CN115423555A

CN115423555A - Commodity recommendation method and device, electronic equipment and storage medium

Info

Publication number: CN115423555A
Application number: CN202211048506.0A
Authority: CN
Inventors: 常德宝; 刘晓庆; 苗晨曦; 李书伟
Original assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Current assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Priority date: 2022-08-29
Filing date: 2022-08-29
Publication date: 2022-12-02

Abstract

The disclosure provides a commodity recommendation method, a commodity recommendation device, electronic equipment and a storage medium, and relates to the technical field of computers, in particular to the fields of natural language processing, big data and the like. The specific implementation scheme is as follows: constructing a user portrait according to the user basic information and the user historical behaviors; the user basic information is used for representing user personalized behaviors and demographic information; constructing a commodity portrait according to commodities related to a user, wherein the commodity portrait comprises a vector image for establishing an incidence relation between the commodities; and responding to the user request, and recommending the target commodity for the user according to the user portrait and the commodity portrait. By adopting the method and the device, the accuracy of personalized recommendation can be improved.

Description

Commodity recommendation method and device, electronic equipment and storage medium

Technical Field

The present disclosure relates to the field of computer technology, and more particularly, to the fields of natural language processing, big data, and the like.

Background

With the development of internet technology, more and more users browse their own interested commodities by using internet platforms or Applications (APP). For a new user, enough data support is lacked, so that commodities recommended for the new user are not personalized and the recommendation precision is low.

Disclosure of Invention

The disclosure provides a commodity recommendation method, an image construction device, an electronic device and a storage medium.

According to an aspect of the present disclosure, there is provided a commodity recommendation method including:

constructing a user portrait according to the user basic information and the user historical behaviors; the user basic information is used for representing user personalized behaviors and demographic information;

constructing a commodity portrait according to commodities related to a user, wherein the commodity portrait comprises a vector diagram for establishing an association relationship between the commodities;

and responding to a user request, and recommending a target commodity for the user according to the user portrait and the commodity portrait.

According to another aspect of the present disclosure, there is provided an image construction method, including:

and constructing a commodity portrait according to commodities related to a user, wherein the commodity portrait comprises a vector map for establishing an association relationship between the commodities.

According to another aspect of the present disclosure, there is provided an article recommendation apparatus including:

the first image construction module is used for constructing a user image according to the user basic information and the user historical behaviors; the user basic information is used for representing user personalized behaviors and demographic information;

the second portrait construction module is used for constructing commodity portrait according to commodities related to the user, wherein the commodity portrait comprises a vector image for establishing association relation among the commodities;

and the recommending module is used for responding to a user request and recommending target commodities for the user according to the user portrait and the commodity portrait.

According to another aspect of the present disclosure, there is provided a portrait construction apparatus including:

the user portrait construction module is used for constructing a user portrait according to the user basic information and the user historical behaviors; the user basic information is used for representing user personalized behaviors and demographic information;

the commodity portrait construction module is used for constructing a commodity portrait according to commodities related to a user, wherein the commodity portrait comprises a vector map for establishing association relations among the commodities.

According to another aspect of the present disclosure, there is provided an electronic device including:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,

the memory stores instructions executable by the at least one processor to cause the at least one processor to perform a method as provided by any one of the embodiments of the present disclosure.

According to another aspect of the present disclosure, there is provided a non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform a method provided by any one of the embodiments of the present disclosure.

According to another aspect of the present disclosure, there is provided a computer program product comprising computer instructions which, when executed by a processor, implement the method provided by any one of the embodiments of the present disclosure.

By adopting the method and the device, the accuracy of personalized recommendation can be improved.

It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present disclosure, nor do they limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description.

Drawings

The drawings are included to provide a better understanding of the present solution and are not to be construed as limiting the present disclosure. Wherein:

FIG. 1 is a diagram illustrating a merchandise vectorization recommendation in the related art;

FIG. 2 is a schematic diagram of another merchandise vectorization recommendation in the related art;

FIG. 3 is a schematic diagram of a distributed cluster processing scenario according to an embodiment of the present disclosure;

FIG. 4 is a flow chart diagram of a merchandise recommendation method according to an embodiment of the disclosure;

FIG. 5 is a schematic diagram of a merchandise vectorization recommendation, according to an embodiment of the present disclosure;

FIG. 6 is a diagram of a scenario for merchandise recommendation, according to an embodiment of the present disclosure;

FIG. 7 is a schematic flow diagram of a representation construction method in accordance with an embodiment of the present disclosure;

FIG. 8 is a schematic diagram of an item recommendation framework in an example of an application according to an embodiment of the present disclosure;

FIG. 9 is a schematic diagram of a sketch constructed based on multiple data sources in an application example according to an embodiment of the present disclosure;

FIG. 10 is a schematic diagram of clusters recommending similar users based on demographic information in an application example according to an embodiment of the present disclosure;

fig. 11 is a schematic diagram of a composition structure of a commodity recommending apparatus according to an embodiment of the present disclosure;

FIG. 12 is a schematic diagram of a component structure of a portrait construction apparatus according to an embodiment of the present disclosure;

FIG. 13 is a block diagram of an electronic device for implementing the merchandise recommendation method/representation construction method of the disclosed embodiments.

Detailed Description

Exemplary embodiments of the present disclosure are described below with reference to the accompanying drawings, in which various details of the embodiments of the disclosure are included to assist understanding, and which are to be considered as merely exemplary. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

The term "and/or" herein is merely an association describing an associated object, meaning that three relationships may exist, e.g., a and/or B, may mean: a exists alone, A and B exist simultaneously, and B exists alone. The term "at least one" herein means any combination of any one or more of a plurality, for example, including at least one of a, B, C, and may mean including any one or more elements selected from the group consisting of a, B, and C. The terms "first" and "second" used herein refer to and distinguish one from another in the similar art, without necessarily implying a sequence or order, or implying only two, such as first and second, to indicate that there are two types/two, first and second, and first and second may also be one or more.

