Communication address resolution service updating method and apparatus, device, medium, and product

Abstract

The application discloses a communication address resolution service updating method and device, equipment, medium and product, and the method comprises the following steps: in response to a communication address resolution request of a user, a first resolution interface of a self-built communication address resolution service is called to process, so that a first resolution result obtained after an application neural network model is obtained is acquired, the model is used for determining a standard address text of an input address text in the request from a standard database; when the first resolution result indicates that the resolution fails, a second resolution interface of a third-party communication address resolution service is called to process, so that a second resolution result is acquired; when any resolution result contains a standard address text and resolution information thereof, the standard address text is stored in the standard database; in response to a timing task triggering event, part of the standard address texts recalled from the standard database according to given address samples are respectively combined with the address samples to form positive samples, and the model is trained. The application can provide efficient and accurate communication address resolution services for e-commerce platforms at a relatively low cost.

Description

Technical Field

[0001] This application relates to the field of e-commerce information processing technology, and in particular to a communication address resolution service update method and its corresponding apparatus, computer equipment, computer-readable storage medium, and computer program product. Background Technology

[0002] Logistics services are particularly important in the e-commerce sector. Logistics services encompass all service activities from receiving a customer's order to delivering the goods to the customer, adding value to the traded products or services. Essentially, it aims to better meet customer needs, ensuring that the goods are delivered on time as requested and that the service meets the customer's requirements.

[0003] In practice, logistics service platforms will identify the address text provided by users, output the latitude and longitude and postal code corresponding to the address text, and send it to downstream channel partners for distribution operations and vehicle positioning operations during the logistics process, thereby improving the efficiency of ordering goods in freight calculation, pickup, distribution and delivery.

[0004] Latitude and longitude coordinates and postal code data are relatively complex and scarce. Currently, they often rely on third-party communication address resolution services to obtain them, which is costly. If the relevant functions are implemented in-house, the implementation cost is even higher and the utilization rate is relatively low. Therefore, it is necessary to explore other feasible methods. Summary of the Invention

[0005] The primary objective of this application is to address the aforementioned problems by providing a communication address resolution service update method and corresponding apparatus, computer equipment, computer-readable storage medium, and computer program product.

[0006] To achieve the various objectives of this application, the following technical solution is adopted:

[0007] A communication address resolution service update method provided for one of the purposes of this application includes the following steps:

[0008] In response to a user's communication address resolution request, the system calls a predefined first resolution interface of the self-built communication address resolution service to process the communication address resolution request, so as to obtain the first resolution result obtained by the self-built communication address resolution service after applying a neural network model. The neural network model is used to determine the standard address text corresponding to the input address text carried in the request from a standard database. The standard database stores multiple standard address texts.

[0009] When the first parsing result indicates parsing failure, the second parsing interface predefined by the third-party communication address parsing service is invoked to process the communication address parsing request, so as to obtain the second parsing result determined according to the input address text carried in the request;

[0010] When the first parsing result or the second parsing result contains the standard address text and its parsing information of the entered address text, it is pushed to the user and stored as a data record in the standard database. The parsing information includes the postal code and / or latitude and longitude corresponding to the standard address text.

[0011] In response to a scheduled task trigger event, the neural network model is applied to determine the standard address text in the standard database that constitutes a positive sample based on the given address sample. The neural network model is then iteratively trained using the positive sample until it reaches a convergent state, after which the service is restarted.

[0012] Optionally, before responding to a scheduled task trigger event, the following steps are included:

[0013] Determine whether the standard database has reached a predetermined period since the last full update. When the predetermined period is reached, iterate through each data record in the standard database, and use the standard address text in each data record as a parameter to call the second parsing interface to obtain and update the parsing information corresponding to the standard address text in the data record.

[0014] Optionally, in response to a user-submitted communication address resolution request, the system calls a predefined first resolution interface of the self-built communication address resolution service to process the request, thereby obtaining a first resolution result obtained by applying a neural network model to the self-built communication address resolution service. The neural network model is used to determine, from a standard database, the standard address text corresponding to the input address text carried in the request. The standard database stores multiple standard address texts, including:

[0015] In response to a user-submitted communication address resolution request, obtain the input address text from the communication address resolution request;

[0016] The system queries the first cache corresponding to the self-built communication address resolution service to see if there is any resolution information corresponding to the entered address text. If the resolution information exists, the system pushes the resolution information to the user.

[0017] When the first cache does not contain the parsing information, the first parsing interface predefined by the self-built communication address parsing service is called to process the communication address parsing request, so as to obtain a first parsing result containing standard address text and its corresponding parsing information.

[0018] When the first parsing result contains the parsing information, the standard address text and the input address text corresponding to the parsing information are stored in the first cache area.

[0019] Optionally, when the first or second parsing result contains the standard address text and its parsing information of the entered address text, it is pushed to the user and stored as a data record in the standard database. The parsing information includes the postal code and / or latitude and longitude corresponding to the standard address text, including:

[0020] When the first parsing result indicates parsing failure, the system queries the second cache corresponding to the third-party communication address parsing service to see if there is parsing information corresponding to the entered address text. If the parsing information exists, the system pushes the parsing information to the user.

