Data processing method in distribution service, medium, equipment and program product

Abstract

The invention discloses a data processing method, medium, equipment and program product in delivery service, and the method comprises the steps: firstly obtaining the address information, latitude and longitude information and Wi-Fi fingerprint information of the delivery places of a plurality of historical delivery orders; a graph network comprising a plurality of nodes is established based on the similarity among address information, longitude and latitude information and Wi-Fi fingerprint information of distribution places of different historical distribution orders, and clustering is carried out to obtain a plurality of initial class clusters, so that the distribution places can be described from three dimensions of addresses, longitudes and latitudes and Wi-Fi fingerprints, and the accuracy of the distribution places is improved. Therefore, the description information of the similar distribution places can be accurately clustered. And then positive samples, negative samples and weak positive samples are selected to adjust the plurality of initial class clusters, so that a plurality of class clusters with higher accuracy can be obtained. And then, determining a target class cluster from the plurality of class clusters, and adjusting the first description information of the delivery place of the current delivery order based on the second description information corresponding to the nodes in the target class cluster.

Description

Technical Field

[0001] The present disclosure relates to the field of Internet technology, and in particular to a data processing method, medium, device and program product in a distribution service. Background Art

[0002] Delivery service is an important and growing service in the home economy. In delivery service, users need to provide a description of the delivery location when submitting a delivery order. The delivery platform determines the delivery location based on the description and dispatches the delivery capacity to deliver the delivery order submitted by the user. However, the delivery information submitted by users is often not standardized, resulting in inaccurate delivery locations, which in turn causes a deviation between the actual arrival location of the delivery capacity and the delivery location. Summary of the invention

[0003] In a first aspect, an embodiment of the present disclosure provides a data processing method in a delivery service, the method comprising: obtaining first description information of a delivery location of a current delivery order; obtaining a plurality of pre-divided clusters, any cluster including at least one node; the plurality of clusters are divided based on the following manner: obtaining second description information of delivery locations of a plurality of historical delivery orders; the second description information includes address information, longitude and latitude information, and Wi-Fi fingerprint information of the delivery locations corresponding to the historical delivery orders; establishing a graph network based on the second description information of the delivery locations of the plurality of historical delivery orders, the graph network comprising a plurality of nodes, each node corresponding to the second description information of a delivery location of a historical delivery order, and the edges between the nodes are based on features between address information in the second description information of the delivery locations of different historical delivery orders. similarity, feature similarity of longitude and latitude information, and feature similarity between Wi-Fi fingerprint information; clustering the multiple nodes based on the graph network to obtain multiple initial clusters; for any node, taking the nodes in the same initial cluster as the node as the positive sample corresponding to the node, and selecting weak positive samples and negative samples from the nodes in different initial clusters from the node according to a preset probability, adjusting the multiple initial clusters based on the positive samples, weak positive samples, and negative samples to obtain multiple clusters; determining the target cluster to which the current delivery order belongs from the multiple clusters based on the feature similarity between the first description information and the second description information corresponding to the nodes in the multiple clusters; adjusting the first description information based on the second description information corresponding to the nodes in the target cluster.

[0004] In a second aspect, an embodiment of the present disclosure provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any embodiment of the present disclosure.

[0005] In a third aspect, an embodiment of the present disclosure provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in any embodiment of the present disclosure when executing the computer program.

[0006] In a fourth aspect, an embodiment of the present disclosure provides a computer program product, including a computer program, which implements the method described in any embodiment of the present disclosure when executed by a processor.

[0007] In the disclosed embodiment, multiple clusters can be determined in an offline manner. Specifically, the address information, longitude and latitude information, and Wi-Fi fingerprint information of the delivery locations of multiple historical delivery orders are first obtained, and a graph network including multiple nodes is established based on the similarity between the address information, longitude and latitude information, and Wi-Fi fingerprint information of the delivery locations of different historical delivery orders, and multiple initial clusters are clustered. In this way, the delivery location can be described from the three dimensions of address, longitude and latitude, and Wi-Fi fingerprint, so that the description information of similar delivery locations can be clustered more accurately. Then, positive samples, negative samples, and weak positive samples are selected to adjust the multiple initial clusters, so that multiple clusters with higher accuracy can be obtained. Then, the target cluster is determined from the above multiple clusters and the first description information of the delivery location of the current delivery order is adjusted based on the second description information corresponding to the nodes in the target cluster, which can improve the accuracy of the first description information, thereby enabling the delivery capacity to reach the delivery location more accurately.

[0008] It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The drawings herein are incorporated into the specification and constitute a part of the present disclosure. These drawings illustrate embodiments consistent with the present disclosure and are used to illustrate the technical solutions of the present disclosure together with the specification.

[0010] Figure 1A It is a schematic diagram of an application scenario of an embodiment of the present disclosure.

[0011] Figure 1B It is a schematic diagram of another application scenario of the present disclosure.

[0012] Figure 2 It is a flowchart of a data processing method in a delivery service according to an embodiment of the present disclosure.

[0013] Figure 3 It is a schematic diagram of multiple clusters of embodiments of the present disclosure.

[0014] Figure 4 It is a schematic diagram of the principle of the feature extraction model of the embodiment of the present disclosure.

[0015] Figure 5 Schematic diagram of positive samples, negative samples and weak positive samples according to an embodiment of the present disclosure.

[0016] Figure 6 It is an overall flow chart of an embodiment of the present disclosure.

[0017] Figure 7 is a schematic diagram of a computer device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0018] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0019] The terms used in this disclosure are only for the purpose of describing specific embodiments and are not intended to limit the disclosure. The singular forms of "a", "said" and "the" used in this disclosure and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items. In addition, the term "at least one" herein means any combination of at least two of any one or more of a plurality of.

[0020] It should be understood that although the terms first, second, third, etc. may be used in the present disclosure to describe various information, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present disclosure, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0021] In order to enable those skilled in the art to better understand the technical solutions in the embodiments of the present disclosure and to make the above-mentioned purposes, features and advantages of the embodiments of the present disclosure more obvious and understandable, the technical solutions in the embodiments of the present disclosure are further described in detail below in conjunction with the accompanying drawings.

