invention for determining at least one relevant object from a multiplicity of objects evaluates these objects for a user search request in order to determine their relevance. The objects are available in a computer network, in particular the Internet, for access and interaction with users. According to the invention, several types of objects are provided. A user can, for example, access electronic documents, in particular HTML, WTML or XTML documents, and view or edit them. For this purpose, the documents can be displayed on a visual display device, in particular any end device equipped with a web browser and connected to the Internet, such as a personal computer (PC), a handheld computer or a mobile phone. The interaction of the user with the electronic document can also take place by navigating in the electronic document (for example scrolling), pressing buttons (for example clicking) or entering information in fields provided for this purpose. The electronic documents can have a content of text, graphics, images, sound and / or language. It is advantageous if parts of the documents themselves again ^'stand as objects. In this way, objects that consist of complex components can in turn be subdivided into individual objects that are themselves available for interaction with users. The method according to the invention can then also determine relevant parts of the complex electronic documents as relevant objects for a search query.

It is also provided that models of real objects are available as objects. The content of the model can describe properties of the objects. The individual objects are assigned to the objects and describe them. For example, such objects can form models of articles to be sold (e.g. vehicles) that are offered for sale via the Internet. The content of such a model of such a vehicle describes its properties, such as type, year of construction, color, mileage and selling price. Such a model can also have an image of the object, for example a photo of a car.

Interaction of a user with such an object can provide for an action that the user performs with this object or with respect to this object. For example, it is possible for the user to be sent further information about the object or about the object described by pressing a button the seller of the item communicates with the user. Furthermore, a message can be sent by an action that is carried out with respect to an object, which indicates that the user would like to acquire such an object described by the object.

The method according to the invention for determining relevant objects makes it possible to specifically determine a suitable and relevant object for a search query from the large number of objects on the Internet, in particular from electronic catalogs of mail order companies. The user can order the item described by the object or request it for viewing.

A user of the method is assigned at least one user profile that describes the characteristic properties of the user. A user profile is a model of the user or a user group that describes his or her characteristic properties, for example in the form of socio-demographic data such as age, gender or place of residence, or his or her behavior. The user profile can have individual aspects, which are described, for example, by attribute-value pairs and / or in the form of a property and knowledge model of the user.

It is also possible to create a complex knowledge model about the user, in which facts and data about the user are stored. These facts and data can be collected, for example, when the user registers for the first time, possibly by evaluating a questionnaire, or through a self-assessment of the user regarding different qualities in different areas and areas of interest. Each user profile expediently consists of a predetermined series of aspects, each of which is assigned a value (attribute-value pair). As in the example above, the values can take a range from 0 to 100 points or another expedient range of values, for example 0-1, and can also assume negative values if necessary. It is useful if all user profiles consist of the same aspects so that they are comparable. Individual parts of a profile may also be inactive, for example if there is no data on the relevant aspect.

Furthermore, a user profile can also be structured hierarchically, with one aspect of the profile being represented in a refined manner by associated sub-aspects. A hierarchical structure of the user profile makes it possible to provide a more precise model of the user when there is sufficient data and to use a more general description of the user properties when the data is not very meaningful. For example, it can be useful to represent the place of residence in a tree-shaped, hierarchical structure. A user's place of residence in Munich is thus represented, for example, by an assignment to the hierarchical aspects of residence-Germany, residence-Bavaria and residence-Munich. Depending on requirements, a correspondingly fine resolution of the aspect can be evaluated. It is also possible to access a higher level of the hierarchical structure in areas in which there is little data available. For example, in a federal state in which only a few users live, a precise distinction between the users' place of residence can be dispensed with.

After registration, several user profiles can also be assigned to a user. For example, it is possible that the user has several virtual identities (for example, one for private and one for professional purposes), each with its own user profile. Depending on the needs, the user can choose between the different virtual identities.

According to the invention the objects are each associated with at least an object profile describes the properties ^"of the object. The object profile, in consideration of at least one user profile, in particular from the profiles of any group of users are formed. The property profile is advantageously of the user profiles and / or the user behavior of the users who accessed or interacted with the object The object profile represents a model of the typical user of an object, in particular it is designed as a medium user profile of the users who accessed the object.

The determination of relevant objects by the method is initiated by a user entering a search request. In the search query, the user can specify the necessary information about the properties of the objects sought.

In a first step of the method according to the invention, a user is determined using a unique user identification. The user identification is a unique identifier of the user, which is assigned, for example, when the user is initially registered. The user identification can be used to uniquely identify the user in the following search queries. For example, a user enters a user ID before a search request. It is also possible to identify the user via the input device used, with for example via a unique identifier for the input device (terminal identifier, cookie).

In a further step, a user profile is determined and assigned to the user. The determined user profile can be determined on the basis of the user identification and can be the typical user profile of the user (standard user profile). Since a user can be assigned for ^'its different virtual identities multiple user profiles, the user can also select a user profile from them.

To determine the relevant objects, object profiles are compared with the determined user profile. The comparison can be carried out with all object profiles of the available objects or with the object profiles of a part of the multitude of objects. The objects can expediently be preselected, the comparison being carried out only with the selected objects. Furthermore, a comparison value is determined for the objects for which a comparison is carried out, based on the correspondence between the determined user profile and the object profile. The comparison value of an object can be used to assess the relevance of the object for the search query. Since the object profile describes the properties of a typical user of the object, it can be concluded that the object is highly relevant to the respective user due to the high degree of agreement between the user profile and the object profile. If, for example, an object is often used by the corresponding user or by users with similar object profiles, the user profile and the object profile are similar. The method according to the invention uses this correspondence in order to assign the object a higher relevance compared to other objects which have been increasingly accessed by other users. Since the user profiles and object profiles expediently have the same structure, ie consist of a similar sequence of aspects and associated values, the user profile and object profile can be compared between the individual values of the aspects of the profiles. The comparison value is advantageously determined as the mean deviation between the values of the individual aspects.

It is expedient if the comparison between the user profile and the object profile is carried out using a suitable metric which determines the degree of correspondence between the respective elements or aspects of the user profile and the corresponding elements or aspects of the object profile. A metric suitable for this purpose can be the Euclitic distance (mean square distance) between the vector from the values of the object profile and the vector from the values of the user profile, in particular if the individual partial distances between the respective elements of the vector (ie between the respective characteristics of the individual aspects). In this case, a weighted Euclitic distance is used as the metric. However, it is also possible to use other customary distance measures to determine the correspondence between the profiles.