[0021] When the parsing information is not found in the second cache, the second parsing interface predefined by the third-party communication address parsing service is invoked to process the communication address parsing request in order to obtain the second parsing result.

[0022] When the second parsing result contains the parsing information, the standard address text and the input address text corresponding to the parsing information are stored in the second cache area.

[0023] Optionally, the self-built communication address resolution service is implemented by performing the following steps:

[0024] Retrieve multiple standard address texts that match the entered address text from the standard database to form a candidate list;

[0025] The neural network model is invoked to construct input information for each standard address text in the candidate list and the input address text, and the neural network model determines the ranking score corresponding to the standard address text.

[0026] Select the standard address text with the highest sorting score that exceeds the preset threshold from the candidate list as the correct text corresponding to the entered address text;

[0027] Determine the latitude, longitude, and / or postal code of the geographical location it points to based on the correct text;

[0028] Using the postal code and / or latitude and longitude as parsing information, the system returns the standard address text corresponding to the correct text, along with the parsing information, as the first parsing result.

[0029] Optionally, the neural network model performs the following steps:

[0030] The application encoding layer encodes the input address text and standard address text in the input information into embedding vectors respectively;

[0031] The application feature extraction layer extracts deep semantic information from the embedded vector to obtain the corresponding address feature vector;

[0032] A linear layer is applied to calculate the similarity between the address feature vectors to obtain a similarity vector;

[0033] The similarity vector is mapped to a preset classification space using a classifier, and the classification probability corresponding to the positive category is used as the ranking score of the standard address text in the input information.

[0034] A communication address resolution service update device provided to meet one of the purposes of this application includes: a self-built service invocation module, a third-party service invocation module, a result push and storage module, and a timed training and update module. The self-built service invocation module is configured to respond to a user-submitted communication address resolution request by invoking a predefined first resolution interface of the self-built communication address resolution service to process the request, thereby obtaining a first resolution result obtained by applying a neural network model to the self-built communication address resolution service. The neural network model is used to determine, from a standard database, the standard address text corresponding to the input address text carried in the request, and the standard database stores multiple standard address texts. The third-party service invocation module is configured to invoke a third-party communication address resolution service when the first resolution result indicates resolution failure. A predefined second parsing interface processes the communication address parsing request to obtain a second parsing result determined based on the entered address text carried in the request. The result push storage module is configured to push the standard address text and its parsing information of the entered address text to the user when the first parsing result or the second parsing result contains the standard address text of the entered address text and its parsing information, and store it as a data record in the standard database. The parsing information includes the postal code and / or latitude and longitude corresponding to the standard address text. The timed training and update module is configured to respond to a timed task trigger event, apply the neural network model to determine the standard address text in the standard database that constitutes a positive sample based on the given address sample, use the positive sample to iteratively train the neural network model, and restart the service after training it to a convergent state.

[0035] A computer device provided for one of the purposes of this application includes a central processing unit and a memory, wherein the central processing unit is configured to invoke and run a computer program stored in the memory to perform the steps of the communication address resolution service update method described in this application.

[0036] A computer-readable storage medium is provided for another purpose of this application, which stores, in the form of computer-readable instructions, a computer program implemented according to the described communication address resolution service update method, which, when invoked by a computer, performs the steps included in the method.

[0037] A computer program product provided for another purpose of this application includes a computer program / instructions that, when executed by a processor, implement the steps of the method described in any embodiment of this application.

[0038] Compared to existing technologies, this application prioritizes handling user-submitted address resolution requests with a self-built address service. Only when the self-built address service fails to obtain the corresponding result is the request forwarded to a third-party address service. This avoids directly calling third-party address services to prevent high resolution costs. Furthermore, the resolution information generated by the third-party address service, due to its authority, is stored in a standard database for reuse. This database can be used in the training and inference phases of the neural network model used by the self-built address service in this application. Multiple standard address texts are retrieved during the training and inference processes. In the training phase, these retrieved standard address texts can be used to construct training samples for the neural network model. In the inference phase, the neural network model can predict the ranking score corresponding to each standard address text, thereby determining the correct text. Based on the correct text, historically obtained resolution information is then returned to the user. Therefore, this application can reuse third-party data at a relatively low cost to provide e-commerce platforms with efficient and accurate address resolution services, and achieve continuous updates to improve the service capabilities of the self-built address resolution service. Attached Figure Description

[0039] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0040] Figure 1 This is a flowchart illustrating one embodiment of the communication address resolution service update method of this application;

[0041] Figure 2 This is a flowchart illustrating the process of responding to a communication address resolution request, calling a self-built communication address resolution service, and caching the resolution information in an embodiment of this application.