[0022] In online delivery services, users need to provide a description of the delivery location when submitting a delivery order. The delivery platform determines the delivery location based on the description and dispatches the delivery capacity to deliver the delivery order submitted by the user. Figure 1A and Figure 1BThe application scenarios shown are respectively used to illustrate this. It can be understood that the application scenarios listed below are only exemplary and are not intended to limit the present disclosure.

[0023] exist Figure 1A In the application scenario shown, the service platform 106 can provide a delivery service, and the user 104 can submit an order (hereinafter referred to as placing an order) to the service platform 106 to purchase the goods sold by the merchant 102 on the service platform 106. The submitted order includes a description of the delivery location, which is used to describe the delivery destination of the goods in the order. After receiving the order submitted by the user 104, the service platform 106 can assign the order to the corresponding merchant 102, and assign the delivery capacity 108 to deliver the goods of the merchant 102. When delivering, the delivery capacity 108 can first go to the merchant 102 to take the goods to be delivered (this process is called picking up the order), and then deliver the goods to the corresponding delivery location according to the description of the delivery location in the order. In this application scenario, the service platform 106 can be an instant delivery platform, such as a takeaway platform, a medicine delivery platform, a fresh food delivery platform, or a flower delivery platform. Accordingly, the goods sold by the merchant 102 can be takeaways, medicines, fresh food, or flowers. It is understandable that this is just an example, and in actual application, the service platform 106 can also be other types of instant delivery platforms.

[0024] exist Figure 1B In the application scenario shown, the service platform can provide delivery services, and the user 104 can submit an order (hereinafter referred to as placing an order) to the service platform 106 to purchase the service provided by the service platform 106. The submitted order includes description information of the delivery location, which is used to describe the delivery location (in this application scenario, the location where the user 104 is located, referred to as the user location). After the service platform 106 receives the order submitted by the user 104, it can allocate the delivery capacity 108 to go to the user's location to pick up the object to be delivered (such as a person or an item), and then deliver the object to be delivered to the destination. In this application scenario, the service platform 106 can be an online car-hailing platform or an errand platform, etc. Accordingly, the service provided by the service platform 106 can be an online car-hailing service or an errand service, etc.

[0025] In the above application scenario, the delivery capacity 108 needs to go to the delivery location according to the description of the delivery location in the order. However, the delivery information submitted by the user is often not standardized. Different users may have different descriptions of the same delivery location, including postal address, entity name or order notes, which leads to the determined delivery location being inaccurate, resulting in the actual arrival location of the delivery capacity 108 being offset from the delivery location. The service platform 106 has to manually intervene through the dispatcher to ensure that the order is delivered on time. In order to improve the efficiency and accuracy of order delivery, the description information of the delivery location needs to be processed to improve the accuracy of the description information.

[0026] In online delivery scenarios, delivery locations are usually referred to as Points of Interest (POIs). The traditional method is to measure and compare the similarity between the user's description of POIs and the standardized description information in the existing POI database, and then select the description information with the highest similarity as the description information of the order POI. In the latest research work, a pre-trained model suitable for geographic text tasks is proposed based on the map-text multimodal architecture, which can complete the processing of POI description information in its downstream tasks. However, these methods have some common problems, namely, they have a strong dependence on large-scale POI annotation. First, the manual collection and annotation of standardized description information of POIs is time-consuming and laborious. Even for map service providers, it is difficult to build a POI database with comprehensive coverage and complete text descriptions and longitude and latitude positioning. Therefore, existing POI datasets often have a large amount of missing data. Secondly, due to the lack of regular manual review, the existing large-scale standardized POI datasets cannot timely reflect the changes of POIs in the real world. The lack of such POI datasets leads to a lack of large-scale annotated data for learning, which has also become the main bottleneck of existing technologies.

[0027] Based on this, the embodiment of the present disclosure provides a data processing method in a delivery service, which generates POI description information in an unsupervised manner based on a large amount of order data generated in an online delivery service. Through the method of the present disclosure, location service providers represented by online delivery can automatically and massively process POI description information, thereby improving the accuracy of the description information, while reducing the reliance on manual annotation, and improving the efficiency and accuracy of location services. For ease of description and understanding, the following is combined with Figure 1A The application scenario shown is used to illustrate the method of the embodiment of the present disclosure. Figure 2 , the method comprising: Step S12: Obtaining first description information of the delivery location of the current delivery order; Step S14: obtaining a plurality of pre-divided clusters, each of which includes at least one node; the plurality of clusters are divided based on the following method: Step S142: obtaining second description information of the delivery locations of multiple historical delivery orders; the second description information includes address information, longitude and latitude information, and Wi-Fi fingerprint information of the delivery locations corresponding to the historical delivery orders; Step S144: establishing a graph network based on the second description information of the delivery locations of the multiple historical delivery orders, the graph network comprising multiple nodes, each node corresponding to the second description information of the delivery location of a historical delivery order, and edges between nodes are determined based on feature similarity between address information in the second description information of the delivery locations of different historical delivery orders, feature similarity between longitude and latitude information, and feature similarity between Wi-Fi fingerprint information; Step S146: clustering the multiple nodes based on the graph network to obtain multiple initial clusters; Step S148: for any node, taking the nodes in the same initial cluster as the node as the positive samples corresponding to the node, and selecting weak positive samples and negative samples from the nodes in different initial clusters from the node according to a preset probability, and adjusting the multiple initial clusters based on the positive samples, weak positive samples and negative samples to obtain multiple clusters; Step S16: determining a target cluster to which the current delivery order belongs from the multiple clusters based on feature similarity between the first description information and the second description information corresponding to the nodes in the multiple clusters; Step S18: adjusting the first description information based on the second description information corresponding to the nodes in the target cluster.