In a further step, the comparison values of the objects are evaluated in order to determine at least one relevant object. The objects can be presented to the user together with their comparison values or can be sent for further processing. It can also be useful to create a list of the relevant objects based on the comparison values and to output them to the user. Furthermore, the order of the relevant objects in the list can be determined based on their respective comparison values. For example, the objects are sized based on their respective comparison values, starting with the object with the largest comparison value and the highest relevance, sorted and output. During the evaluation, it is possible to output all objects, a predetermined number of the most relevant objects or the objects whose comparison values exceed a predetermined value. Furthermore, only the most relevant object can be determined.

According to an advantageous embodiment of the invention, it is provided to record a search criterion for a search query. The search criterion can include information about the type (electronic document or model of an object) and / or the content of searched objects. The information about the content of searched objects can relate to the type of content (e.g. text, graphics, image, sound or language) and / or describe specific searched content, in particular as search text, image or sound description. Such a search criterion enables the electronic documents searched for to be described in terms of their content and to be searched in the large number of available electronic documents. Furthermore, the search criterion can also specify the content of such searched objects, which are assigned to real objects and describe their properties. It is thus possible to specify the desired properties of the objects sought and to determine the objects assigned to them. For example, the desired properties of a searched vehicle or a description of a searched video are specified in the search criterion. Information about the content sought, in particular search terms or search texts, and / or the type of content, for example text or audio, can be given for describing the Internet pages sought.

The search criterion is advantageously compared with the respective content of the objects and / or the object itself. The comparison values of the individual objects can take into account the match can be determined between the search criteria and the content of the object and / or the object itself. If there is a high level of agreement, it is advisable to increase the comparison value of the object accordingly in order to improve the relevance rating of the object. The comparison value of the object is advantageously determined taking into account the correspondence between the user profile and the object profile and the correspondence between the search criterion and the content of the object, since the relevance of the objects can result both from the importance of their content for the search query and their typical use.

According to an advantageous embodiment of the invention, it is provided that objects are preselected. The preselection can take place on the basis of the respective match between the search criterion and the content of the object and / or the object itself. A comparison value is expediently determined only for the selected objects. Such a procedure can reduce the number of objects for which the comparison of user profile and object profile is to be carried out. It is advantageous to select a predetermined number of objects and to determine their relevance, or to determine a number of the selected objects dynamically on the basis of the respective agreement with the search criterion. For example, the relevance for the best 10 objects, the best 10% of the objects or for the objects whose match with the search criterion lies above a predetermined value can be determined. These various options automatically make it possible to select a reasonable number of objects from the possibly very large number of existing objects.

According to a particularly advantageous embodiment of the invention, the comparison values of the individual objects can be determined taking into account a relevance weighting of the respective objects. In particular it can be provided to take into account the frequencies of access to the objects by the users and / or the time derivation of the access frequencies. Furthermore, the intensity of use for accessing the objects can be used as an "intrinsic value" of an object, which describes the relative importance of the object. The intrinsic value of an object can, for example, be determined separately for each user, for a group of users or for all users together for a specific object. The relevance weighting of an object can include all influencing variables for the importance or relevance of the object that are independent of the object profile and the search criterion.

The so-called action equivalence can be used as a further influencing variable on the importance of an object. The equivalence of action increases the importance of objects with high usage intensity. The usage identity of an individual user or any user group is taken into account. The influence of action equivalence on the relevance weighting of the objects makes it possible to increase the relevance of user- or group-specific favorites (i.e. objects with high usage intensity).

It is furthermore advantageous if the relevance weighting of an object is carried out taking into account an assessment of the object, which is given by a user accessing the object. After each access to an object, it is possible for the user to evaluate the relevance of the object for the corresponding search query. The evaluation can take place, for example, on the basis of a predefined scale from which the user selects a suitable evaluation measure. The ratings of an object by users can contain information about the importance of the object and can help to ensure that unimportant and irrelevant objects are rated with low relevance in further search queries. The evaluation The evaluations can be carried out separately for the individual users or collectively for one object for all users.

It is particularly advantageous to determine the comparison values of the objects, taking into account the correspondence between the user profile and the object profile, the correspondence with the search criterion, and the relevance weighting. By taking these factors into account when determining the comparison values, the various influencing factors can be taken into account, which determine the relevance of an object for a search query of a specific user. It is advisable to appropriately calculate the individual influencing factors on the comparison values, for example as a product or a sum. Furthermore, the individual influencing factors can be included in this calculation, weighted according to their respective importance.

In a special embodiment of the invention, it is provided to provide user groups, each with its own user profile. The groups of users can be predefined (for example, all women) or determined automatically (for example, all users who access a certain object). A single user can be assigned to any number of groups. The assignment to a group can be initiated by a user himself, ie can be done explicitly, for example by logging into a group or community. Furthermore, the assignment can also be made implicitly, for example on the basis of the presence of certain properties of the user. The profile of a user group is expediently formed on the basis of the user profiles of the group members, in particular as a medium user profile of the group members. For a search query, a user can select the profile of a user group in order to use their characteristic properties for the relevance assessment of the objects. For example, by choosing a group profile for auto mechanics for a search query with a search criterion that relates to automobiles, objects with a more pronounced technical aspect are determined. By selecting the group profile, the assigned user profile of the user for the search query is the middle user profile of the group members.

User groups can also advantageously be used to determine the ^{equivalence of} action. By selecting a corresponding reference group, the objects can be determined that have the highest usage intensity in the reference group. For example, by selecting the group of auto mechanics, the collective favorites of this group can be given special consideration when determining the relevance weighting of the objects.

A particularly advantageous embodiment of the method provides for the user profile of a user to be changed over time in order to be able to detect changes in the properties of the user and in his usage behavior. It is advantageous to form the user profile taking into account the objects that the respective user has accessed in the past. In this way, the usage behavior of the user can influence the user profile. The degree of this influence can be determined by the intensity of the use of the object by the user, for example by the time a user spends looking at an electronic document. It is also possible to take into account an evaluation of the object by the user (for the change of the user profile by the object profile). The influence of the intensity of use of the object and / or the evaluation of the object can take place individually for each user, for a group of users or for the total of the users. In the latter case, be for an object determines an absolute usage intensity and an absolute rating from the corresponding values of the individual users.