[0042] Figure 3 This is a schematic diagram of the response and caching process when the self-built communication address resolution fails in the embodiments of this application;

[0043] Figure 4This is a flowchart illustrating the working process of the self-built communication address resolution service in the embodiments of this application;

[0044] Figure 5 This is a schematic diagram of the network architecture of the neural network model used in this application as an example;

[0045] Figure 6 This is a schematic diagram illustrating the workflow of the neural network model described in this application;

[0046] Figure 7 A schematic block diagram of the communication address resolution service update device of this application;

[0047] Figure 8 This is a schematic diagram of the structure of a computer device used in this application. Detailed Implementation

[0048] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0049] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this application means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0050] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0051] Those skilled in the art will understand that the terms "client," "terminal," and "terminal device" as used herein include both devices that receive wireless signals, devices that only possess wireless signal receiver capabilities without transmission capabilities, and devices with receiving and transmitting hardware, devices that have receiving and transmitting hardware capable of bidirectional communication over a bidirectional communication link. Such devices may include: cellular or other communication devices such as personal computers or tablets, having single-line displays, multi-line displays, or cellular or other communication devices without multi-line displays; PCS (Personal Communications Service) that can combine voice, data processing, fax, and / or data communication capabilities; PDA (Personal Digital Assistant) that may include a radio frequency receiver, pager, internet / intranet access, web browser, notepad, calendar, and / or GPS (Global Positioning System) receiver; and conventional laptops and / or handheld computers or other devices that have and / or include radio frequency receivers. As used herein, "client," "terminal," and "terminal device" can be portable, transportable, installed in a means of transportation (air, sea, and / or land), or suitable and / or configured to operate locally and / or in a distributed manner, operating in any other location on Earth and / or in space. "Client," "terminal," and "terminal device" as used herein can also be a communication terminal, an internet access terminal, or a music / video playback terminal, such as a PDA, a MID (Mobile Internet Device), and / or a mobile phone with music / video playback capabilities, or a smart TV, set-top box, etc.

[0052] The hardware referred to by the names "server," "client," and "service node" in this application is essentially an electronic device with the equivalent capabilities of a personal computer. It is a hardware device with the necessary components revealed by the von Neumann architecture, such as a central processing unit (including an arithmetic logic unit and a control unit), memory, input devices, and output devices. The computer program is stored in its memory, and the central processing unit loads the program stored in the secondary storage into the main memory to run it, execute the instructions in the program, and interact with the input and output devices to complete specific functions.

[0053] It should be noted that the concept of "server" used in this application can also be extended to the case of server clusters. Based on the network deployment principles understood by those skilled in the art, the servers should be logically divided. Physically, these servers can be independent of each other but accessible through interfaces, or they can be integrated into a single physical computer or a computer cluster. Those skilled in the art should understand this flexibility and should not use it to constrain the implementation of the network deployment method in this application.

[0054] One or more of the technical features of this application, unless explicitly specified herein, can be deployed on a server and accessed by a client remotely calling the online service interface provided by the server, or can be directly deployed and run on a client for access.

[0055] Unless otherwise specified, the neural network models referenced or potentially referenced in this application may be deployed on a remote server and invoked remotely on the client, or deployed on a client with the capability to invoke directly. In some embodiments, when running on the client, the corresponding intelligence may be acquired through transfer learning in order to reduce the requirements on the client's hardware resources and avoid excessive consumption of the client's hardware resources.

[0056] Unless otherwise specified, all data involved in this application may be stored remotely on a server or on a local terminal device, as long as it is suitable for use by the technical solution of this application.

[0057] Those skilled in the art will understand that although the various methods in this application are described based on the same concept and thus present commonality among them, they can be performed independently unless otherwise specified. Similarly, the various embodiments disclosed in this application are all based on the same inventive concept; therefore, concepts expressed in the same way, as well as concepts that are appropriately changed for convenience but are expressed differently, should be understood equivalently.

[0058] Unless otherwise expressly stated, the various embodiments disclosed in this application can be combined in a cross-cutting manner to flexibly construct new embodiments, as long as such combination does not depart from the inventive spirit of this application and can meet the needs of the prior art or solve a certain deficiency in the prior art. Those skilled in the art should be aware of such modifications.

[0059] The communication address resolution service update method of this application can be programmed into a computer program product and deployed in a client and / or server to run. For example, in the e-commerce platform application scenario of this application, it can be implemented in the website page by coordinating the information interaction between the client and the server.

[0060] Please see Figure 1 In one embodiment of the communication address resolution service update method of this application, the following steps are included:

[0061] Step S1100: Respond to the communication address resolution request submitted by the user, call the first resolution interface predefined by the self-built communication address resolution service to process the communication address resolution request, so as to obtain the first resolution result obtained by the self-built communication address resolution service after applying the neural network model. The neural network model is used to determine the standard address text corresponding to the input address text carried in the request from the standard database. The standard database stores multiple standard address texts.

[0062] In an exemplary e-commerce platform application scenario, this application is used to respond to communication address resolution requests submitted by users accessing various independent websites. These users can be either merchant users or consumer users on the independent websites. The independent websites are used to deploy online stores and are not affiliated with other independent websites, but they can access various interfaces provided by the computer program product that has deployed the technical solution of this application. The server for this application is typically provided by the e-commerce platform service provider so that, after the computer program product is installed, a self-built communication address resolution service can be run to centrally process various communication address resolution requests forwarded from various independent websites.

[0063] When a consumer user enters an address text as a delivery address during the ordering process at an online store on an independent website, this address text, as the entered address text, is encapsulated in a communication address resolution request and sent to the server of this application. The server is requested to parse the entered address text to obtain its corresponding postal code and / or latitude and longitude. The postal code can be used to complete the relevant information required for the delivery address, and the latitude and longitude can be used to locate the destination on a map for relevant users to refer to.