[0028] The disclosed embodiment can be executed by the service platform 106. In the disclosed embodiment, the service platform 106 can pre-acquire the second description information of the delivery location of the historical delivery order, and cluster the second description information of the delivery location of the historical delivery order to obtain multiple initial clusters. Since the initial clusters are determined based on the feature similarity between the address information in the second description information of the delivery locations of different historical delivery orders, the feature similarity between the longitude and latitude information, and the feature similarity between the Wi-Fi fingerprint information, and the feature similarity between the second description information of the same delivery location is often higher than the feature similarity between the second description information of different delivery locations, the nodes corresponding to the second description information describing the same delivery location can be roughly divided into the same initial cluster in the above manner. After dividing multiple initial clusters, the initial clusters are adjusted by selecting positive samples, weak positive samples, and negative samples, which can further improve the accuracy of cluster division. After obtaining the first description information of the delivery location of the current delivery order, the target cluster to which the current delivery order belongs can be determined. The delivery location described by the second description information corresponding to the node in the target cluster and the delivery location described by the first description information can be considered to be the same or close delivery location, so that the first description information can be adjusted using the second description information corresponding to the node in the target cluster to improve the accuracy of the first description information, thereby enabling the delivery capacity to reach the delivery location more accurately. Considering the privacy of the user, the adjusted first description information disclosed in the present invention can be refined only to the building corresponding to the delivery location, rather than the room number or floor number.

[0029] In step S12, the first description information is used to describe the delivery location of the current delivery order, so that the delivery transport 108 can go to the delivery location according to the first description information. The delivery location can be the starting point or destination point for the delivery by the delivery transport 108. Figure 1A In the example shown, the delivery transport 108 delivers the goods to the delivery location, so the delivery location is the destination location for delivery by the delivery transport 108. Figure 1B In the example shown, the delivery capacity 108 goes to the delivery location to pick up the object to be delivered. For example, the delivery capacity is an online-hailing vehicle, and the object to be delivered is a passenger. At this time, the delivery location is the starting point for the delivery by the delivery capacity 108.

[0030] In some embodiments, the first description information may include address information, longitude and latitude information, and / or Wi-Fi fingerprint information of the delivery location of the current delivery order.

[0031] The address information may be text description information (text information for short), and may include multiple address segments, such as province, city, district, street, community, direction, etc. The address information may be reported to the service platform 106 through the client of the user 104. For example, a user interface may be provided on the client, and the user 104 may input and submit the address information on the user interface, and the client may upload the address information submitted by the user 104 to the service platform 106.

[0032] The longitude and latitude information may be coordinate information, including longitude and latitude, used to indicate the longitude and latitude of the delivery location. The longitude and latitude information may include first longitude and latitude information collected by a positioning module (such as a GPS module) on the client of the user 104 and reported to the service platform 106. In addition, the longitude and latitude information may also include second longitude and latitude information collected by the client of the delivery capacity 108 in the process of delivering the current delivery order. In the process of the delivery capacity 108 delivering the current delivery order, the client of the delivery capacity 108 may collect longitude and latitude information at a certain frequency to form a longitude and latitude sequence, i.e., the second longitude and latitude information. The first longitude and latitude information can reflect the longitude and latitude of the location of the user 104, but since the user 104 is usually in an indoor environment, the accuracy of the first longitude and latitude information is low. The second longitude and latitude information can reflect the longitude and latitude characteristics around the delivery location. Therefore, reporting the first longitude and latitude information and the second longitude and latitude information together to the service platform 106 can enable the service platform 106 to obtain richer longitude and latitude information to make up for the shortcomings caused by the low accuracy of the first longitude and latitude information.

[0033] The Wi-Fi fingerprint information is used to represent the Wi-Fi signal set around the delivery location, and may include the Basic Service Set Identifier (BSSID) and / or Service Set Identifier (SSID) of each Wi-Fi signal in the Wi-Fi signal set. The Wi-Fi fingerprint information may include the first Wi-Fi fingerprint information scanned and reported by the client of the user 104. In addition, the Wi-Fi fingerprint information may also include the second Wi-Fi fingerprint information collected by the client of the delivery capacity 108 during the delivery of the current delivery order. During the delivery of the current delivery order by the delivery capacity 108, the client of the delivery capacity 108 may collect Wi-Fi fingerprint information at a certain frequency (for example, every 2 minutes) to form a sequence of Wi-Fi fingerprint information, namely, the second Wi-Fi fingerprint information. The first Wi-Fi fingerprint information can reflect the Wi-Fi fingerprint information of the location of the user 104, but due to the short effective coverage range of the Wi-Fi signal, the accuracy of the first Wi-Fi fingerprint information may be low. The second Wi-Fi fingerprint information can reflect the Wi-Fi characteristics around the delivery location. Therefore, reporting the first Wi-Fi fingerprint information and the second Wi-Fi fingerprint information together to the service platform 106 can enable the service platform 106 to obtain richer Wi-Fi characteristics to make up for the deficiency caused by the low accuracy of the first Wi-Fi fingerprint information.

[0034] It can be understood that in the above embodiment, since the longitude and latitude information and the Wi-Fi fingerprint information each have certain deficiencies, the address information, the longitude and latitude information and the Wi-Fi fingerprint information are used together as the first description information. In other examples, only the address information can be used as the first description information, or the address information and the longitude and latitude information can be used as the first description information, or the address information and the Wi-Fi fingerprint information can be used as the first description information. The first description information can also include other information according to actual needs, which will not be listed here one by one.

[0035] In step S14, multiple clusters may be pre-divided, each of which includes at least one node. Each node corresponds to the second description information of the delivery location of a historical delivery order. In some embodiments, the feature similarity between the second description information corresponding to the nodes in the same cluster is greater than the feature similarity between the second description information corresponding to the nodes in different clusters. This step may be performed simultaneously with step S12, or before or after step S12, and the present disclosure does not limit the execution order of step S12 and step S14.

[0036] In some embodiments, a graph network can be established based on the feature similarity between the second description information of multiple historical delivery orders, and the graph network includes multiple nodes, and the multiple nodes respectively correspond to the second description information of the delivery locations of the multiple historical delivery orders, and the nodes can be connected by edges. Similar to the first description information, the second description information may include the address information, longitude and latitude information, and Wi-Fi fingerprint information of the delivery location of the historical delivery order. The methods for obtaining the above three types of information can be found in the first description information, which will not be repeated here.