In order to adapt to a changed usage behavior of the user, a dynamic part of the user profile can be updated after each interaction of the user with an object. The user profile can expediently have a dynamic and a static part. The dynamic part of the user profile can have a neutral expression of the individual aspects immediately after the registration of the user and can adapt to the usage behavior of the user through the interaction of the user with objects, in that the dynamic part through the object profiles of those objects that the user accesses, is changed. When updating the user profile, it is advantageous to take the object profiles into account as a function of the time of access to the object and / or the intensity of use of the object. The object profiles of objects that have been recently accessed and used by the user with high intensity can have a greater influence on the updating of the user profile than those of other objects that the user has accessed recently and only briefly.

By taking the object profiles into account when updating the user profiles, the usage behavior of the individual users can be automatically detected and taken into account in the method according to the invention. Since it is possible to change the user profile through the object profile after each interaction of the user with an object, individualized user profiles are created automatically, which correspond to the preferred interests of the users. For example, the political aspect of a user is reinforced by increased and intensive access to documents with political topics. Due to the increased political aspect the user's interest in further search queries is taken into account by giving preference to those objects which also have a pronounced political aspect when evaluating the relevance of objects.

In order to reduce the influence of incorrect or unintentional access by a user to irrelevant objects, it is expedient to change the object profile by the user profile of a user accessing the object, taking into account an assessment of the object by the user. If a user recognizes after accessing an object that it is of no importance for the search query and / or his interests, a poor evaluation of the object can reduce the influence of his user profile on the object profile and thus the formation of "wrong" properties prevent in the object profile.

Furthermore, it can be advantageous if the object profile of an object is changed by the user profile of a user accessing the object, taking into account an intensity of the use of the object by the user. Use with high intensity can thus lead to a stronger influence of the user profile on the object profile than short-term access to the object. In particular, the number of accesses to the object, their chronological sequence, the duration of an access and the type and number of interactions of the user with the object can be taken into account in order to determine the influence of the user profile on the object profile.

A further embodiment of the invention provides for the object profile to be determined using a cluster method, the user profiles of the users who have accessed the object being taken into account. The cluster process allows the user profiles of these users to be assigned to one or more typical user profiles. The object profile can be determined taking these typical user profiles into account. In particular, a certain typical user profile can be assigned to the object profile, the user profiles not being taken into account by non-typical users. In this way, the influence of random access by non-typical users on the object profile can be reduced.

At least some of the objects are expediently designed as electronic documents, in particular HTML, WTML or XTML documents, the respective contents of which have text, graphics, images, sound and / or language. By evaluating the comparison values of the objects, the respective relevance of the object is determined, which indicates the importance or the rank of the object for the relevant search query. The method according to the invention can determine the relevant electronic documents for the search request from the large number of electronic documents available on the Internet. The search criterion expediently has information about the type and / or content of objects. For the determination of relevant electronic documents, this information can be in the form of text, graphics, images, sound and / or language.

The determination of the similarity between the respective textual content of the objects, in particular the content of electronic documents, and a search text can be carried out, for example, using a full-text analysis method. The full-text analysis method allows a search text and / or search term to be compared with the textual content of electronic documents using a similarity measure for character strings, for example the Levinstein distance, whereby omissions and / or insertions of words can be permitted. In order to determine the correspondence between the search criteria and the content of the object, it is advisable to use a full-text analysis method. Furthermore, the full-text analysis method can be used to preselect objects, which are then subjected to a further relevance assessment by the method. The preselection can, for example, be based on the similarity of the search text and the textual content of the object.

The text analysis can be carried out, for example, by known Internet search engines. The Internet search engines determine a large number of potentially relevant objects for a given search text based on the textual correspondence between the search text and the textual content of the electronic documents. It is advantageous to take the evaluation of the textual match by the search engines into account when determining the comparison values of the objects. It is also advantageous to have the text analysis carried out by several different search engines and to take the different results into account. For example, intersection or union sets of the search engine results can be formed and the respective evaluations of the individual documents can be combined with one another. It is particularly advantageous to weight the results of the individual search engines for the determination of the comparison values of the method according to the invention, in order to take into account different properties and qualities of the search engines.

The present invention makes it possible to determine objects relevant to a search query from a large number of objects, for example the websites available on the Internet. The relevance of the object for the user can be determined individually by comparing the user profile and the object profile. Since the characteristic properties of the user flow into the relevance assessment of the objects, the same search query can be carried out for different users with different results. lead results. Since the object profile is formed taking into account the user profiles of the users who accessed the object, it describes the typical properties of the average user of the object. The method according to the invention provides that the object profiles are formed automatically by the accesses and interactions of the users with objects, without manual intervention, for example by cataloging the objects or assigning properties, being necessary.

Since it is also possible according to the invention to design the user profiles taking into account those object profiles which the user has accessed, the characteristic properties of the users can be formed automatically through their usage behavior. The relevance of the objects for a search query is determined by evaluating the comparison values of the objects and indicates the importance or the rank of the object for the search query.

Further special features and advantages of the invention can be found in the figures and the following description of preferred exemplary embodiments. Show it:

1 shows a flow chart to explain the method according to the invention;

Fig. 2 is a schematic representation of a user profile

Fig. 3 is a flow chart for explaining the update of user and object profiles;

4a and b show two possible implementations of the method according to the invention;

Fig. 5 is an illustration for explaining a possible

Realization of the invention; and 6a and b representations to explain an embodiment.

1 shows a flow chart to explain the method according to the invention. In step 100, a user is determined using a unique user identification. The user has previously logged on with the user identification, which was assigned to him, for example, when registering for the first time. By identifying the user, it is also possible to assign individualized user profiles to the user in order to use user properties when determining relevant objects.

In step 110, the user is assigned a user profile. This can be, for example, the standard profile of the user, which is assigned to the user on the basis of the user identification. Furthermore, it can be provided that the user has several virtual identities, each of which is provided with its own user profile. So it is possible that the user uses one virtual identity for private and another for professional purposes. The user can choose between these different virtual identities for a search request, the respective user profile being assigned to him.