[0064] Similarly, online store merchants can also trigger the aforementioned communication address resolution request to further correct the relevant address information when the address information provided by the consumer user is incomplete, and further determine the postal code and / or latitude and longitude based on the address text entered by the consumer user, so as to improve the business data required for the logistics processing of the order, making the corresponding packages easier to distribute and deliver during the logistics process.

[0065] When the server receives the communication address resolution request, it can call the first resolution interface predefined in its self-built communication address resolution service to process the request. In one embodiment, the server can also determine whether to use the first resolution interface to process the communication address resolution request based on whether the traffic of its self-built communication address resolution service has reached a threshold. If the traffic has not reached the threshold, the first resolution interface is used by default. If the traffic has reached the threshold, a second resolution interface predefined in a third-party communication address resolution service can be called to process the request.

[0066] After the first parsing interface is called by default to process the communication address parsing request, the self-built communication address parsing service obtains the input address text carried in the communication address parsing request, calls a pre-trained neural network model to convergence state to correct the input address text to obtain its corresponding standard address text, and then queries and determines its corresponding postal code and / or latitude and longitude based on the standard address text. The postal code and / or latitude and longitude are the parsing information corresponding to the input address text. The standard address text is associated with the parsing information and encapsulated into a first parsing result, which is returned to the caller. The caller user then obtains the parsing information corresponding to the input address text.

[0067] The neural network model is implemented to predict the similarity between two given address texts, typically including an input address text and a standard address text, based on their deep semantic information. This similarity is then used as a ranking score. Multiple standard address texts are associated with the input address text to form multiple address pairs. Each address pair is input into the neural network model to determine its corresponding ranking score. Finally, the self-built communication address resolution service can determine that the standard address text with the highest ranking score is the correct text of the input address text. Therefore, the resolution information determined based on the standard address text is correct.

[0068] The standard address text can be retrieved from a standard database based on the entered address text for further input into the neural network model to determine its ranking score. The retrieval strategy can be to retrieve all standard address texts in the standard database, or to determine a subset of standard address texts based on rule matching or close edit distance matching. In short, by applying the retrieval strategy, a retrieved address set can be obtained, which is a subset of the standard address database. Subsequently, the neural network model can be applied to each standard address text in the retrieved address set to determine its ranking score. Therefore, it is easy to understand that the correlation between the coarsely retrieved standard address text and the entered address text is not precise enough. However, by using the neural network model of this application to determine the corresponding ranking score for each standard address text, the correlation between the standard address text and the entered address text can be more accurately represented. Based on this, the corresponding standard address text can be determined as the correct text for the entered address text to obtain the corresponding parsing information, which is more accurate.

[0069] The aforementioned standard database may include multiple standard address texts. In one embodiment, these standard address texts can be obtained by responding to a large number of user-submitted communication address resolution requests through a third-party communication address resolution service, and these historical standard address texts and resolution information are stored in a backlog. Since the resolution results of the third-party communication address resolution service, i.e., the standard address texts and resolution information it determines, are authoritative, storing them in the standard database allows for recall of address texts when a self-built communication address resolution service needs to provide specific resolution services, effectively reducing the high-frequency use of the third-party communication address resolution service.

[0070] Of course, the self-built communication address resolution service may fail to recall any standard address text from the standard database, or, although some standard address text may be recalled, after each of them is paired with the entered address text and the neural network model is used to determine the ranking score, none of the ranking scores exceed the preset threshold. In this case, it means that the self-built communication address resolution service has failed to provide effective resolution information for the entered address text, and therefore can return a null value or other specific identifier as the first resolution result to indicate that the resolution has failed.

[0071] Step S1200: When the first parsing result indicates parsing failure, the second parsing interface predefined by the third-party communication address parsing service is invoked to process the communication address parsing request, so as to obtain the second parsing result determined according to the input address text carried in the request;

[0072] When the first parsing result indicates that the self-built communication address parsing service fails to parse the entered address text, the second parsing interface predefined by the third-party communication address parsing service can be called to process the communication address parsing request. Specifically, the entered address text carried in the communication address parsing request can be passed as a parameter to the second parsing interface, and then the second parsing result returned by the second parsing interface can be waited for.

[0073] The third-party address resolution service is a standardized and well-known service. Its function is to determine the correct standard address text and its corresponding postal code and / or latitude and longitude based on the address text entered by the caller of the second resolution interface. For this application, as long as the second resolution interface is called, the corresponding second resolution result can be expected. The specific implementation of the third-party address resolution service is flexibly set by the third party and does not affect the inventive spirit of this application.

[0074] Step S1300: When the first parsing result or the second parsing result contains the standard address text and its parsing information of the entered address text, it is pushed to the user and stored as a data record in the standard database. The parsing information includes the postal code and / or latitude and longitude corresponding to the standard address text.