[0037] In some embodiments, the nodes corresponding to the second description information whose feature similarity is greater than a preset similarity threshold are connected by an edge, and there is no edge between the nodes corresponding to the second description information whose feature similarity is less than or equal to the preset similarity threshold. For example, when both the first description information and the second description information include longitude and latitude information, the feature similarity may be the similarity between the longitude and latitude. Specifically, the distance between the delivery locations described by different second description information can be determined based on the longitude and latitude information in the second description information, and the distance may reflect the similarity between the longitude and latitude. If the distance between the delivery locations described by the two second description information is greater than the preset distance threshold, there is no edge between the nodes corresponding to the two second description information, otherwise, there is an edge between the nodes corresponding to the two second description information. In the example where both the first description information and the second description information include Wi-Fi fingerprint information, the feature similarity may also be the similarity of the Wi-Fi fingerprint information. When both the first description information and the second description information include address information, the feature similarity may also be the similarity between the address information of the delivery locations.

[0038] In some embodiments, the first description information and the second description information both include latitude and longitude information, Wi-Fi fingerprint information and address information. On this basis, the graph network can be established in the following manner: for any first node among the multiple nodes of the graph network and any second node among the multiple nodes that is different from the first node, the edge between the first node and the second node is determined based on the following manner: obtaining the first feature similarity between the address information corresponding to the first node and the address information corresponding to the second node, the second feature similarity between the longitude and latitude information corresponding to the first node and the longitude and latitude information corresponding to the second node, and the third feature similarity between the Wi-Fi fingerprint information corresponding to the first node and the Wi-Fi fingerprint information corresponding to the second node; determining the maximum similarity among the first feature similarity, the second feature similarity and the third feature similarity; if the maximum similarity is greater than the preset similarity threshold, determining the edge corresponding to the maximum similarity as the edge between the first node and the second node. If the maximum similarity is less than or equal to the preset similarity threshold, it is determined that there is no edge between the first node and the second node. Optionally, the above-mentioned various similarity thresholds can be 0 or a value greater than 0.

[0039] After determining the nodes and edges, the clusters can be preliminarily divided. Specifically, the nodes corresponding to the second description information in the same cluster are interconnected, and the nodes can be divided into clusters according to whether there are edges between the nodes. Figure 3 , the dots represent nodes, and the line segments between nodes represent edges. It can be seen that the white dots are connected by edges, so the white dots are connected; similarly, the black dots are connected by edges, so the black dots are also connected; the white dots and the black dots are not connected. Therefore, Figure 3 The illustrated embodiment includes two clusters, one including a cluster including black dots and the other including white dots.

[0040] The above embodiment first constructs a multimodal directed graph network ,in, Represents a node, Indicates the edge corresponding to the address information, Indicates the edge corresponding to the longitude and latitude information, The edges corresponding to the Wi-Fi fingerprint information can capture and utilize multiple information to enhance the expressiveness of the graph network. Then, the highest weight edge among the three edges, that is, the edge corresponding to the maximum similarity among the first feature similarity, the second feature similarity, and the third feature similarity, is retained to reduce the complexity of the network, avoid excessive redundant calculations, and keep the key information intact, thereby optimizing the use of computing resources.

[0041] The first feature similarity may be determined by the Levenshtein edit distance between the address information, the second feature similarity may be determined by the Haversine geographic distance between the longitude and latitude information, and the third feature similarity may be determined by the set overlap between the Wi-Fi signal sets. The set overlap is the ratio of the number of Wi-Fi signals in the intersection of two Wi-Fi signal sets to the number of Wi-Fi signals in the union of the two Wi-Fi signal sets.

[0042] In the above embodiment, the feature similarities between multiple types of information are simultaneously considered to determine the graph network, thereby clustering the second description information of the delivery locations of multiple historical delivery orders. This can effectively solve the problem of insufficient accuracy and reliability of the extracted features caused by insufficient accuracy of a single piece of information, thereby improving the robustness of the solution.

[0043] In some embodiments, the first feature similarity represents the similarity between the features of the address information in the second description information corresponding to different nodes. For each piece of address information, the address element parsing model can be used in advance to segment the address elements and filter out the remark information in the address information. Since the content in the remark information is usually irrelevant to the address, filtering out the remark information can reduce the interference of content irrelevant to the address on feature extraction and feature comparison. Then, a pre-trained large language model can be used to extract features from the address information after filtering out the remark information to obtain a feature embedding representation of the address information.

[0044] The second feature similarity represents the similarity between the features of the longitude and latitude information in the second description information corresponding to different nodes. The features of the longitude and latitude information can be extracted by a pre-trained feature extraction model, and the feature extraction model can be a masked language model (MLM), such as a BERT model. The feature extraction model can be trained based on the following method: obtaining a sample longitude and latitude sequence, wherein the sample longitude and latitude sequence includes multiple longitude and latitude information collected during the delivery of a sample delivery order by the delivery capacity; randomly masking at least one longitude and latitude information in the sample longitude and latitude sequence, performing feature extraction on the masked sample longitude and latitude sequence through the feature extraction model, and predicting the masked longitude and latitude information based on the extracted features; and training the feature extraction model based on the difference between the predicted longitude and latitude information and the masked longitude and latitude information.

[0045] like Figure 4 As shown, assuming that a sample longitude and latitude sequence is {(lng 1 ,lat 1 ), (lng 2 ,lat 2), (lng 3 ,lat 3 ), (lng 4 ,lat 4 ), ...}, where lng represents longitude and lat represents latitude. At least one longitude and latitude information in the above sequence can be randomly masked, for example, (lng 2 ,lat 2 ), the feature extraction model is used to extract the features of the masked sample longitude and latitude sequence, and the masked longitude and latitude information is predicted based on the extracted features, which is recorded as (lng p ,lat p Then, based on (lng 2 ,lat 2 ) and (lng p ,lat p ) to train the feature extraction model. The trained feature extraction model can be used to extract features from longitude and latitude information to obtain feature embedding representation of longitude and latitude information.