Furthermore, in the following detailed description, numerous specific details are set forth in order to provide a better understanding of the present disclosure. It will be understood by those skilled in the art that the present disclosure may be practiced without some of these specific details. In some instances, methods, means, elements and circuits that are well known to those skilled in the art have not been described in detail so as not to obscure the present disclosure.

In the e-commerce industry, target commodities are recommended for users, a global hot commodity library is built mostly based on a recommendation mechanism of the global hot commodity library, namely based on historical click data of commodities, commodities matched with user requests are screened out from the global hot commodity library and serve as the target commodities; or, the target commodity is obtained through commodity vectorization recommendation based on 'graph walk'.

Considering that, in addition to the commodity recommendation for the old user, the commodity recommendation for the new user is also involved, in comparison, the number of the new user is higher and higher, the occupation ratio in the total user data is higher, but the behaviors in the station (such as an internet platform or APP) are sparse and low-frequency, for example, the duration of stay in the station is shorter, the path length of a behavior track is shorter, or the interval of access time (also called session) is longer, and the history data of the new user is less (for example, the number of the historical commodities of the new user is not higher, and the newly released commodities cannot be covered in the history data), which results in a problem that the recommendation precision is lower when the target commodity is recommended for the new user.

In the process of research, the inventor finds the following two aspects:

1) The historical behaviors of the new user are few, and the historical behaviors of other people cannot copy the personalized requirements of the new user;

2) The commodity vectorization recommendation of the 'graph walk' only records the behavior track of a user and considers a single commodity, and the association relation among a plurality of commodities is not established.

In the aspect 1), the personalized requirements of the new user may be considered, instead of being inferred by relying on the historical behaviors of the new user or the historical behaviors of other users, so that the recommendation accuracy of the target commodity may be achieved based on the personalized behaviors of the new user. Secondly, the personalized requirements of the new user can be considered, attributes which are fixed and unchangeable to the new user, namely demographic information (such as workplaces or industries and the like) can be considered, and compared with the historical behaviors of the new user, the demographic information reflecting the basic attributes of the new user can better reflect the personalized requirements which are in line with the new user.

As for the aspect 2), as shown in fig. 1, which is an example of "graph walk" commodity vectorization, a vector graph only records a track of a user behavior, and takes a commodity clicked by a user as a node in the vector graph, specifically, the commodity 101, the commodity 102, and the commodity 103 are all commodities clicked by the user a, a connecting line formed among the commodity 101, the commodity 102, and the commodity 103 is taken as an edge of the vector graph, and only a "collinear relationship" (that is: user a clicks on the three items, an edge is formed between the three items). As shown in fig. 2, which is another example of the "graph walk" product vectorization, the vector graph only records the track of the user behavior, specifically, the product 201, the product 202, and the product 203 are all the products clicked by the user a, the product 202, the product 203, and the product 204 are all the products clicked by the user B, and the connecting line between the product 201 and the product 204 is taken as the edge of the vector graph, and only records the "collinear relationship" (i.e., the user a and the user B both click three products and form an edge between the three products, where the product 202 and the product 203 are the products clicked by the user a and the user B), as shown in fig. 2: the vector diagram does not reflect the attributes of the edges, as shown in fig. 1, the edges of the vector diagram only record the behavior tracks of the user a, as shown in fig. 2, the edges of the vector diagram only record the behavior tracks of the user a and the user B, which are different, and the commodity database formed based on the vector diagram does not have an "association relationship" between multiple commodities (for example, multiple commodities belong to the same region, or multiple commodities belong to the same brand, etc.). In view of this, the incidence relation of multiple commodities is established, the attributes of the edges except for the nodes are added to the vector graph, the incidence relation between the commodities corresponding to the user, the incidence relation between the similar or similar commodities of different users and the like can be embodied by using the edges, and the recommendation accuracy of the commodity vectorization recommendation can be improved by reflecting the incidence relation through the attributes of the edges.

Fig. 3 is a schematic diagram of a distributed cluster processing scenario according to an embodiment of the present disclosure, where the distributed cluster system is an example of a cluster system, and exemplarily describes that commodity recommendation can be performed by using the distributed cluster system. As shown in fig. 3, in the distributed cluster system 300, a plurality of nodes (e.g., server cluster 301, server 302, server cluster 303, server 304, and server 305) are included, where the server 305 may further be connected to electronic devices, such as a cell phone 3051 and a desktop 3052, and the plurality of nodes and the connected electronic devices may jointly perform one or more related goods recommendation tasks. Optionally, a plurality of nodes in the distributed cluster system may perform commodity recommendation by using a data parallel relationship, and then the plurality of nodes may perform a commodity recommendation task based on the same processing logic, or the plurality of nodes in the distributed cluster system may perform a commodity recommendation task based on different processing logics (for example, some nodes perform processes such as portrait construction for commodity recommendation, some nodes perform commodity recommendation processing, some commodities perform processes such as merging, deduplication, intervention, and sorting for improving commodity recommendation accuracy, and these different processing logics may also perform combined processing according to business requirements). Optionally, after each round of training of the relationship extraction model is completed, data exchange (e.g., data synchronization) may be performed between multiple nodes.

According to an embodiment of the present disclosure, a method for recommending a commodity is provided, and fig. 4 is a schematic flowchart of the method for recommending a commodity according to the embodiment of the present disclosure, and the method may be applied to a commodity recommending apparatus, for example, the apparatus may be deployed in a situation where a terminal, a server, or other processing devices in a single-machine, multi-machine, or cluster system execute, and may implement processing such as commodity recommendation. The terminal may be a User Equipment (UE), a mobile device, a Personal Digital Assistant (PDA), a handheld device, a computing device, a vehicle-mounted device, a wearable device, or the like. In some possible implementations, the method may also be implemented by a processor calling computer readable instructions stored in a memory. As shown in fig. 4, the method is applied to any node or electronic device (e.g. desktop) in the cluster system shown in fig. 3, and includes:

s401, constructing a user portrait according to user basic information and user historical behaviors; the user basic information is used for representing the personalized behavior and the demographic information of the user.