[0075] Whether the self-built address resolution service is used by default to handle address resolution requests, or a third-party address resolution service is called after the self-built service fails, the corresponding resolution result will be obtained, namely the first resolution result or the second resolution result. Therefore, the final resolution result can be identified. If it returns standard address text and its resolution information, i.e., postal code and / or latitude and longitude, the resolution of the entered address text can be confirmed as successful, and it can then be pushed to the user who submitted the address resolution request as a response. Thus, the self-built address resolution service, in conjunction with the third-party address resolution service, can reuse historical data generated by the third-party service, avoiding frequent calls to the second resolution interface provided by the third-party service, and providing a faster and lower-cost address resolution service.

[0076] On the other hand, as mentioned earlier, since the final determined standard address text and its parsing results are authoritative, they can be stored in the standard database for use by self-built communication address resolution services. In one embodiment, for the standard address text and its parsing information contained in the first parsing result, since it is taken from the standard database, storing it in the standard database is actually replacing the old data record; therefore, it is not necessary to perform a storage operation on the first parsing result. That is, the focus is on storing the standard address text and its postal code and / or latitude and longitude in the second parsing result into the standard database. If necessary, duplicate data records can be deduplicated during the storage process, or for parsing results that already have duplicate data records, no storage processing is required.

[0077] Step S1400: Respond to the timed task trigger event, apply the neural network model to determine the standard address text in the standard database that constitutes a positive sample based on the given address sample, use the positive sample to perform iterative training on the neural network model, train it to a convergent state and then restart the service.

[0078] As mentioned above, the historical data generated by the third-party address resolution service is stored in the standard database. The standard address text in the standard database can provide a recall source for the self-built address resolution service. In addition, the standard address text in the standard database can also be used to train the neural network model described in this application, so as to achieve high reuse of the resolution results generated by the third-party address resolution service, increase the marginal revenue of paying for the third-party address resolution service, and effectively reduce the marginal cost of responding to each address resolution request.

[0079] The training of the neural network model can be performed periodically. Therefore, a timed task can be preset, such as once every six months, or once a predetermined period is reached from the date of the last full update of the standard database, which will trigger the corresponding timed task trigger event and instruct the execution of training of the neural network model.

[0080] Before training, a list of address samples is obtained, which provides a subset of address samples. Then, for each address sample, the trained neural network model is invoked, and its standard address text is determined as described above. The address sample and its corresponding standard address text are then used to construct a positive sample. Alternatively, the address text can be combined with other standard address texts in the standard database to construct a negative sample. Each address sample can generate its corresponding positive and negative samples. Thus, the positive and negative samples can be used as training samples to form a training dataset. Any training sample in this training dataset can be used to retrain the neural network model.

[0081] Training the neural network model can begin by inputting a single training sample from the training dataset into the model. The model encodes two address texts (the address sample and its associated standard address text) in the training sample to obtain corresponding embedding vectors. Then, deep semantic information is extracted from each embedding vector, and their similarity is calculated based on this deep semantic information to obtain a similarity vector. Finally, the similarity vector is mapped by a classifier. The classifier's loss value is calculated based on whether the training sample is positive or negative. If the loss value does not reach a preset threshold or the number of iterations does not reach a preset number, the weights of the neural network model are adjusted, and the next training sample is input from the training dataset to iteratively train the model until the loss value reaches the preset threshold or the number of iterations reaches the preset number. At this point, the neural network model is considered to have converged, and retraining can be terminated.

[0082] The classification results in the classifier include the confidence scores corresponding to each category in the classification space defined by the classifier, expressed as probability values. One category is designated as the positive category. When the positive category has the highest confidence score, it indicates that the input training sample is identified as a positive sample; otherwise, the input training sample is identified as a negative sample. Therefore, the loss value can be calculated according to this principle. Subsequently, when applying the classification results of the neural network model, it is easy to understand that the confidence score obtained by the positive category can be directly determined as the ranking score obtained by the standard address text in the corresponding address pair.

[0083] After the neural network model has been trained on historical standard address text accumulated over a period of time, the standard database has accumulated even more historical data during this period. This effectively expands the scale of training data required for the neural network model. Therefore, retraining the neural network model can improve its generalization ability and the accuracy of its ranking score determination. It is evident that this application not only reuses historical data generated by third-party address resolution services but also continuously improves the accuracy of the neural network model in the self-built address resolution service in determining ranking scores through a continuous cyclical update mechanism, thereby continuously enhancing the overall service capability of the self-built address resolution service.

[0084] As can be seen from the above embodiments, this application prioritizes processing user-submitted communication address resolution requests through a self-built communication address service. Only when the self-built communication address service fails to obtain the corresponding result is the communication address resolution request forwarded to a third-party communication address service. This avoids directly calling third-party communication address services to prevent high resolution costs. Furthermore, the resolution information generated by the third-party communication address service in processing the communication address resolution request is authoritative and will be stored in a standard database for reuse. This database can be used for the training and inference phases of the neural network model used by the self-built communication address service in this application, recalling multiple standard address texts for the training and inference process. During the training phase, these recalled standard address texts can be used to construct training samples required for the neural network model training. During the inference phase, the neural network model can predict the ranking score corresponding to each standard address text to determine the correct text, so that the historically obtained resolution information can be retrieved based on the correct text and returned to the user. Therefore, this application can reuse third-party data in a relatively low-cost manner to provide e-commerce platforms with efficient and accurate communication address resolution services, and achieve cyclical updates to improve the corresponding service capabilities of the self-built communication address resolution service.