[0046] Compared with text or images, latitude and longitude coordinates have more limited semantic richness and lower dimensionality. This limitation hinders their application in deep learning models in location-based services. However, the inventors found that although there are challenges in identifying the semantic similarity between two latitude and longitude coordinates, the spatial association of latitude and longitude coordinates can be used to extract features from latitude and longitude information.

[0047] In the related art, masked language models are often used to extract features from natural language (such as a paragraph), which allows the representation of words to be learned by capturing bidirectional contextual semantic correlations in sentences. Just as there is semantic coherence between words in a sentence, there is a natural continuity between longitude and latitude points on any road on a map. Therefore, in a similar way to natural language processing, the continuous longitude and latitude values ​​on the map can be discretized into a grid of a certain length, and each longitude and latitude coordinate is regarded as a "word". Each series of longitude and latitude coordinates from a delivery route can be considered as a "sentence". By randomly masking the longitude and latitude coordinates in each route, the pre-trained language model BERT is fine-tuned to learn the feature embedding representation of each longitude and latitude coordinate. By masking certain longitude and latitude coordinates and letting the model predict them, the model is able to capture the spatial relationship between longitude and latitude coordinates, thereby learning the intrinsic structure of geographic information. This approach is expected to improve the accuracy and efficiency of location-based services, especially in tasks that require understanding and predicting geographic location patterns.

[0048] Furthermore, the multiple initial clusters divided in the above manner may not be accurate enough, so the multiple initial clusters may be further adjusted. In the present disclosure, the initial clusters are adjusted by contrastive learning.

[0049] Specifically, first, for any node, the nodes in the initial cluster where the node is located can be used as positive samples of the node, and negative samples and weak positive samples of the node can be selected from initial clusters other than the initial cluster where the node is located.

[0050] See also Figure 5 , assuming that there are two clusters, cluster A and cluster B, cluster A includes nodes X1, X2, and X3, and cluster B includes nodes X4, X5, and X6. Then for the current node X1 (the node that needs to determine positive samples, negative samples, and weak positive samples), nodes X2 and X3 in cluster A with node X1 are both positive samples, and each node in cluster B can be determined as a negative sample or a weak positive sample of node X1. For example, node X4 can be used as a weak positive sample, and nodes X5 and X6 can be used as negative samples.

[0051] In some embodiments, for any first node and a second node in a different initial cluster from the first node, the preset probability that the second node is selected as a weak positive sample corresponding to the first node is positively correlated with the feature similarity between the second node and the first node, and the preset probability that the second node is selected as a negative sample corresponding to the first node is negatively correlated with the feature similarity between the second node and the first node. The feature similarity can be determined by the distance between the two nodes in the feature space, and the above distance can be a normalized distance.

[0052] Specifically, the probability distribution of selecting negative samples And the probability distribution of selecting weak positive samples They can be recorded as:

[0053]

[0054] in, represents the first node, represents the second node, express and The distance between represents the initial cluster where the first node is located, Represents other nodes in the initial cluster where the first node is located. By adopting the above means, the second node with higher feature similarity to the first node can be selected as a weak positive sample of the first node, and the second node with lower feature similarity to the first node can be selected as a negative sample of the first node.

[0055] In some embodiments, a loss function can be established based on positive samples, weak positive samples and negative samples; wherein the loss function is anti-correlated with the feature similarity between the second description information corresponding to the node and the second description information corresponding to the positive sample of the node, anti-correlated with the feature similarity between the second description information corresponding to the node and the second description information corresponding to the weak positive sample of the node, and positively correlated with the feature similarity between the second description information corresponding to the node and the second description information corresponding to the negative sample of the node. After establishing the loss function, the features of the second description information in the multiple initial clusters can be adjusted based on the loss function. The above loss function can be a QuadraInfoNCE loss function. Exemplarily, the loss function can be expressed as:

[0056] in, Indicates the current node. express The positive sample of express Weak positive samples of represents the nodes in other initial clusters other than the initial cluster where the current node is located, N represents the total number of nodes in other initial clusters other than the initial cluster where the current node is located, express The total number of positive samples, express The total number of weak positive samples.

[0057] After adjusting the features of the second description information in the multiple initial clusters, the adjusted second description information in the multiple initial clusters can be clustered, for example, by re-clustering the adjusted second description information in the multiple clusters using K-Means clustering, hierarchical clustering, or DBSCAN clustering. Feature adjustment and re-clustering can be performed in the above manner, and after several iterations, if the features of the second description information corresponding to each node and the clustering results converge, clustering is stopped, and the multiple clusters obtained at this time (i.e., the multiple clusters obtained in the last iteration) are determined as the final clusters and used in subsequent steps.

[0058] In step S16, the target cluster to which the current delivery order belongs may be determined, that is, the target cluster to which the node corresponding to the first description information of the delivery location of the current delivery order belongs may be determined.

[0059] When both the first description information and the second description information include address information, longitude and latitude information, and Wi-Fi fingerprint information, a first feature similarity between the address information included in the first description information and the address information included in the second description information corresponding to the nodes in the multiple clusters, a second feature similarity between the longitude and latitude information included in the first description information and the longitude and latitude information included in the second description information corresponding to the nodes in the multiple clusters, and a third feature similarity between the Wi-Fi fingerprint information included in the first description information and the Wi-Fi fingerprint information included in the second description information corresponding to the nodes in the multiple clusters can be obtained, and based on the first feature similarity, the second feature similarity and the third feature similarity, the target cluster to which the current delivery order belongs is determined from the multiple clusters.