It is also provided that the user can select a user profile from a large number of group profiles for a search query. The group profiles each describe the characteristic properties of the average user of the respective group, whereby each user can belong to any number of groups. The assignment of a user to a group (community) can be done explicitly by the user himself, for example by logging into a community. Furthermore, a user can join a virtual community be ordered, the assignment being based on certain properties. If a user fulfills these properties, he is automatically assigned to the corresponding group. In this way, a large number of virtual communities are available, for example the group of all women or the user of a specific document. The respective group profile is formed on the basis of the user profiles of the group members, for example as a medium user profile, and describes the typical properties of the group.

In step 120, a search request is entered by the user. The search query can have a search criterion that describes the content of the searched objects. The search criterion contains information about the type (electronic document or model of a real object) and / or content of searched objects. The information about the content can relate to the type of content (text, graphics, image, sound, language) and / or describe specific content sought. For example, search terms, search texts or descriptions of images or music can be used as search criteria.

In step 130, the search criterion is evaluated and compared with the respective content of the objects. A comparison value is formed for each object, taking into account the correspondence between the search criterion and the content of the object. For the search for electronic documents with a content of text, graphics, images, sound and / or language, the respective type of content is taken into account when determining the match with the search criterion. In the case of a textual content, the correspondence between the search criterion and the object content can be determined using a full-text analysis method. In order to determine the similarity to a search text, methods of Information retrival, for example semantic indexing, can be used.

To reduce the effort required for the further procedure, it is possible for a preselection of objects to be carried out on the basis of the match between the search criterion and the object content. This preselection can limit the number of objects that are still to be processed by the method. Furthermore, the results of different text analyzes can be used by summarizing the different results and weighting the individual evaluations of the objects by the different text analysis methods.

In step 140, the comparison of the user profile with the object profiles takes place, wherein only the object profiles of those objects that were selected by the preselection for further processing may be taken into account. Based on the correspondence between the user profile and the object profile, a comparison value is formed for each object. The correspondence between the profiles can be determined, for example, by a Euclitean distance measure. The comparison values of the objects are determined taking into account the distance between the profiles and the match with the search criterion. Furthermore, it is useful to consider a relevance weighting of the objects that is independent of these criteria. The relevance weighting of an object indicates its importance or relevance regardless of the respective search query.

In step 150, the comparison values are evaluated to determine the relevance of the individual objects. The comparison values can assume values between 0 and 100, for example. The evaluation of the comparison values can be based on an absolute criterion, for example all equivalent values> 90, or a relative criterion, for example the largest or the largest three values or the upper third of the comparison values. The objects can be displayed to the user in a sorted list, possibly together with their comparison values or their ranks in the list. It is also possible to determine and output the rank of an individual object. The result of a search query for determining relevant objects is presented to the user after the evaluation of the comparison values. The user now has the option of selecting individual objects and accessing them. He can interact with the objects and, for example, send an order for an object that is described by the selected object. The described method enables the user to locate relevant objects and to access them.

In step 160, the method is ended or, by branching to step 120, a new search query is entered.

2 shows a schematic representation of a user profile. Each user of the method is assigned at least one user profile that describes its characteristic properties. The user profile can consist of a static and a dynamic part. Furthermore, each object, in particular an electronic document on the Internet, is assigned an object profile which represents a model of the typical user of the object. In addition, the user groups are assigned group profiles, each of which forms an average user profile of the group. The profiles characterize usage behavior and properties of the users to whom they are assigned. Each profile consists of a predetermined number of aspects, with each aspect being assigned a value which is the expression of this aspect for the assigned Describes users or the assigned user group. All profiles consist of the same aspects and are therefore comparable. However, it is possible that parts of a profile are not activated and are not taken into account in a comparison.

The static part of a user profile is determined, for example, when the user is registered. A static profile part is assigned to each virtual identity of a user. Each registration date leads to the strengthening or weakening of individual aspects. In this way, a limited number of clearly defined and comparable aspects can be formed from a large number of registration data, which are stored, for example, in a complex knowledge model. Examples of static profile data are socio-demographic aspects, which are formed from the registration data, and technical aspects, which are formed from a self-assessment of the user on technical qualities in different areas and areas of interest, for example sports, music, politics, entertainment.

2 shows an example of a section from such a user profile. The individual aspects of the user profile are listed in the first column. The second column shows the values that correspond to the formation of the aspects in the static profile area. The third column concerns the dynamic profile area. The overall expression of an aspect is determined by the respective value of the static and dynamic profile area. In the example shown, unchangeable properties of the user are also taken into account in the user profile. The example shows the user profile of a female user of the method between the ages of 20 and 40. In the basic state, the dynamic values of the user profile are initially neutral. Every interaction with objects, for example access to an electronic document, leads to a dynamic influence on the profile by changing the dynamic values of the aspects through the interaction. The dynamic profile area reinforces or weakens the aspects of the static part in the overall profile.

In the example shown, it can be seen from the dynamic profile area that the user does not predominantly access objects which are preferred by female users, since the dynamic value of the gender-female aspect is “58”. Furthermore, it can be seen from the relatively high dynamic value "70" of the technology aspect that, contrary to the expectations that have arisen from the registration data, the user is interested in objects with technical properties.

3 shows a flowchart to explain the updating of user profiles and object profiles through user interactions. In step 200, the user accesses an object. The object can be accessed by displaying the object, for example an electronic document, on the user's visual display device. The user can interact with the object by scrolling, scrolling, or clicking buttons.

In step 210, the user decides whether to evaluate the object.

When the user evaluates the object in step 220, the user evaluates the importance and relevance of the object. This can be done, for example, on a scale from which the user selects a value and assigns it to the object. For the evaluation Different evaluation standards such as relevance, clarity, graphic representation, intelligibility or other possible.

In step 230, a usage analysis is carried out to determine an intensity factor for the interaction of the user with the object. An intensity factor I for the interaction is determined on the basis of the active time for the access to the object by the user or the active time for the interaction with the object and / or the evaluation for the object. The intensity factor I indicates how pronounced the interaction of the user with the object was and how the user evaluated the object. After each interaction, the value for the usage intensity of the objects can be recalculated.