S402, constructing a commodity portrait according to commodities related to a user, wherein the commodity portrait comprises a vector diagram for establishing association relations among the commodities.

S403, in response to the user request, recommending the target commodity for the user according to the user portrait and the commodity portrait.

In an example of S401-S403, the user personalized behavior may be a type of commodity, a brand of the commodity, a purchase intention of the commodity, etc. that the user pays attention to; the demographic information can be the workplace of the user (such as Beijing, shanghai, shenzhen and the like) and the industry of the user (such as the mechanical industry, the electric power industry, the chemical industry and the like), and based on the basic information of the user, the user portrait which is more in line with the personalized requirements of the user can be constructed by combining the historical behaviors of the user. The goods related to the user may be goods that are obtained by the user initiating a user request (such as a request generated by clicking a certain goods, or a request generated by searching a certain goods, etc.) before, and the goods include: according to the commodities directly fed back by the user request, and/or the recommendations related to the commodities, more accurate commodity images can be constructed based on the commodities related to the users, in addition, the attributes of edges are increased based on the commodity images constructed by the vector graph (as shown in figure 5), not only can the user behavior tracks be recorded by using the nodes of the vector graph, but also the association relation between the corresponding commodities of the users can be embodied by using the edges, the association relation between the same type or similar commodities of different users and the like can be reflected by the attributes of the edges, the recommendation accuracy of the commodity vectorization recommendation can be improved by reflecting the association relation, and finally, the user request can be responded, and more accurate target commodities can be recommended for the users according to the user images and the commodity images.

By adopting the embodiment of the disclosure, the user portrait can be constructed according to the user basic information (the user basic information is used for representing the personalized behaviors and the demographic information of the user) and the historical behaviors of the user, and the commodity portrait can be constructed according to the commodities related to the user (the commodity portrait comprises a vector diagram for establishing the association relationship between the commodities). Since the basic information of the user is not only the individual information unique to the user, but also the information such as demographic information (such as workplace and industry) is static, can be kept constant for a long time and can reflect the requirements of the user, the commodity portrait comprises a vector diagram establishing the association relationship between commodities except commodities clicked or searched by the user and the association between the commodities, and therefore, the target commodity is recommended to the user according to the user portrait and the commodity portrait in response to the request of the user, and the accuracy of individual recommendation can be improved.

In one embodiment, the method further comprises: acquiring high-level information of a user, and updating the user portrait according to the high-level information of the user; the user high-level information is used for representing a short-term access behavior of the user, which can also be called a short-term session behavior.

It should be noted that, in terms of session behaviors, colloquially, if an internet platform or APP is regarded as a big market, a user behaves like a customer scanning goods in the big market, for the market, the customer purchases a visit from entering a market to leaving the market, a series of behaviors in the middle buy the visit, and a corresponding behavior is called as the session behavior (or visit behavior). session behavior may have 5 elements (who, time, place, how, specific event) to describe one behavior of the user, namely: when, where and what specific events the user does, therefore, through the user behavior recorded by the session behavior, it can be determined when the user enters the internet platform or APP, when the user buys something, and so on.

In some examples, the high-level user information may be user preference information, the user preference information is classified as a short-term session behavior or a long-term session behavior, and the short-term session behavior is mainly targeted in consideration of personalized commodity recommendation for a new user.

In some examples, the short-term session behavior may include, but is not limited to: cookie information for jumping to the home page from the station (e.g., internet platform or APP).

By adopting the embodiment, the user data can be enriched through the user advanced information, so that the user portrait can be updated according to the user advanced information, and the purpose of recommending personalized commodities can be better realized.

In some embodiments, further comprising: obtaining a plurality of commodities according to an operation object corresponding to a user clicking operation or a user searching operation, determining similar or similar commodities to be associated from the commodities, and constructing a vector diagram according to the commodities and the association relation between the commodities to be associated. The nodes of the vector graph are used for representing a plurality of commodities, and the edges of the vector graph are used for representing the association relation between the commodities to be associated.

In some examples, the operation object may be a product clicked or indirectly obtained by a searched keyword.

In some examples, as shown in the vector diagram of fig. 5, the association relationships between multiple commodities clicked by multiple users (user 511, user 512, and user 513) and multiple commodities may be reflected (the association relationships are represented by weights, and the weight values corresponding to the association relationships may be the same or different). Wherein the plurality of merchandise includes: product 501, product 502, product 503, product 504, product 505, product 506, and product 507. The vector graph considers not only the nodes as the commodities but also the attributes of the edges, reflects the association relationship among a plurality of commodities through the attributes of the edges, such as whether the commodities belong to the same brand or belong to the same region, and constructs a commodity portrait (the association relationship among the commodities and the commodities) based on the association relationship, so that a recall path describing personalized demands can be established based on the commodity portrait.

Optionally, the nodes in the vector graph may include nodes serving as users in addition to nodes serving as commodities, and similarly, association between a user and a related commodity may also be established through edges between the nodes of the user and the nodes of the commodities (association may also be represented by using weights).

Optionally, the vector graph may also adopt a directed graph in addition to the undirected graph shown in fig. 5 (a vector graph with no direction at its edge is called an undirected graph), and the advantage of the directed graph compared with the undirected graph is that: besides recording the track of the user and the association relationship thereof, a certain time sequence also exists (for example, what commodity the user buys in what time period, or what commodity the user buys in what area/different areas of the same area, etc.), so that better and more accurate personalized recommended commodities can be screened out on the basis of a directed graph.

By adopting the embodiment, the relevance information among the commodities is established through the edges among the nodes of the vector graph, so that the commodity vector not only comprises the commodities but also comprises the relevance relation among the commodities, and after the vector graph is established according to the relevance relation, the commodity portrait obtained based on the vector graph is more accurate, and the recommendation precision of subsequent commodity recommendation is improved.