[0060] For example, for any second description information, the first feature similarity between the address information in the first description information and the address information in the second description information, the second feature similarity between the longitude and latitude information in the first description information and the longitude and latitude information in the second description information, and the third feature similarity between the Wi-Fi fingerprint information in the first description information and the Wi-Fi fingerprint information in the second description information may be weighted to obtain the feature similarity between the first description information and the second description information. Alternatively, the maximum similarity among the first feature similarity between the address information in the first description information and the address information in the second description information, the second feature similarity between the longitude and latitude information in the first description information and the longitude and latitude information in the second description information, and the third feature similarity between the Wi-Fi fingerprint information in the first description information and the Wi-Fi fingerprint information in the second description information may be determined, and the maximum similarity may be determined as the feature similarity between the first description information and the second description information. The above method can be used to determine the feature similarity between the first description information and the second description information corresponding to each node, and then, for the second description information corresponding to the nodes in the same cluster, the feature similarity between the first description information and the second description information corresponding to each node in the cluster is weighted to obtain the total similarity between the first description information and the second description information corresponding to the nodes in the cluster. If the total similarity between the first description information and the second description information corresponding to the nodes in a cluster (assuming cluster A) is greater than the total similarity between the first description information and the second description information corresponding to the nodes in other clusters, cluster A is determined to be the target cluster to which the current delivery order belongs.

[0061] In step S18, the first description information may be adjusted based on the second description information in the target cluster.

[0062] In the case where each second description information corresponds to a node in the graph network, and the nodes corresponding to the second description information whose feature similarity is greater than a preset similarity threshold are connected by edges, a target node set can be determined, wherein the target node set includes the target node with the largest in-degree in the target cluster and neighboring nodes whose distance to the target node is less than a preset distance threshold, and the first description information is adjusted based on the second description information corresponding to each node in the target node set.

[0063] The in-degree of a node refers to the number of edges pointing to the node. If a node has the largest in-degree, it means that the node receives the most connections from other nodes, which usually indicates that the node plays a central or hub role in the network and has a high influence or importance. Therefore, the node can be determined as a target node (also called a core node). The neighboring nodes whose distance to the target node is less than the preset distance threshold usually also have a high influence or importance. Therefore, the first description information can be adjusted in combination with the target node and its neighboring nodes.

[0064] For example, when both the first description information and the second description information include longitude and latitude information, the longitude and latitude information included in the second description information corresponding to each node in the target node set can be weighted to obtain the adjusted longitude and latitude information included in the first description information. Alternatively, the sum of the distances between the longitude and latitude information included in the first description information and the longitude and latitude information included in each second description information in the target node set can be minimized as a goal, and the longitude and latitude information included in the first description information is adjusted based on the optimization goal. The adjusted longitude and latitude information included in the first description information can be recorded as:

[0065] in, Indicates the longitude and latitude information included in the second description information corresponding to the i-th node in the target node set, Indicates the longitude and latitude information included in the first description information.

[0066] For another example, when both the first description information and the second description information include address information, and the address information includes description information corresponding to multiple address segments, for any one of the multiple address segments, the occurrence frequency of the description information corresponding to the address segment can be determined from the second description information corresponding to each node in the target node set, and the description information corresponding to the address segment with the highest occurrence frequency is determined as the description information corresponding to the address segment. For example, assuming that the address segment includes an address segment corresponding to "district" and an address segment corresponding to "street", in the second description information corresponding to each node in the target node set, for the address segment corresponding to "district", the occurrence frequency of M1 district is 5 times, the occurrence frequency of M2 district is 3 times, and the occurrence frequency of M3 district is 8 times, then M3 district can be determined as the description information corresponding to the "district" address segment. Similarly, for the address segment corresponding to "street", assuming that the occurrence frequency of B1 street is 6 times, the occurrence frequency of B2 street is 10 times, and the occurrence frequency of B3 street is 2 times, then B2 street can be determined as the description information corresponding to the "street" address segment. Therefore, the first description information is M3 District B2 Street.

[0067] Figure 6 The overall process of the embodiment of the present disclosure is shown, including the following steps: Step 1, feature extraction of POI text description (i.e., address information) and latitude and longitude description (i.e., latitude and longitude information). For any delivery location in the current delivery order (hereinafter referred to as the query POI), extract the historical records (i.e., second description information) of POI descriptions in the same area (e.g., in the same geographical administrative region) as the query POI. For the text description of the POI in each record, the address element parsing model (e.g., natural language processing model) is used in advance to segment the address elements and filter out the remark information. Then, a pre-trained large language model is used to extract its feature embedding representation. Specifically, when the processing language is Chinese, a pre-trained ChineseBERT model can be used. Based on the observation that the delivery capacity 108 will collect a series of longitude and latitude sequences including a series of longitude and latitude information during delivery, the masked language model can be referred to to extract spatial features of the POI longitude and latitude information according to the longitude and latitude sequence. Specifically, for each delivery trajectory, some longitude and latitude information in the trajectory can be randomly masked, and a pre-trained BERT model can be fine-tuned with the goal of predicting the longitude and latitude information corresponding to the mask. Then, the fine-tuned BERT model is used to extract feature embedding representations of longitude and latitude information.

[0068] Step 2, nearest neighbor clustering of POI multimodal information. The present disclosure measures the multimodal distance between the second description information in the historical delivery orders, including the Levenshtein edit distance between text descriptions, the Haversine geographic distance between longitude and latitude descriptions, and the set overlap between Wi-Fi signal sets, that is, the ratio of the intersection and union size of two Wi-Fi signal sets. Then, taking the second description information as a node, multi-type edges between nodes are constructed according to the multimodal distance, thereby constructing a multimodal directed graph network. , represents a text edge, whose weight corresponds to the first feature similarity between nodes; Represents geographic longitude and latitude edges, whose weights correspond to the similarity of the second feature between nodes; represents the Wi-Fi edge, and its weight corresponds to the third feature similarity between nodes. Select the nearest neighbor subgraph of graph G , for each node, only the highest weight edges (i.e. the edges corresponding to the maximum similarity among the first feature similarity, the second feature similarity and the third feature similarity) of its text edge, geographic edge and Wi-Fi edge are retained. The second description information is clustered by nearest neighbors based on the directed strongly connected piece with the highest weight edge to form multiple initial clusters.