The intensity values for each object used are stored, for example, on the user's computer (client). The ten largest intensity values of the user are transmitted to a central computer system (server) together with object identifiers, for example the URLs (Universal Resource Locators) of the objects. The usage intensities of the individual users for the different objects are stored there in an absolute usage intensity matrix. The intensity factors I of the respective interactions and / or the interaction times and user ratings for the object can be transmitted to the central computer. In the latter case, an absolute evaluation matrix of the individual evaluations can be stored by the different users for the respective objects on the central computer system. The individual elements of the rating matrix can be determined as the average rating of the user for the respective object. Furthermore, an absolute interaction time matrix can be stored in the central computer system, in which the sum of the interaction times of the individual users is formed for the different objects. Based on the usage data determined in the usage analysis, the usage intensity of an object can be determined by an individual user or by a group of users, or the totality of users. The specific and absolute usage intensities can be used when updating the profiles to determine the influence of the interaction on the profile and / or to determine the comparison values of the objects when determining the relevance criteria.

The mth user interaction can be described as follows:

Am = (c _m , k _m , ti _m , B _m ), where:

Cm - is the user index from the set of users for the mth user interaction; k _m - is the object index from the set of objects for the mth user interaction; ti _m - is the active interaction time for the mth user interaction; and

Bm - represents the user rating for the mth user interaction.

In general:

U = {UW _» ... _» ^ ⁾ } is the set of users;

O = {Ol ¹ >, ..., Ota ⁾ } is the set of objects; and

C = {CW, ... _» ^ ⁾ } is the set of groups.

The intensity factor describes the intensity of the use of an object by the user. It describes the usage behavior of the user as an elementary variable of the system. The intensity factor of a usage and the absolute usage intensity for an object can be differentiated. The intensity factor relates to the current interaction on. The absolute usage intensity accumulates all interactions of a user for an object.

The intensity factor I for an interaction is determined by analyzing the interaction of the user, taking into account the actually active time for the interaction with the object and / or the evaluation of the object by the user:

Im + i = f (ti _m + ι, Bm + ι); in which

Im + i is the intensity factor of the m + lth interaction; ti m + i is the active interaction time; Bm + i is the user rating for the property; and f grows with both parameters.

After each interaction, the value for the usage intensity of this object is recalculated. The intensity values for each object used can be stored in user computer 1 (client). The N largest intensity values of each user can be transmitted to a central computer 2 (server). A favorite matrix with freely segmentable user groups can be determined there.

The elements of the absolute usage intensity matrix I ^A (U, 0) contain the usage intensity I ^A (Uj, Ok) of the kth object by the jth user:

; B _m ^A _{+ l} (U _j , O _k )), where ti ^ J., O _k ) is the absolute interaction time of the jth user for the kth

Object and B _{+ l} (U., O _k ) is the absolute rating of the jth user for the kth object. The elements of the absolute rating matrix B ^A (U, 0) contain the average rating B ^A (Uj, Ok) of the jth user for the kth object:

BV _k )

The elements of the absolute interaction time matrix ti ^A (U, 0) contain the sum of the interaction times ti ^A (Uj, Ok) of the jth user for the kth object: vj _{m + l} ≠ j

Each profile consists of a predetermined number n of aspects, each of which is assigned a value. In order to make the profiles comparable, all profiles consist of the same aspects.

The user profiles consist of a static and a dynamic part. The profiles of the users are the basis for the profiles of the groups and objects.

The static profile area of the user profile is determined when registering and creating virtual identities. A static profile part is assigned to each virtual identity. Each registration date leads to the strengthening or weakening of individual aspects in the static part of the user profile in order to reduce the theoretically unlimited amount of possible registration data to a limited number of clearly defined and comparable aspects. The following applies to the static user profile structure:

Us ^ü =

Usi ^j) ), where

Usü ^{) is} the vector of the static profile aspect values of the j-th user.

The dynamic profile area of the user profile has the same aspects as the static area. The aspect values of the dynamic user profile are initially equally strong in the basic state, i.e. neutral training. Every interaction (use of documents, etc.) leads to a time-sensitive, dynamic influence on the profile. The dynamic profile area reinforces or weakens the aspects of the static part in the overall profile.

The following applies to the dynamic profile structure:

in which

Udmü ^{) is} the vector of the dynamic profile aspect values of the jth user after the mth interaction.

The overall user profile results from the sum of the static and dynamic profile values.

Every object that is accessed by a user (eg a web page displayed in the browser window) is indexed in a database. An average user profile for the respective object is formed from the user profiles of the users, who are known as users of an object. The profile of an object has the same structure as the user profiles. It describes the typical user of the object and depends on their profile values and the frequency and intensity of use.

Each use of an object by a user leads to a change in the profile of the object in question. It gets a little adjustment to the user's profile. The degree of alignment depends on the intensity of use (intensity factor I). In addition, the frequency and chronological sequence of access can have an impact on the object profile.

The following applies to the object profile structure:

₀ ^(k) ₌ (QW, ..., θJ2), where O _m < ^k > is the vector of the profile values of the kth object after the mth interaction.

In step 240, the influence of the interaction on the user profile of the user is determined. For this purpose, user profile influence values are determined which determine the influence of the interaction, the intensity of use, the evaluation of the object and the object profile 320 on the dynamic user profile 300. The user profile influence values depend on the intensity I of the interaction with the object and the difference between the object profile 320 and the static user profile 310.

The user profile influence values OU_ determine the influence of the user interaction (usage intensity and user rating) and the object profile on the dynamic user profile. The user profile influence values depend on the intensity of the underlying use (I _m + i) and the aspect difference from the object profile and the static user profile:

V /

, in which Oi, _m ⁽⁾ the i-th profile aspect value of the k-th object after the m-th

Interaction and h is growing with both parameters.

The dynamic part of the user profile 300 is determined as the average value of the user profile influence values.

The dynamic user profile values Udι- _n are calculated as the average value of all previous user profile influence values:

^Ua i, m + l ~ n <Λ - + '> + l' ^W ° ^bel

pm ^{ü) is} the number of user interactions of the jth user after m interactions.