[0085] Based on any embodiment of this application, before responding to the timed task triggering event, the method includes: determining whether the standard database has reached a predetermined period since the last full update; when the predetermined period is reached, traversing each data record in the standard database, using the standard address text in each data record as a parameter, calling the second parsing interface to obtain and update the parsing information corresponding to the standard address text in the data record.

[0086] Considering the dynamic changes in administrative divisions of various countries and other information updates that may lead to changes in the authoritative information of third-party address resolution services, all data in the standard database of this application can be periodically updated. After each full update, the update time is marked. The initial full update time can be set to the creation time of the standard database. Thus, through periodic full updates, the latest information changes of third-party address resolution services can be adapted, so that the resolution results obtained by the self-built address resolution service are as synchronized as possible with the actual results of third-party address resolution services.

[0087] Therefore, prior to responding to the scheduled task trigger event, it can be determined whether a predetermined period, such as six months, has been reached since the last full update of the standard database. Once the predetermined period is reached, a full update can be performed on the standard address text in each data record of the standard database. Each standard address text is used as a parameter to call the second parsing interface of the third-party communication address resolution service. The latest standard address text and its corresponding parsing information are obtained through the second parsing interface. Then, for data that has changed relative to the corresponding historical data record in the standard database, the latest standard address text and its corresponding parsing information can be used to replace the historical data record.

[0088] As can be seen from the above embodiments, by periodically traversing and updating the standard database, all historically accumulated standard address texts and their corresponding resolution information can be updated in a timely manner, ensuring that users are provided with the most accurate and timely communication address resolution service.

[0089] Based on any embodiment of this application, please refer to Figure 2 In response to a user-submitted communication address resolution request, the system calls a predefined first resolution interface of the self-built communication address resolution service to process the request, thereby obtaining a first resolution result obtained by applying a neural network model to the self-built communication address resolution service. The neural network model is used to determine, from a standard database, the standard address text corresponding to the input address text carried in the request. The standard database stores multiple standard address texts, including:

[0090] Step S1110: Respond to the communication address resolution request submitted by the user and obtain the input address text in the communication address resolution request;

[0091] When a user submits the communication address resolution request and it is received by the server, the server parses the request and extracts the input address text carried therein. The input address text may include prefix information for representing administrative divisions at various levels and suffix information for representing the user's specific detailed address. In one embodiment, only the prefix information can be processed. Therefore, the prefix information can be directly used as the input address text, while the suffix information can be determined by the user.

[0092] Step S1120: Query the first cache corresponding to the self-built communication address resolution service to see if there is resolution information corresponding to the entered address text. If the resolution information exists, push the resolution information to the user.

[0093] In this embodiment, the self-built communication address resolution service corresponds to a preset first cache area, which is used to cache the first resolution result obtained after responding to any communication address resolution request. The input address text of each communication address resolution request and its corresponding first resolution result are cached in the first cache area. Thus, when responding to a communication address resolution request, the first cache area is first queried based on the input address text carried by the communication address resolution request. If a first resolution result corresponding to the input address text exists, specifically confirming the existence of resolution information in the first resolution result, namely postal code and / or latitude and longitude, the resolution information can be directly pushed to the user who submitted the communication address resolution request to complete the resolution service without continuing with other steps.

[0094] Step S1130: When the first cache does not contain the parsing information, the first parsing interface predefined by the self-built communication address parsing service is called to process the communication address parsing request, so as to obtain a first parsing result containing standard address text and its corresponding parsing information.

[0095] If, after querying, it is confirmed that the first cache does not contain the parsing information corresponding to the entered address text, then the communication address parsing request can be processed by calling the first parsing interface of the self-built communication address parsing service, as described in step S1100 above, in order to obtain the corresponding first parsing result. The subsequent process is the same as disclosed in the previous embodiments.

[0096] Step S1140: When the first parsing result contains the parsing information, the standard address text and the input address text corresponding to the parsing information are stored in the first cache area.

[0097] It should be noted that after obtaining the corresponding first resolution result by calling the self-built communication address resolution service in step S1130, it can be associated with the entered address text and stored in the first cache area, thereby realizing the association and caching of the entered address text, standard address text and corresponding resolution information for subsequent query.

[0098] According to the above embodiments, the self-built communication address resolution service can provide faster resolution services through its first cache, avoid high server load, reduce computational load, and obtain economic benefits from scaled utility.

[0099] Based on any embodiment of this application, please refer to Figure 3When the first or second parsing result contains the standard address text and its parsing information of the entered address text, it is pushed to the user and stored as a data record in the standard database. The parsing information includes the postal code and / or latitude and longitude corresponding to the standard address text, including:

[0100] Step S1210: When the first parsing result indicates parsing failure, query the second cache corresponding to the third-party communication address parsing service to see if there is parsing information corresponding to the entered address text. If the parsing information exists, push the parsing information to the user.

[0101] Similar to the previous embodiment, this application also pre-defines a second cache area for caching the second resolution results returned by the third-party communication address resolution service. Therefore, when the first resolution result returned by the self-built communication address resolution service indicates a resolution failure, the system first checks the second cache area to see if it contains resolution information corresponding to the entered address text. If such resolution information exists, it can be directly invoked and returned to the user who submitted the communication address resolution request. This saves calls to the second resolution interface and allows for faster result retrieval for the user.