[0069] Step 3, use contrastive learning with weak positive samples to iteratively optimize the multiple initial clusters obtained in step 2. In this step, the multiple initial clusters obtained in step 2 can be used as pseudo labels for the second description information to perform contrastive learning on the second description information. , select all nodes belonging to the same initial cluster as positive samples for comparative learning, and select all nodes belonging to different initial clusters with probability distribution Select several negative samples for contrastive learning and distribute them among all nodes belonging to different initial clusters with probability distribution Select several weak positive samples for contrastive learning. Then, based on the InfoNCE loss function InfoNCE, the QuadraInfoNCE loss function is proposed to support the distance between weak positive samples and x in the feature space. Specifically, . Under the guidance of this loss function, it is proposed to update the feature embedding representation of all nodes during the training process with each node as the anchor point for comparative learning. During the training process, the updated feature embedding representation is used for nearest neighbor clustering every several training rounds until the feature embedding representation and clustering results converge. The target cluster can be determined from the nearest neighbor clustering results of the last round as the basis for adjusting the first description information. For the description information of the query POI (i.e., the first description information), all nodes belonging to the same target cluster are matched with it.

[0070] Step 4, query the selection of the core node set (i.e., the target node set) in the target cluster. First, for all nodes in the target cluster, define their node density in the directed graph network G as their node in-degree. Then, sort all nodes in the target cluster according to the node density, and select the node with the highest node density as the core node in the cluster (i.e., the target node in the aforementioned embodiment). Thus, it can be determined that the target node set to be used to generate the final address information and longitude and latitude information is the union of the core node and its one neighbor node set.

[0071] Step 5: Unification of longitude and latitude information and standardization of address information. Based on the target node set, it is planned to generate unified longitude and latitude information and standardized address information of the query POI. For the generation of longitude and latitude information, the following optimization goals are formulated to determine the optimal longitude and latitude: . Combining the pre-set standard address format and user description habits, for each address element, the description information with the highest frequency of use in all address information of the target node set is selected according to the "majority principle" as the standardized representation of the address segment. Finally, the description information of all address segments is spliced ​​in the order of the standard address format as the standardized address information for querying POIs.

[0072] An embodiment of the present disclosure further provides a computer device, which at least includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method described in any of the above embodiments is implemented.

[0073] Figure 7 A more specific hardware structure diagram of a computer device provided by an embodiment of the present disclosure is shown, and the device may include: a processor 202, a memory 204, an input / output interface 206, a communication interface 208, and a bus 210. The processor 202, the memory 204, the input / output interface 206, and the communication interface 208 are connected to each other through the bus 210 in the device.

[0074] The processor 202 can be implemented by a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, etc., and is used to execute relevant programs to implement the technical solutions provided by the embodiments of the present disclosure. The processor 202 can also include a graphics card, which can be a Nvidia titan X graphics card or a 1080Ti graphics card, etc.

[0075] The memory 204 may be implemented in the form of a read-only memory (ROM), a random access memory (RAM), a static storage device, a dynamic storage device, etc. The memory 204 may store an operating system and other application programs. When the technical solution provided by the embodiment of the present disclosure is implemented by software or firmware, the relevant program code is stored in the memory 204 and is called and executed by the processor 202.

[0076] The input / output interface 206 is used to connect the input / output module to realize information input and output. The input / output module can be configured in the device as a component (not shown in the figure), or it can be externally connected to the device to provide corresponding functions. The input device may include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device may include a display, a speaker, a vibrator, an indicator light, etc.

[0077] The communication interface 208 is used to connect a communication module (not shown) to realize communication interaction between the device and other devices. The communication module can realize communication through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, Wi-Fi, Bluetooth, etc.).

[0078] The bus 210 comprises a pathway for transmitting information between the various components of the device, such as the processor 202 , the memory 204 , the input / output interface 206 , and the communication interface 208 .

[0079] It should be noted that, although the above device only shows the processor 202, the memory 204, the input / output interface 206, the communication interface 208 and the bus 210, in the specific implementation process, the device may also include other components necessary for normal operation. In addition, it can be understood by those skilled in the art that the above device may also only include the components necessary for implementing the embodiments of the present disclosure, and does not necessarily include all the components shown in the figure.

[0080] An embodiment of the present disclosure provides a computer program product, including a computer program. When the computer program is executed by a processor, the method described in any embodiment of the present disclosure is implemented.

[0081] The present disclosure also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the method described in any of the above embodiments is implemented.

[0082] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computer device. As defined in this article, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0083] Each embodiment in the present disclosure is described in a progressive manner, and the same and similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. The device embodiment described above is merely schematic, wherein the modules described as separate components may or may not be physically separated, and the functions of each module can be implemented in the same one or more software and / or hardware when implementing the embodiment of the present disclosure. It is also possible to select some or all of the modules according to actual needs to achieve the purpose of the embodiment. A person of ordinary skill in the art can understand and implement it without paying creative labor.

[0084] The above is only a specific implementation of the embodiment of the present disclosure. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the embodiment of the present disclosure. These improvements and modifications should also be regarded as the protection scope of the embodiment of the present disclosure.

Claims

1. A data processing method in a delivery service, the method comprising: Get the first description information of the delivery location of the current delivery order; A plurality of pre-divided clusters are obtained, each of which includes at least one node; the plurality of clusters are divided based on the following method: Acquire second description information of the delivery locations of the plurality of historical delivery orders; the second description information includes address information, longitude and latitude information, and Wi-Fi fingerprint information of the delivery locations corresponding to the historical delivery orders; Establishing a graph network based on the second description information of the delivery locations of the multiple historical delivery orders, the graph network comprising multiple nodes, each node corresponding to the second description information of the delivery location of a historical delivery order, and edges between the nodes are determined based on feature similarity between address information in the second description information of the delivery locations of different historical delivery orders, feature similarity between longitude and latitude information, and feature similarity between Wi-Fi fingerprint information; Clustering the multiple nodes based on the graph network to obtain multiple initial clusters; For any node, the nodes in the same initial cluster as the node are used as positive samples corresponding to the node, and weak positive samples and negative samples are selected from nodes in different initial clusters from the node according to preset probabilities, and the multiple initial clusters are adjusted based on the positive samples, weak positive samples and negative samples to obtain multiple clusters; Determining a target cluster to which the current delivery order belongs from the multiple clusters based on feature similarities between the first description information and the second description information corresponding to the nodes in the multiple clusters; The first description information is adjusted based on the second description information in the target cluster.