In one embodiment, in response to a user request, a target commodity is recommended for a user according to a user portrait and a commodity portrait, and the method comprises the following steps: and extracting information to be matched from the user request, and matching the information to be matched with a database obtained based on the user portrait and the commodity portrait to obtain a matching result. And obtaining a target commodity according to the matching result and the multi-dimensional recall strategy.

In some examples, obtaining the target product according to the matching result and the multi-dimensional recall strategy includes: and obtaining the target commodity according to the matching result and one of the multidimensional recall strategies. In other words, the database and the multi-dimensional recall strategy obtained based on the user portrait and the commodity portrait support at least one specific application scene, and the recall strategy is adopted for the corresponding scene in a targeted manner, so that the recall efficiency is improved.

In some examples, obtaining the target product according to the matching result and the multi-dimensional recall policy includes: and obtaining candidate commodities according to the matching result and the multi-dimensional recall strategy, performing at least one of merging, filtering, duplicate removal and intervention on the candidate commodities, then sequencing to obtain a commodity sequencing list, and taking at least one commodity sequenced in front in the commodity sequencing list as a target commodity. In other words, the database and the multi-dimensional recall strategy obtained based on the user portrait and the commodity portrait support at least one specific application scene, and the optimal target commodity is obtained by simultaneously adopting the multi-dimensional recall strategy, so that the recommendation results obtained from a plurality of scenes are combined, filtered, deduplicated and intervened in at least one mode, and finally sequenced to obtain the optimal target commodity.

By adopting the embodiment, after the information to be matched is extracted from the user request, the information to be matched and the database obtained based on the user portrait and the commodity portrait are matched to obtain the matching result, so that the accurate target commodity can be obtained according to the matching result and the multi-dimensional recall strategy.

In one embodiment, the multi-dimensional recall policy includes: the system comprises at least one of a commodity similar recall strategy based on user real-time search, a commodity similar recall strategy based on user historical search, a vector recall strategy based on user real-time commodity clicking, a vector similar recall strategy based on user real-time commodity clicking, a label recall strategy based on user real-time commodity clicking, a recall strategy based on user historical preference, a hot commodity recall strategy based on demographic information and a user vector recall strategy based on demographic information.

In some examples, as shown in fig. 6, in the multiple user requests, for example, a first user request may be a request triggered by a user click operation, an nth (N is an integer greater than 1) user request may be a request triggered by a user search operation, after information to be matched is extracted from any user request, matching processing may be performed on the information to be matched and a database obtained based on a user portrait and a commodity portrait, for example, taking a commodity portrait based on vectorization as an example, after a matching result is obtained, commodity recommendation may be performed according to the matching result and a multi-dimensional recall policy, so as to obtain an accurate target commodity.

More specifically, the above-described multidimensional recall strategy is described as follows:

for the commodity similarity recall strategy based on real-time search of the user, similarity recall can be performed based on a real-time search request of the user, for example, a search request based on a short-term behavior of a user session obtains categories, vectors and the like of commodities on line so as to recall commodities with similar contents;

for the similar commodity recall strategy based on the user history search, the recall can be performed based on the user history search request, for example, categories, vectors and the like of commodities are obtained on line based on a user history request list so as to recall commodities with similar contents;

for the vector recall policy of the user clicking the commodity in real time, the user can recall the commodity based on the Item Collaborative Filtering (ICF) of the commodity vector Item clicked in real time, for example, the commodity vector is obtained online and the commodity with similar content is recalled based on the clicked commodity of the user session short-term behavior. The ICF is a mode in collaborative filtering, and by analyzing the categories of commodities, commodities with high similarity are found and classified as a category of commodities, and then commodity recommendation is carried out;

for the vector similarity recall strategy of the user clicking the commodity in real time, the content vector similarity recall of the commodity can be clicked in real time by the user, for example, the commodity vector is acquired on line based on the clicked commodity of the user session short-term behavior, so as to recall the commodity with similar content;

for the label recall strategy of clicking the commodity in real time by the user, the recall based on the commodity label can be realized by clicking the commodity in real time by the user, for example, if the commodity label is established in advance, the commodity category corresponding to the commodity label is acquired on line based on the clicked commodity of the user session short-term behavior so as to recall the commodity with the same category;

in the recall strategy based on the historical preference of the user, the recall strategy can be based on the historical preference of the user, such as acquiring a commodity core word of a new user from a user characteristic positive bank so as to recall popular commodities which correspond to the historical preference of the user;

for the popular product recall strategy based on demographic information, a popular recall can be performed based on demographic information, for example, a User Authentication Service (UAS) based on online request obtains a field (field), such as a workplace and an industry of a User, to recall the popular product under each field;

for the user vector recall strategy based on demographic information, the user vector recall based on demographic information can be used, for example, the user vector under each field is obtained based on fields obtained by online request UAS, such as the workplace and industry of the user, and the commodities clicked by similar users are recalled.

By adopting the embodiment, at least one of the multidimensional recall strategies can be selected to realize recall according to the service scene of the user, and the method has diversity.

According to an embodiment of the present disclosure, a method for recommending a commodity is provided, and fig. 7 is a flowchart of a representation construction method according to an embodiment of the present disclosure, which may be applied to a representation construction apparatus, for example, the apparatus may be deployed in a terminal or a server or other processing devices in a single-machine, multi-machine or cluster system to implement commodity recommendation and the like. The terminal may be a User Equipment (UE), a mobile device, a Personal Digital Assistant (PDA), a handheld device, a computing device, a vehicle-mounted device, a wearable device, or the like. In some possible implementations, the method may also be implemented by a processor calling computer readable instructions stored in a memory. As shown in fig. 7, the method is applied to any node or electronic device (e.g. desktop) in the cluster system shown in fig. 3, and includes:

s701, constructing a user portrait according to user basic information and user historical behaviors; the user basic information is used for representing the personalized behaviors and the demographic information of the user.