[0102] Step S1220: When the second cache does not contain the parsing information, the second parsing interface predefined by the third-party communication address parsing service is invoked to process the communication address parsing request in order to obtain the second parsing result;

[0103] Similarly to the previous embodiment, if after querying it is confirmed that the second cache does not contain the parsing information corresponding to the entered address text, then the communication address parsing request can be processed by calling the second parsing interface of the third-party communication address parsing service according to the process described in step S1200 above, in order to obtain the corresponding second parsing result. The subsequent process is the same as disclosed in the previous embodiment.

[0104] Step S1230: When the second parsing result contains the parsing information, store the standard address text and the input address text corresponding to the parsing information in the second cache area.

[0105] It should be noted that after obtaining the corresponding second resolution result by calling the third communication address resolution service in step S1220, it can be associated with the entered address text and stored in the first cache area, thereby realizing the association and caching of the entered address text, standard address text and corresponding resolution information for subsequent query.

[0106] According to the above embodiments, the second resolution result obtained by the third-party communication address resolution service is associated with the entered address text and stored in the second cache area. The second cache area can provide a faster resolution service, which can effectively reduce the frequent calls to the second resolution interface and achieve the economic benefits brought by the scale effect.

[0107] Based on any embodiment of this application, please refer to Figure 4 The self-built communication address resolution service is implemented by performing the following steps:

[0108] Step S2100: Retrieve multiple standard address texts that match the entered address text from the standard database to form a candidate list;

[0109] When responding to a communication address resolution request, the self-built communication address resolution service first obtains the input address text in the request through the parameters passed to its first resolution interface. Then, it can perform word segmentation on the input address text using a word segmentation algorithm to obtain its word segmentation sequence and convert it into word segmentation vectors. Similarly, each standard address text in the standard database can also pre-obtain its word segmentation vector by performing word segmentation using the same word segmentation algorithm. Thus, by using the word segmentation vector of the input address text to calculate the data distance between it and the word segmentation vectors of each standard address text in the standard database, the distance value of each standard address text can be determined. This distance value represents the semantic similarity between the word segmentation vector of the corresponding standard address text and the word segmentation vector of the input address text. Based on this, a preset threshold is used to filter out standard address texts with distance values ​​higher than the preset threshold and construct a candidate list. This candidate list is actually the recall address set obtained from the coarse recall of the input address text from the standard database.

[0110] Step S2200: Call the neural network model to construct input information for each standard address text in the candidate list and the input address text respectively, and determine the ranking score corresponding to the standard address text by the neural network model;

[0111] As described above, the neural network model of this application is suitable for determining the confidence level of mapping two address texts to a positive category for an address pair. When the two address texts are an input address text and a standard address text, the confidence level can be used to characterize the ranking score of the standard address text. Therefore, for the candidate list, each standard address text can be constructed as an address pair with the input address text. Then, the address pair is used as input information and input into the neural network model, which determines its corresponding ranking score. This ranking score is then associated with the standard address text in the address pair and stored for later use. Thus, each standard address text in the candidate list can obtain its corresponding ranking score.

[0112] Step S2300: Select the standard address text with the highest sorting score that exceeds the preset threshold from the candidate list as the correct text corresponding to the entered address text;

[0113] For each standard address text in the candidate list, the one with the highest ranking score is theoretically the closest to the entered address text. However, the confidence level of the standard address text in the positive category mapped by the neural network model, i.e., the ranking score, also characterizes the degree of similarity between the standard address text and the entered address text. Therefore, to determine the correct standard address text for the entered address text, a preset threshold can be provided. This preset threshold can be an empirical threshold or an experimental threshold. Then, the highest ranking score in the candidate list is compared with the preset threshold. Only when the highest ranking score is higher than the preset threshold is the standard address text corresponding to the highest ranking score determined as the correct text for the entered address text. Applying this mechanism effectively corrects the entered address text submitted by the user to an authoritative standard address text.

[0114] Step S2400: Determine the latitude and longitude and / or postal code of the geographical location to which the correct text points;

[0115] To accommodate the mapping relationship between standard address text and corresponding parsing information, a parsing database is pre-existing. This database stores the mapping relationship between administrative divisions of various countries and their corresponding latitude, longitude and / or postal codes. Accordingly, once a standard address text is confirmed as the correct text for entering an address, the database can be queried based on the correct text, and the latitude, longitude and / or postal code corresponding to the correct text can be determined using the mapping relationship.

[0116] Step S2500: Using the postal code and / or latitude and longitude as parsing information, return the standard address text corresponding to the correct text and the parsing information as the first parsing result.

[0117] After obtaining the postal code and / or latitude and longitude corresponding to the correct text of the entered address text, encapsulate it into parsing information, associate it with the standard address text corresponding to the correct text, for example, construct it as a key-value pair, and return it as the first parsing result.