In step 250, the influence of the interaction on the object profile 320 is determined. The object profiles 320 have the same structure and the same structure as the user profiles 300, 310 and are designed as an average user profile for the respective object. Object profile 320 describes the typical user of the object. Each use of an object by a user leads to a change in the object profile 320 of the relevant object by an adaptation to the profile of the respective user. The degree of alignment depends on the intensity of use. In addition, the frequency and chronological sequence of the accesses can affect the object profile 320.

The object profile influence values determine the influence of the interaction with the object, the intensity of use, the evaluation by the user and the static user profile 310 on the object profile 320. The object profile influence values are dependent on the intensity I of the interaction and the static user profile 310. The object profile 320 is determined as the average value of the object profile influence values.

The object profile influence values UO_ determine the influence of the user interaction (intensity of use and user rating) and the static user profile on the object profile. The object profile influence values depend on the intensity of the underlying use I _m + ι and the profile aspect values of the static user profile:

vz

h (Us, ^{(Λ, + l)} , / _{m + 1} ), where h is growing with both parameters.

The calculation of the object profile values (object) Oι- _n takes place as the average value of all previous object profile influence values:

, in which

rm ^{t> is} the number of object interactions of the kth object after m interactions.

In step 260, the influence of the interaction on the group profiles 330 is determined. The profiles of the groups are generated from the profiles of the users on the basis of a member directory 340 of the groups. They describe the average (typical) user of a group. A specific typical user with specific properties can be used by the group profiles 330 in the relevance assessment of objects. The group profiles 330 can be determined after each interaction as an average of all members of the group. The profiles of the groups are generated from the profiles of the users. They describe the average (typical) user of a group.

The following applies to the group profile structure: C =

, in which

C _m ^{(1) is} the vector of the profile aspect values of the 1st group after the mth interaction.

The i-th profile aspect value of the 1st group after m interactions is calculated as the average value of the profile aspect values of all members of the group:

In step 270, if the user does not want to cancel the method, the method branches back to step 200 and the method is continued with another access to a new object.

FIG. 4a shows a possible implementation of the method in which search queries by a user are recorded on a user computer 1. Based on the search query, the search criterion, for example a search text, is transmitted to a full-text search engine 3 (I). The search engine 3 carries out the comparison of the search criterion with the content of the objects and generates a hit list with evaluation for selected objects. This hit list is then transmitted back to user computer 1 (II). The user computer 1 forwards the URLs of the hit list to the central computer 2 of the system (III). There the hit list is expanded by the object profiles of the respective objects and sent back to the user computer 1 (IV). The user computer 1 now carries out the comparison of the user profile with the object profiles and determines the comparison values for the selected objects. The comparison values are then evaluated by the user computer 1 and the relevant objects determined in this way are presented to the user.

4b shows another possible implementation of the method according to the invention, in which a search request is recorded by a user and transmitted from the user computer 1 to the central computer 2 (I). The central computer 2 starts corresponding search queries on the full-text search engine 3 (II) using the search criterion. The search engine 3 compares the search criterion with the object contents and, taking into account the correspondence between the search criterion and the object contents, determines an evaluation for the objects, on the basis of which the objects are preselected. The selected objects, together with their evaluation, are sent back to the central computer 2 in a hit list (III). The central computer 2 carries out the comparison of the object profiles with the user profile of the user and determines the comparison values for the objects. After the evaluation of the comparison values, a list of the relevant objects is transmitted to the user computer 1 (IV) and presented there to the user. This implementation option is particularly suitable for carrying out a comparison of the search criterion with the object contents of different search engines using different text analysis methods. The central computer 2 can send a search request to several search engines 3 and combine the respective results in a suitable manner.

5 shows a further exemplary embodiment of the method. A software program according to the invention is installed on the user computer 1 as a plug-in 10 for a commercially available internet browser 11. The software program 10 logs the accesses and interactions of the user with the Internet browser 11 to objects 4 on the Internet. The Information about the objects (object identifier URL, usage intensity I) is forwarded to the central computer 2. A search query of the user, as described, is either forwarded to the central computer 2 or to the search engine 3.

6a and 6b show an example for the determination of the usage intensities and the inner values of the objects. The intensity of use I is determined after ^" every interaction on the user computer 1 for the respective object. Furthermore, the inner value W of the object, which describes its relative importance, is determined after each interaction. The inner value of an object can be used as part of the relevance weighting for the determination of the comparison value of the object can be used (step 330).

In the event of a significant change in the usage intensities or the inner values of the objects as a result of the interaction, the updated values are transmitted to the central computer 2 (step 350). This can happen, for example, if there is a change in a list of objects with the best ratings. The usage intensities and inner values for the objects which the user has accessed or interacted with are stored in the user computer 1 (client) in a memory area 12.

The received values of the usage intensities and inner values of the objects are stored in a memory area 13, 14 in the central computer 2 (server). The ratings for the most important objects of the user are stored there for each user, for example the top ten ratings. Furthermore, 2 memory areas are provided in the central computer for storing the absolute evaluation matrix B (U, O) and the absolute interaction time matrix ti (U, O). The usage data of the users for all objects are stored in a memory area 14. Using the data stored in the memory area 14, the absolute inner value of the objects can be determined as an average value for all users and stored in a memory area 15. An example of the stored data is shown in Fig. 6b.

The comparison values of the objects are determined on the basis of various criteria, which are based on the correspondence between the user profile and the object profile, on the correspondence between the search criterion and the content of the object or other relevance criteria, for example the intrinsic value of the object or the frequency of accesses to the object. A comparison value is determined for each object, which lies, for example, in a value range from 0 to 100.

Each object is evaluated according to the individual criteria. A value is assigned to each criterion, the weighting of which can be varied. The comparison value of the object results, for example, as the product or the sum of all values. Each of the criteria is weighted and can be varied depending on the type of search query.

The comparison value for the kth object and for the jth user after m system interactions is calculated using the following formula:

TRm (k) = p ((WOm (k), zl), (WCm (k), z2), (WUm (k), z3), (GRm (k), z4), (OC m (k), z5), (d (Us (j), Om (k)), z6), (HA (k), z7)), where p is a function that grows with all parameters and decreases with the sixth parameter; zl, ..., z4 are the weights for the corresponding criteria of TR; and d (Us (j), Om (k)) is the difference between the profile aspects of the k-th object Om (k) and the j-th user Us (j) after m interactions.