In some examples, the user personalized behavior may be a commodity type, a commodity brand, a commodity purchase intention, and the like, which are focused on by the user; the demographic information can be the workplace of the user (such as Beijing, shanghai, shenzhen and the like) and the industry of the user (such as the mechanical industry, the electric power industry, the chemical industry and the like), and based on the basic information of the user, the user portrait which is more in line with the personalized requirements of the user can be constructed by combining the historical behaviors of the user.

S702, constructing a commodity portrait according to commodities related to a user, wherein the commodity portrait comprises a vector diagram for establishing association relations among the commodities.

In some examples, the items related to the user may be items that the user has previously initiated a user request (such as a request generated by clicking on a certain item, or a request generated by searching for a certain item, etc.), and the items include: the commodities and/or recommendations related to the commodities directly fed back according to the user request.

By adopting the embodiment of the disclosure, the user portrait can be constructed according to the user basic information (the user basic information is used for representing the user personalized behaviors and the demographic information) and the user historical behaviors, and the commodity portrait can be constructed according to the commodities related to the user (the commodity portrait comprises a vector diagram for establishing the association relation between the commodities). Because the basic information of the user is not only the individual information unique to the user, but also the information such as demographic information (such as workplace and industry) is static, can be kept constant for a long time and can reflect the requirements of the user better, the commodity portrait comprises a vector diagram establishing the association relationship between commodities except commodities clicked or searched by the user and the association between the commodities, therefore, the target commodity can be recommended to the user according to the database after the user request is responded based on the database obtained by the user portrait and the commodity portrait, and the accuracy of individual recommendation can be improved.

In one embodiment, the method further comprises: acquiring high-level information of a user, and updating the user portrait according to the high-level information of the user; wherein the user high-level information is used for characterizing the short-term access behavior of the user (e.g., the short-term access behavior may include cookie information for the user to jump from the intra-site page to the top page). By adopting the embodiment, the user portrait can be updated in time, so that the accuracy of personalized recommendation is improved. It is noted that the user representation includes: a user representation of a buyer user and a user representation of a seller user. The multi-type images of the buyer user and the seller user are obtained through global analysis of the users, and finally, the commodities are recommended to the buyer more accurately.

In one embodiment, the method further comprises: obtaining a plurality of commodities according to an operation object corresponding to a user clicking operation or a user searching operation, determining the same type or similar commodities to be associated from the commodities, and constructing a vector graph according to the commodities and the association relation between the commodities to be associated, wherein nodes of the vector graph are used for representing the commodities, and edges of the vector graph are used for representing the association relation between the commodities to be associated. By adopting the embodiment, the accuracy of personalized recommendation can be improved by utilizing the association relation.

In an embodiment, the user historical behavior further carries at least one piece of timestamp information, where the at least one piece of timestamp information is used to represent an execution time of the user historical behavior or a time sequence formed by the execution times of the user historical behavior. By adopting the embodiment, the historical behaviors of the user carry the time stamps, a behavior sequence for recording the historical behaviors can be formed, the time sequence of the historical behaviors is increased besides the historical behaviors, and the accuracy of personalized recommendation is improved.

In an application example of commodity recommendation, for personalized commodity recommendation of a new user, in addition to global popular commodity recommendation, that is, based on historical click data of commodities, a global popular commodity library is established, and personalized behaviors, demographic information and the like of the user are considered when the commodities are recommended. In the graph migration-based commodity vectorization recommendation, not only are commodities clicked or searched by a user recorded in history, but also the association relation between the commodities is recorded by each node in a vector graph, and then a commodity vector (comprising commodity attributes represented by the nodes and association attributes between the commodities represented by the edges) is obtained by graph migration, and commodities (such as required commodities, similar commodities of the required commodities and the like) can be recalled online on the basis of real-time clicking or searching of a new user, so that the recommendation precision is improved.

Fig. 8 is a block diagram of an overall implementation of the application example, which mainly includes 5 parts: recalling, establishing a library offline, storing in a classified mode, recalling online, combining and sequencing multiple data sources. The specific description is as follows:

1. recall of multiple data sources

The multi-path data source recall aims at expanding user coverage based on the multi-path data source, and improving the action path length of a user so as to improve the commodity recommendation precision of subsequent recall. The method mainly comprises the following steps: historical behavior, session behavior and demographic information of the user in the station (such as an internet platform or an APP), and a series of behaviors from the user entering the station to the user leaving the station is a session behavior. And analyzing the session behaviors, namely finding out a user session object with a specific behavior from all the user session behaviors so as to screen the user session behavior records. Wherein the specific behavior comprises: users who have searched for certain keywords, users who have visited within a certain time period, users who have been within a certain age, users who have been within a certain occupation, users who are in a certain city, etc. The demographic information may include, among other things: if the work place is 'Beijing', a user buys local goods in the Beijing, if the work place is 'Shanghai', the user visits a supermarket at the Shanghai to buy the goods; such as "mechanical industry" in industry, users buy books in mechanical industry, such as "electronic industry" in industry, users buy products in electronic industry, etc

FIG. 9 shows a picture database constructed based on multiple data sources, the picture database comprising: a database of merchandise representations and a database of user representations, wherein the database of merchandise representations may comprise: commodity vectors, commodity categories, commodities, etc.; the database of user images includes images of buyer users and images of seller users, and may be configured as independent databases, or integrated into a database, and in the case of integrating into a database, the database may include: user vectors, user behaviors, user characteristics, and the like.

2. Off-line building warehouse

The method is mainly used for realizing personalized recommended commodities by supporting an online strategy. Comprises two parts:

(1) Commodity recommendation library

A demographic information hot store, a preference tag hot store may be included.

(2) User behavior library and user feature library

May include a user history request list library, a user feature positive rank library, and a demographic information vector library.