2. According to the method of claim 1, the first description information includes address information, longitude and latitude information and Wi-Fi fingerprint information of the delivery location of the current delivery order.

3. The method according to claim 2, wherein the address information included in the first description information is reported by a client of the user who submitted the current delivery order; The longitude and latitude information included in the first description information includes the first longitude and latitude information collected when the user's client submits the current delivery order, and the second longitude and latitude information collected by the client of the delivery capacity that delivers the current delivery order in the process of delivering the current delivery order; The Wi-Fi fingerprint information included in the first description information includes first Wi-Fi fingerprint information collected when the user's client submits the current delivery order, and second Wi-Fi fingerprint information collected by the client of the delivery transportation capacity that delivers the current delivery order during the process of delivering the current delivery order.

4. The method according to claim 3, wherein determining the target cluster to which the current delivery order belongs from the multiple clusters based on the feature similarity between the first description information and the second description information corresponding to the nodes in the multiple clusters comprises: For any node, obtaining a first feature similarity between address information included in the first description information and address information included in second description information corresponding to the node, a second feature similarity between longitude and latitude information included in the first description information and longitude and latitude information included in the second description information corresponding to the node, and a third feature similarity between Wi-Fi fingerprint information included in the first description information and Wi-Fi fingerprint information included in the second description information corresponding to the node; Based on the first feature similarity, the second feature similarity, and the third feature similarity, a target cluster to which the current delivery order belongs is determined from the multiple clusters.

5. According to the method of claim 4, the second feature similarity is used to characterize the similarity between the features of the longitude and latitude information included in different second description information, and the features of the longitude and latitude information are extracted by a pre-trained feature extraction model, and the feature extraction model is trained based on the following method: Acquire a sample longitude and latitude sequence, wherein the sample longitude and latitude sequence includes a plurality of longitude and latitude information collected during the process of delivery of the sample delivery order by the delivery transport capacity; Randomly masking at least one longitude and latitude information in the sample longitude and latitude sequence, extracting features from the masked sample longitude and latitude sequence using the feature extraction model, and predicting the masked longitude and latitude information based on the extracted features; The feature extraction model is trained based on the difference between the predicted longitude and latitude information and the masked longitude and latitude information.

6. The method according to claim 1, wherein the step of establishing a graph network based on the second description information of the delivery locations of the plurality of historical delivery orders comprises: For any first node among the plurality of nodes of the graph network and any second node among the plurality of nodes that is different from the first node, an edge between the first node and the second node is determined based on the following method: Obtaining a first feature similarity between address information corresponding to the first node and address information corresponding to the second node, a second feature similarity between longitude and latitude information corresponding to the first node and longitude and latitude information corresponding to the second node, and a third feature similarity between Wi-Fi fingerprint information corresponding to the first node and Wi-Fi fingerprint information corresponding to the second node; Determine a maximum similarity among the first feature similarity, the second feature similarity, and the third feature similarity; If the maximum similarity is greater than a preset similarity threshold, the edge corresponding to the maximum similarity is determined as the edge between the first node and the second node.

7. According to the method described in claim 1, for any first node and a second node in a different initial cluster from the first node, the preset probability that the second node is selected as a weak positive sample corresponding to the first node is positively correlated with the feature similarity between the second node and the first node, and the preset probability that the second node is selected as a negative sample corresponding to the first node is negatively correlated with the feature similarity between the second node and the first node.

8. The method according to claim 1, wherein adjusting the plurality of initial clusters based on the positive samples, weak positive samples and negative samples comprises: A loss function is established based on the positive samples, weak positive samples and negative samples; wherein the loss function is anti-correlated with the feature similarity between the second description information corresponding to the node and the second description information corresponding to the positive sample of the node, anti-correlated with the feature similarity between the second description information corresponding to the node and the second description information corresponding to the weak positive sample of the node, and positively correlated with the feature similarity between the second description information corresponding to the node and the second description information corresponding to the negative sample of the node; Adjusting features of the second description information corresponding to the nodes in the multiple initial clusters based on the loss function; The multiple initial clusters are adjusted based on the adjusted features of the second description information corresponding to each node.

9. The method according to claim 1, wherein adjusting the first description information based on the second description information corresponding to the nodes in the target cluster comprises: Determine a target node set, wherein the target node set includes a target node with the largest in-degree in the target cluster and neighboring nodes of the target node, and a distance between the neighboring nodes and the target node is less than a preset distance threshold; The first description information is adjusted based on the second description information respectively corresponding to each node in the target node set.

10. The method according to claim 9, wherein the first description information includes the longitude and latitude information of the delivery location of the current delivery order, and the adjusting the first description information based on the second description information corresponding to each node in the target node set comprises: Weighting the longitude and latitude information included in the second description information corresponding to each node in the target node set to obtain adjusted longitude and latitude information included in the first description information; or, Determine an optimization target, and adjust the longitude and latitude information included in the first description information based on the optimization target; wherein the optimization target includes: minimizing the sum of distances between the longitude and latitude information included in the first description information and the longitude and latitude information included in the second description information corresponding to each node in the target node set.

11. The method according to claim 9, wherein the first description information includes address information of the delivery location of the current delivery order, and the address information includes description information corresponding to a plurality of address segments; and adjusting the first description information based on the second description information corresponding to each node in the target node set, comprises: For any one of the multiple address segments, determine the frequency of occurrence of the description information corresponding to the address segment from the second description information corresponding to each node in the target node set, and determine the description information with the highest occurrence frequency corresponding to the address segment as the description information corresponding to the address segment.

12. A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method according to any one of claims 1 to 11.

13. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 11 when executing the computer program.

14. A computer program product, comprising a computer program, which implements the method according to any one of claims 1 to 11 when executed by a processor.

Images