The following applies to the individual criteria:

WOm (k) is the absolute inner value of the kth object;

WUm (k) is the user-specific inner value of the kth object;

WCm (k) is the intrinsic value of any user group for kth object;

GRm (k) is the basic priority of an object;

OCm (k) is the communication weight of the object; and

HA (k) is the action equivalence of the object.

The action equivalence of an object HA (k) is important for evaluating the object and / or for determining favorites. The basic idea of equivalency of action is that what a user has already viewed has an impact on their future expectation for searched objects. However, the objects sought - especially when searching for a new topic - should go beyond the user's level of knowledge. The action equivalence determines a weighting of objects / documents with the highest usage intensity of any reference group. The reference group can be a basic community of the requester or the users of the objects in a hit list.

The user's own action equivalence allows the user's personal favorites, ie the objects with the highest user intensity, to be determined. This makes it easier to find your personal favorites for later search queries because they are have a higher comparison value and are therefore potentially relevant objects.

By considering the cumulative usage intensity for objects of any group in the collective action equivalency, collective favorites of the group can be determined. The groups can be of any nature, for example a real community, any virtual community, or selected according to regional aspects. All users of an object can also form a reference group.

The intrinsic value of an object W describes its "relative importance". The intrinsic value is determined by the number of users who have used this object and their actuality, ie the time when this object was used and how often it was used. Each Use and possibly a positive evaluation by a user increases the intrinsic value of the object. The time sensitivity of the intrinsic value leads to an automatic - time-dependent gradation of the intrinsic value. The more users have positively evaluated the object, the more current this evaluation is and the more access on the object, the greater the intrinsic value of the object. Objects with great intrinsic value are relatively important, ie they are potentially of great relevance.

In general:

ψ w _n , »„ i + l ψ rt _{m + i} - t _m ; ψ asymptotically restricted with 0; ψ> 0

The inner value is calculated before or after each user interaction: W _o = 0; t _o = 0

^, + ι = - ^~ - ⁽ ^ + + ι)

" ^{, + 1} ">, where

Where is the absolute intrinsic value for an object;

Wu is the custom intrinsic value for an object; and

WC is the intrinsic value of an object for any group.

Taking into account the intrinsic value of the objects used by a user, for example, the user's personal favorites can be determined.

Different (full-text) search engines can compare the content of the objects with the search criterion and, if appropriate, preselect objects with a high degree of agreement between the search criterion and the object content. Each object can be assigned to one or more search engine hit lists. A basic priority of the object is determined on the basis of the results of the search engines for the search query. The basic priority GR of an object depends on a weighting of the search engine and the rating calculated by the search engine. It is possible that for a certain search query an object is number five in a search engine's hit list and number one in another search engine.

In order to take into account the quality of the results of the individual search engines, each search engine used can have its own weighting: Sg ^ Sgt ⁾ , ... ^ ⁸⁾

It is possible to provide the search engines with a manual weighting. In addition to the search technology, for example full-text analysis methods, the weighting can also depend on the possibility of transferring the search question and may vary depending on the type of query to be submitted. Search engines that are more suitable for certain types of search queries can be given a higher weight.

Another possibility provides that the weightings of the search engines are determined dynamically. This can be done taking into account the amount of interactions with the hits of each search engine that were made after a search query by the user. A comparison with the previous intensity factors can determine how the weightings of the search engines are changed. It is also possible to provide specific search engine weights, for example for each user, for individual communities. Depending on the type of search query, the respective weightings of the search engines can be used to calculate the basic priority.

Each search engine assigns its own priority rating: R _z ⁽ⁱ⁾ to each object. This priority assessment and the resulting order of the objects can be used to determine the basic priorities. For example, the position of the individual documents in the hit list of the respective search engine can be evaluated with a factor. It is also possible that each search engine forms its hit list as a set of pairs of hits (object) and priority of the hit (search engine-specific information;%, points, etc.):

T «H (Tι«, Rι ^e, ), (2 ⁽ⁱ J, R2ffl), ..., {Tpi ⁽ », Rpi < ^i} )}, whereby

R _Z W is a priority rating of the zth hit from the ith search engine.

The search engine-specific priority ratings can be normalized and offset. For the calculation of the basic priority, it is further assumed that the priority ratings increase with the importance of the assigned objects in a value range from 0 to 1. The basic priority of an object (hit) can be calculated as follows:

GR.W = pr (R _z W, Sg ⁽ⁱ⁾ ), where pr is a function that grows with both parameters.

If a hit is in the hit lists of several search engines, the largest value of the hit's basic priorities from the hit lists or the average priority value can be used as the basic priority of the assigned object.

The profile comparison determines a measure d for the match between the user profile and the object profile (average user profile of the hit). This is done by comparing the individual aspects of the searcher's user profile and the average user profile of the object (object profile). In this way, the average deviation d of the individual aspects can be determined. The lower d, the higher the priority (relevance) of the object. If the search query is selected by selecting a group (e.g. a community with a thematic specification), the group profile of the selected community is used instead of the user profile.

The profile comparison for a search request in which the user profile of the user was selected is carried out by comparing the user profile and the object profile and determining a dimension d for the distance of all profile aspect values between the object and user profile. Each aspect has a weight for the value of the distance: , in which

Gi is the weighting of the individual profile aspects, with:

"m V ^ m '^ 2,111» - »^ n ,,,,)

The profile comparison for a search request using a group profile is done by comparing the group profile with the object profiles. A dimension d for the comparison between group and object profile is determined as follows:

, in which

Gi is the weighting of the individual profile aspects, with:

Furthermore, it can be determined how many users of an object are members of which community, ie the distribution of the memberships of the users of the object in user groups. A community weighting can be assigned to each object, which is based on the usage history of this object by users with memberships in various which communities are determined. The following weighting results for the kth object: OC ^(k) = (OCι ^(k) , OC ^(k) , ..., OCι ⁽⁾ ). After m + lth use of the kth object, every ith element of the community weighting changes.