3. Classified storage

The method mainly solves the problems of online acquisition, offline analysis and application expansion of other scenes of a commodity recommendation library, a user behavior library and a user characteristic library which are built. The overall storage mode is divided into the following 3 types:

1) For a commodity recommendation library and a demographic information vector library, the commodity recommendation library and the demographic information vector library can be stored in a database in a redis (database based on key value pairs) format, and the online acquired time response requirement can be met;

2) For the user history request list library, because the user history request list library is obtained by mining based on natural search results and commercial advertisements, the data volume is large, and the user history request list library is stored in the database in the ava format and can meet the requirement of storage capacity;

3) The user feature forward-ranking library can be stored in a database in an SNDB format, and can meet the stability requirement.

4. Online recall

The method mainly builds an individualized multidimensional recall strategy, recalls commodities based on individual historical behaviors and session short-term behaviors, and solves the problems of insufficient individuation based on global popular commodity recommendation and insufficient commodity coverage and insufficient display based on graph-walk commodity vectorization recommendation. The multidimensional recall strategy comprises the following contents:

(1) Similar recalls are carried out based on the real-time search request of the user, for example, based on the search request of the user session short-term behavior, categories, vectors and the like of commodities are obtained on line, and the commodities with similar contents are recalled.

(2) Recalling based on the user history search request, for example, obtaining categories, vectors and the like of the commodities on line based on the user history request list so as to recall the commodities with similar contents.

(3) And (4) calling back by clicking the commodity vector ICF in real time based on the user, for example, calling the commodity based on the short-term behavior of the user session, acquiring the commodity vector on line, and calling back the commodity with similar content.

(4) And (4) recalling the similar vectors of the commodity contents based on real-time clicking of the commodity by the user, for example, recalling the commodity with similar contents by acquiring the commodity vectors online based on the clicked commodity of the short-term behavior of the user session.

(5) The method comprises the steps of realizing the recall based on a commodity label by clicking the commodity in real time based on a user, for example, if the commodity label is established in advance, clicking the commodity based on the short-term behavior of the user session, and obtaining the commodity category corresponding to the commodity label on line so as to recall the commodity with the same category.

(6) And recalling the popular commodities corresponding to the historical preference of the user based on the historical preference recall of the user, such as acquiring the commodity core words of the new user from a user characteristic positive bank.

(7) Trending based on demographic information, for example, fields are obtained based on the UAS of the online request, such as the user's workplace and industry, to recall trending items under each of the above fields. Taking the commodity clicked by the user in the mechanical industry as an example, the commodity recommendation library in the mechanical industry is constructed in advance, and then, if the other users analyze the commodity recommendation library based on the demographic information, the other users: and the machine industry user also can recommend the user individually from the commodity recommendation library of the machine industry, so that the user can recall the machine industry user directly to obtain the required commodity.

(8) User vector recalls based on demographic information, such as fields based on online requests UAS, such as the user's workplace and industry, obtain a user vector under each field, recall items clicked on by similar users. As shown in fig. 10, a clustering process for recalling similar users is shown, the demographic information 100 includes a mechanical industry and a chemical industry, users in the mechanical industry and users in the chemical industry respectively construct user vectors, and perform clustering to obtain a cluster 1001 in the mechanical industry and a cluster 1003 in the chemical industry, where a central point of the cluster 1001 in the mechanical industry is a user vector 1002, a central point of the cluster 1003 in the chemical industry is a user vector 1004, and the user vectors in the respective industries are used to respectively find similar users nearest to the cluster 1001, so as to recall a commodity clicked by the similar user.

5. Merging and sorting

Mainly aims to meet the requirements of diversity and relevance, and recommends commodities which better meet the personalized requirements for users, thereby improving the user experience. Comprises 5 parts: the method is beneficial to a multi-dimensional recall strategy to obtain candidate commodities of each recall channel, combines the candidate commodities, filters low-quality and sensitive commodity categories and the like, removes the duplication of historical recommended commodities of a user, sequences the recall channels after intervention to obtain a recommendation list of target commodities, and preferentially recommends the target commodities ranked in the front to the user.

By adopting the application example, the user coverage is enlarged and the action path length of the user is improved through the recall of the multiple data sources; the method comprises the steps of establishing a library offline, and supporting an online strategy to realize personalized recommended commodities; by storing, the problems of online acquisition, offline analysis and application expansion of other scenes of the established commodity library, the user behavior library and the user characteristic library are solved; through online recall, the problems of insufficient personalization, insufficient commodity coverage and insufficient display of a new user are solved; by combining and sequencing, the improvement of diversity and relevance is realized, commodities which meet personalized requirements better are recommended to users, and the user experience is improved.

According to an embodiment of the present disclosure, there is provided a product recommendation device, fig. 11 is a schematic structural diagram of a component of the product recommendation device according to an embodiment of the present disclosure, and as shown in fig. 11, the product recommendation device includes: a first sketch constructing module 1101, configured to construct a user sketch according to user basic information and user historical behaviors; the user basic information is used for representing user personalized behaviors and demographic information; a second portrait construction module 1102, configured to construct a commodity portrait according to a commodity relevant to a user, where the commodity portrait includes a vector diagram for establishing an association relationship between commodities; and the recommending module 1103 is used for responding to a user request and recommending target commodities for the user according to the user portrait and the commodity portrait.

In one embodiment, the system further comprises an updating module, configured to obtain user high-level information, and update the user portrait according to the user high-level information; wherein the user high-level information is used to characterize a short-term access behavior of the user.

In one embodiment, the method further comprises: the operation module is used for obtaining a plurality of commodities according to an operation object corresponding to a user click operation or a user search operation; the determining module is used for determining the same type or similar commodities to be associated from the commodities; the graph construction module is used for constructing the vector graph according to the incidence relations among the commodities to be correlated; the nodes of the vector graph are used for representing the commodities, and the edges of the vector graph are used for representing the association relation among the commodities to be associated.

In one embodiment, the recommending module 1103 is configured to extract information to be matched from the user request; matching the information to be matched with a database obtained based on the user portrait and the commodity portrait to obtain a matching result; and obtaining the target commodity according to the matching result and the multi-dimensional recall strategy.

In an embodiment, the recommending module 1103 is configured to obtain the target product according to the matching result and one of the multidimensional recall policies.

In one embodiment, the recommending module 1103 is configured to obtain a candidate product according to the matching result and a multidimensional recall policy; performing at least one of merging, filtering, duplicate removal and intervention on the candidate commodities, and then sequencing to obtain a commodity sequencing list; and taking at least one commodity sequenced at the front in the commodity sequencing list as the target commodity.

According to an embodiment of the present disclosure, there is provided a portrait building apparatus, fig. 12 is a schematic structural diagram of a portrait building apparatus according to an embodiment of the present disclosure, and as shown in fig. 12, the portrait building apparatus includes: the user portrait construction module 1201 is used for constructing a user portrait according to user basic information and user historical behaviors; the user basic information is used for representing user personalized behaviors and demographic information; the merchandise portrait constructing module 1202 is configured to construct a merchandise portrait according to a merchandise related to a user, where the merchandise portrait includes a vector diagram for establishing an association relationship between the merchandise.

In one embodiment, the system further comprises a user portrait updating module, configured to obtain user high-level information, and update the user portrait according to the user high-level information; wherein the user high-level information is used to characterize the short-term access behavior of the user.

In one embodiment, the method further comprises: the operation object acquisition module is used for acquiring a plurality of commodities according to an operation object corresponding to a user click operation or a user search operation; the commodity determining module is used for determining similar or similar commodities to be associated from the commodities; the vector graph building module is used for building the vector graph according to the incidence relations among the commodities and the commodities to be correlated; the nodes of the vector graph are used for representing the commodities, and the edges of the vector graph are used for representing the association relation between the commodities to be associated.

In one embodiment, the user representation includes: a user representation of a buyer user and a user representation of a seller user.

In an embodiment, the historical user behavior further carries at least one piece of timestamp information, where the at least one piece of timestamp information is used to represent an execution time of the historical user behavior or a time sequence formed by the execution times of the historical user behavior.

In one embodiment, the short term access behavior includes access information for a user to jump from an intra-site page to a home page.

In the technical scheme of the disclosure, the acquisition, storage, application and the like of the personal information of the related user all accord with the regulations of related laws and regulations, and do not violate the customs of public sequences.

The present disclosure also provides an electronic device, a readable storage medium, and a computer program product according to embodiments of the present disclosure.

Fig. 13 shows a schematic block diagram of an example electronic device 1300 that can be used to implement embodiments of the present disclosure. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be examples only, and are not intended to limit implementations of the disclosure described and/or claimed herein.

As shown in fig. 13, the electronic device 1300 includes a computing unit 1301 that can perform various appropriate actions and processes according to a computer program stored in a Read Only Memory (ROM) 1302 or a computer program loaded from a storage unit 1308 into a Random Access Memory (RAM) 1303. In the RAM 1303, various programs and data necessary for the operation of the electronic apparatus 1300 can also be stored. The calculation unit 1301, the ROM 1302, and the RAM 1303 are connected to each other via a bus 1304. An input/output (I/O) interface 1305 is also connected to bus 1304.

A number of components in the electronic device 1300 are connected to the I/O interface 1305, including: an input unit 1306 such as a keyboard, a mouse, and the like; an output unit 1307 such as various types of displays, speakers, and the like; a storage unit 1308 such as a magnetic disk, optical disk, or the like; and a communication unit 1309 such as a network card, modem, wireless communication transceiver, etc. The communication unit 1309 allows the electronic device 1300 to exchange information/data with other devices through a computer network such as the internet and/or various telecommunication networks.

Computing unit 1301 may be a variety of general and/or special purpose processing components with processing and computing capabilities. Some examples of computing unit 1301 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various specialized Artificial Intelligence (AI) computing chips, various computing units running machine learning model algorithms, a Digital Signal Processor (DSP), and any suitable processor, controller, microcontroller, and so forth. The calculation unit 1301 performs the respective methods and processes described above, such as the goods recommendation method/portrait construction method. For example, in some embodiments, the merchandise recommendation method/representation construction method may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as storage unit 1308. In some embodiments, part or all of the computer program can be loaded and/or installed onto the electronic device 1300 via the ROM 1302 and/or the communication unit 1309. When the computer program is loaded into the RAM 1303 and executed by the computing unit 1301, one or more steps of the merchandise recommendation method/representation construction method described above may be performed. Alternatively, in other embodiments, computing unit 1301 may be configured in any other suitable manner (e.g., by way of firmware) to perform the merchandise recommendation method/representation construction method.

Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuitry, field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), system on a chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, receiving data and instructions from, and transmitting data and instructions to, a storage system, at least one input device, and at least one output device.

Program code for implementing the methods of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the program codes, when executed by the processor or controller, cause the functions/operations specified in the flowchart and/or block diagram to be performed. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which a user may provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic, speech, or tactile input.

The systems and techniques described here can be implemented in a computing system that includes a back-end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), wide Area Networks (WANs), and the Internet.

The computer system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server may be a cloud server, a server of a distributed system, or a server with a combined blockchain.

It should be understood that various forms of the flows shown above, reordering, adding or deleting steps, may be used. For example, the steps described in the present disclosure may be executed in parallel or sequentially or in different orders, and are not limited herein as long as the desired results of the technical solutions of the present disclosure can be achieved.

The above detailed description should not be construed as limiting the scope of the disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made in accordance with design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present disclosure should be included in the protection scope of the present disclosure.

Claims

1. A method of merchandise recommendation, comprising:

2. The method of claim 1, further comprising:

acquiring high-level information of a user, and updating the user portrait according to the high-level information of the user; wherein the user high-level information is used to characterize the short-term access behavior of the user.