Furthermore, the number Nk, m of the previous community interactions with the k-th object and the number Guj _{m of} the communities to which the user j _{m of} the m-th interaction belongs, and the community weighting can be determined:

N _m if user belongs to i - community

O ^{L /} c ^l - '; ^w , »ι + ι • oc2 +. ^{N + Gu} J. ^N ^ ^{+ Gu} J _m 0, otherwise

N _{k, m + l} = N _m + Gu _Jm

Based on the determined comparison values of the objects, an evaluation is carried out to determine the relevant objects from the large number of available objects. The evaluation can be carried out, for example, by sorting the objects on the basis of their comparison value, a high comparison value indicating great relevance and a smaller comparison value indicating less relevance. The objects are presented in a sorted hit list, possibly with their rating, to the user, who can select individual objects and access them. Since the personal properties of the user are taken into account in the comparison values of the objects, the method according to the invention enables an individual determination of relevant objects for a search query that is tailored to the personal properties of the user. Objects with a high level of correspondence between the user profile and the object profile, as well as objects with a high relevance criterion, are assigned relatively high comparison values. These objects that are important for the user can be determined in the evaluation on the basis of their comparison value and presented to the user in a preferred form. The intensity of the use of the objects, their assessment by the users accessing them, the frequency and / or time of access to the objects can be recorded, evaluated and used in a variety of ways in the method according to the invention. For example, you can influence the following processes:

- the determination of the comparison values;

- the dynamic part of the user profile;

- the relevance weighting of an object; and

- the change of the object profile.

The relevant objects can be presented by sorting the objects according to their relevance and presenting the objects in the order of their relevance, starting with the most relevant object. Furthermore, particularly relevant objects can be marked in color or differently during the presentation. The hit lists, which contain the relevant objects and are presented to the user, can have an object identification, for example the URL (Universal Resource Locator) of the object, a brief summary or description of the content of the object and an indication of the relevance assessment of the object by the method ,

Claims

Expectations

1. Method for determining at least one relevant object from a multiplicity of objects for a search query, wherein at least one user profile is assigned to a user, which describes characteristic properties of the user, and at least one object profile is assigned to the objects, the properties of the object describes and is formed taking into account at least one user profile, the method comprising the following steps:

Determining a user by means of a clear user identification,

Assigning a determined user profile to the user,

- Comparison of the determined user profile with object profiles and determination of a comparison value for objects based on the correspondence between the user profile and the object profile and

- Evaluation of the comparison values of the objects to determine at least one relevant object.

2. The method of claim 1, wherein a search criterion for at least one object to be determined is detected, compared with the respective content of the objects and the comparison values of the individual objects are determined taking into account the correspondence between the search criterion and the content of the object.

3. The method according to claim 1 or 2, wherein the comparison values of the individual objects are determined taking into account a relevance weighting of the objects, in particular taking into account the Frequency of access to the objects by the user and / or the temporal derivation of the access frequency.

4. The method according to claim 2 or 3, wherein a preselection of objects is carried out, in particular on the basis of the respective match between the search criterion and the content of the object, and a comparison value is determined only for the selected objects.

5. The method according to at least one of claims 1 to 4, wherein the object profile is formed from the user profiles and / or the user behavior of the users who have accessed the object.

6. The method according to at least one of claims 1 to 5, wherein the object profile is designed as an average user profile of a predetermined group of users who have accessed the object.

7. The method according to at least one of claims 1 to 6, wherein the assigned user profile is the user profile of the determined user.

8. The method according to at least one of claims 1 to 6, wherein the assigned user profile is the average user profile of a group of users.

9. The method according to at least one of claims 1 to 6, wherein the assigned user profile is a predetermined user profile selected by the determined user.

10. The method according to at least one of claims 1 to 9, wherein the user profile of a user is temporally variable and under loading taking into account the objects that the respective user has accessed.

11. The method according to at least one of claims 1 to 10, wherein a list of the relevant objects is created on the basis of their respective comparison values and is output to the user.

12. The method according to claim 11, wherein the order of the relevant objects in the list is determined on the basis of their respective comparison values.

13. The method according to at least one of claims 1 to 12, wherein the objects are electronic documents or parts of the content of electronic documents, in particular HTML, WTML or XTML documents, the respective contents of which are text, graphics, images, sound and / or language exhibit.

14. The method according to at least one of claims 2 to 13, wherein the search criterion has information about the type and / or content of objects, in particular in the form of text, graphics, images, sound and / or language.

15. The method according to claim 14, in which the similarity between the respective textual content of the objects and a search text is determined by a full-text analysis method, in particular using a similarity measure for character strings, and used to determine the relevance assessment of the individual objects and / or the preselection of objects becomes.

16. The method according to at least one of claims 1 to 15, wherein the user profiles and / or the object profiles are designed in the form of attribute-value pairs with predetermined attributes and determined values.

17. The method according to at least one of claims 1 to 15, wherein the user profile contains elements which each contain personal data of the user, in particular gender, age, profession and place of residence, and preference data which indicate a preferred use of certain objects in the form of areas of interest, especially sports, politics, music and entertainment.

18. The method according to at least one of claims 1 to 17, wherein the comparison of the determined user profile with the object profiles is carried out using a suitable metric which determines the degree of correspondence between the respective elements of the user profile and corresponding elements of the object profiles.

19. The method according to at least one of claims 10 to 18, wherein a dynamic part of the user profile adjusts to a changing usage behavior of the user, in particular when updating the user profile object profiles depending on the time the user accessed objects and / or in Dependent on the intensity of use of the objects by the user.

20. The method according to at least one of claims 3 to 19, wherein the relevance weighting of an object taking into account an assessment of the object by the users accessing the object and / or the intensity of use of the object by the users accessing the object.

21. The method according to at least one of claims 5 to 20, wherein the object profile is changed by the user profile of a user accessing the object, taking into account an assessment of the object by the user and / or the frequency and / or time of access to the object ,

22. The method according to at least one of claims 5 to 21, wherein the object profile is changed by the user profile of a user accessing the object, taking into account an intensity of use of the object by the user, in particular the number of times the user accesses the object , the chronological order, the duration of an access and the interaction of the user with the object are taken into account.

23. The method according to at least one of claims 5 to 22, wherein the object profile is determined using a cluster method by the user profiles of the users who have accessed the object.

24. Device for carrying out the method according to at least one of claims 1 to 23.