3. The method of claim 2, further comprising:

obtaining a plurality of commodities according to an operation object corresponding to a user click operation or a user search operation;

determining the same type or similar commodities to be associated from the commodities;

constructing the vector diagram according to the incidence relation among the commodities to be correlated;

the nodes of the vector graph are used for representing the commodities, and the edges of the vector graph are used for representing the association relation between the commodities to be associated.

4. The method of any of claims 1-3, wherein said recommending, in response to a user request, a target good for a user based on the user representation and the good representation, comprises:

extracting information to be matched from the user request;

matching the information to be matched with a database obtained based on the user portrait and the commodity portrait to obtain a matching result;

and obtaining the target commodity according to the matching result and the multi-dimensional recall strategy.

5. The method of claim 4, wherein the obtaining the target product according to the matching result and a multidimensional recall strategy comprises:

and obtaining the target commodity according to the matching result and one of the multidimensional recall strategies.

6. The method of claim 4, wherein the obtaining the target product according to the matching result and a multidimensional recall strategy comprises:

obtaining candidate commodities according to the matching result and a multi-dimensional recall strategy;

performing at least one of merging, filtering, duplicate removal and intervention on the candidate commodities, and then sequencing to obtain a commodity sequencing list;

and taking at least one commodity which is sequenced at the front in the commodity sequencing list as the target commodity.

7. The method of claim 5 or 6, wherein the multi-dimensional recall policy comprises: the system comprises at least one of a commodity similar recall strategy based on user real-time search, a commodity similar recall strategy based on user historical search, a vector recall strategy based on user real-time commodity clicking, a vector similar recall strategy based on user real-time commodity clicking, a tag recall strategy based on user real-time commodity clicking, a recall strategy based on user historical preference, a hot commodity recall strategy based on demographic information and a user vector recall strategy based on demographic information.

8. An image construction method, comprising:

and constructing a commodity portrait according to commodities related to a user, wherein the commodity portrait comprises a vector image for establishing an association relation between the commodities.

9. The method of claim 8, further comprising:

10. The method of claim 8 or 9, further comprising:

constructing the vector graph according to the incidence relation among the commodities to be correlated;

11. The method of claim 8 or 9, wherein the user representation comprises: a user representation of a buyer user and a user representation of a seller user.

12. The method of claim 8, wherein the user historical behavior further carries at least one piece of timestamp information, and the at least one piece of timestamp information is used to characterize execution time of the user historical behavior or a time sequence formed by the execution time of the user historical behavior.

13. The method of claim 9, wherein the short term access behavior comprises access information for a user to jump from an intra-site page to a home page.

14. An article recommendation device comprising:

the first image construction module is used for constructing a user image according to the user basic information and the user historical behavior; the user basic information is used for representing user personalized behaviors and demographic information;

the second portrait construction module is used for constructing a commodity portrait according to commodities related to a user, wherein the commodity portrait comprises a vector diagram for establishing association relations among the commodities;

15. The apparatus of claim 14, further comprising an update module for obtaining user high level information, updating the user representation based on the user high level information; wherein the user high-level information is used to characterize the short-term access behavior of the user.

16. The apparatus of claim 15, further comprising:

the operation module is used for obtaining a plurality of commodities according to an operation object corresponding to a user click operation or a user search operation;

the determining module is used for determining similar or similar commodities to be associated from the commodities;

the graph building module is used for building the vector graph according to the incidence relations among the commodities to be correlated;

the nodes of the vector graph are used for representing the commodities, and the edges of the vector graph are used for representing the association relation among the commodities to be associated.

17. The apparatus of any of claims 14-16, wherein the recommendation module is to:

extracting information to be matched from the user request;

18. The apparatus of claim 17, wherein the recommendation module is to:

and obtaining the target commodity according to the matching result and one recall strategy in the multidimensional recall strategies.

19. The apparatus of claim 17, wherein the recommendation module is to:

20. The apparatus of claim 18 or 19, wherein the multidimensional recall policy comprises: the system comprises at least one of a commodity similar recall strategy based on user real-time search, a commodity similar recall strategy based on user historical search, a vector recall strategy based on user real-time commodity clicking, a vector similar recall strategy based on user real-time commodity clicking, a label recall strategy based on user real-time commodity clicking, a recall strategy based on user historical preference, a hot commodity recall strategy based on demographic information and a user vector recall strategy based on demographic information.

21. A sketch constructing apparatus, comprising:

the commodity portrait construction module is used for constructing a commodity portrait according to commodities related to a user, wherein the commodity portrait comprises a vector image for establishing an incidence relation between the commodities.

22. The apparatus of claim 21, further comprising a user profile update module to:

acquiring high-level information of a user, and updating the user portrait according to the high-level information of the user; wherein the user high-level information is used to characterize a short-term access behavior of the user.

23. The apparatus of claim 21 or 22, further comprising:

the operation object acquisition module is used for acquiring a plurality of commodities according to an operation object corresponding to user click operation or user search operation;

the commodity determining module is used for determining similar or similar commodities to be associated from the commodities;

the vector graph constructing module is used for constructing the vector graph according to the incidence relations among the commodities and the commodities to be associated;

24. The apparatus of claim 21 or 22, wherein the user representation comprises: a user representation of a buyer user and a user representation of a seller user.

25. The apparatus of claim 21, wherein the historical user behavior further carries at least one piece of timestamp information, and the at least one piece of timestamp information is used to characterize execution time of the historical user behavior or a time sequence formed by the execution time of the historical user behavior.

26. The apparatus of claim 22, wherein the short term access behavior comprises access information for a user to jump from an intra-site page to a home page.

27. An electronic device, comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein, the first and the second end of the pipe are connected with each other,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-13.

28. A non-transitory computer readable storage medium having stored thereon computer instructions for causing the computer to perform the method of any one of claims 1-13.

29. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any one of claims 1-13.