[0118] As can be seen from the above embodiments, the self-built communication address resolution service of this application can effectively utilize the standard database. First, a candidate list storing standard address text is obtained from the standard database in a coarse manner. Then, a neural network model is used to determine the ranking score of each standard address text in the candidate list. The standard address text corresponding to the highest ranking score that exceeds the preset threshold is determined as the correct text of the input address text. Based on this, its resolution information is determined. The entire business process does not need to directly rely on a third-party communication address resolution service, but can use local resources to obtain the resolution information of the input address text more quickly. The processing efficiency is high and the processing cost is low.

[0119] Based on any embodiment of this application, please refer to Figure 5 The network architecture shown is based on Figure 5 It can be seen that the neural network model is structured as a dual-tower model, which contains two processing branches. Each processing branch contains the same network structure. In each processing branch, along the direction from input to output, there are encoding layers and feature extraction layers. Then, the outputs of the two processing branches are respectively connected to linear layers to calculate semantic similarity. Finally, the similarity vector is input into the classifier to determine the ranking score.

[0120] Further integration Figure 6 The flowchart shown is as follows. Figure 5 An exemplary neural network model performs the following steps:

[0121] Step S3100: The input address text and standard address text in the input information are encoded into embedding vectors by the application encoding layer;

[0122] First, for two address texts that need to be encoded, such as input address text and standard address text, the two address texts are respectively input into the two branches of the neural network model. They first enter the encoding layer of each branch for encoding. Through encoding, multiple embedding vectors corresponding to each address text are obtained.

[0123] In the encoding layer of each branch, for each address text, two granularities of word segmentation are performed. For example, the N-Gram algorithm can be used, where N is set to two grammars (binary and trigram), effectively implementing two word segmenters with different granularities. These two segmenters perform parallel word segmentation on the same address text, thus obtaining corresponding word segments for the two grammars and three grammars. The two-gram segmentation sequence is the set of word segments obtained by sliding a window of 2 characters at a time on the address text, while the three-gram segmentation sequence is the set of word segments obtained by sliding a window of 3 characters at a time on the address text.

[0124] It is evident that for two address texts that require semantic similarity calculation, word segmentation sequences can be constructed in parallel for each other, and for each address text, word segmentation sequences of different granularities can also be constructed in parallel.

[0125] For each of the two address texts, multiple word segmentation sequences have been obtained through word segmentation, such as the bigram and trigram word segmentation sequences. Accordingly, for each word segmentation sequence, by referencing a preset reference lexicon, the encoded value of each word in the segmentation sequence is queried. Based on the corresponding position of each word in its respective word segmentation sequence, the encoded values ​​of all words are systematically constructed into an embedding vector, thus achieving word embedding for the segmentation sequence. Therefore, as shown in the previous example, each of the two address texts can obtain its corresponding bigram and trigram embedding vectors.

[0126] As can be seen from this step, in the vectorization process before calculating the semantic similarity between two address texts (i.e., the input standard text and any standard address text in the candidate list), each address text is first simultaneously segmented at different granularities to obtain multiple embedding vectors. Segmentation at different granularities more accurately achieves the semantic expression of the address text. For example, for "Guangdong Province, Guangzhou City, Huangpu District", binary segmentation can obtain the segmentation sequence as [Guangdong; Dongsheng; Shengguang; Guangzhou; Zhouzhou; Shihuang; Huangpu; Pu District]. The word segmentation sequence obtained through ternary segmentation is [Guangdong Province; Guangdong Province; Guangzhou City; Huangpu District; Huangpu District]. This demonstrates that the word segmentation [Guangdong Province; Guangzhou City; Huangpu District] in the binary segmentation sequence can accurately represent an address, and the word segmentation [Guangdong Province; Guangzhou City; Huangpu District] in the ternary segmentation sequence can also accurately represent an address. Therefore, word segmentation at different granularities can more accurately discover various possible expressions of address information, providing the necessary information foundation for subsequent semantic mining and making the determination of semantic similarity more accurate.

[0127] Step S3200: Apply the feature extraction layer to extract deep semantic information from the embedded vector to obtain the corresponding address feature vector;

[0128] After obtaining all the embedding vectors of the corresponding address text in the encoding layer of each processing branch, all the embedding vectors can be input into the feature extraction layer of the processing branch to construct the corresponding address feature vector.

[0129] In one embodiment, the feature extraction layer may include two modules and a concatenation layer, namely an attention layer module and a pooling operation module. The attention layer module is used to determine the key feature vector corresponding to each embedding vector as a query vector. The pooling operation module is used to determine the compressed feature vector of each embedding vector. Then, the concatenation layer concatenates all the key feature vectors and all the compressed feature vectors to obtain the address feature vector of the corresponding address text.

[0130] In one embodiment of the feature extractor, the key feature vectors corresponding to each embedding vector can be determined by the following process:

[0131] The first step is to take each of the embedding vectors for each address text and use it as a query vector, while using the remaining embedding vectors as key and value vectors, and input them into the attention layer to perform attention operations, thereby obtaining the key feature vectors corresponding to each embedding vector:

[0132] To achieve deeper semantic mining, an attention layer is used to perform attention operations on all embedding vectors of each address text to obtain the corresponding address feature vectors.

[0133] The mechanism of attention operation is to use a given query vector to retrieve key weight information from a given key vector, and then extract key feature vectors from the value vector based on the normalized result of the key weight information.

[0134] Based on this mechanism, in this embodiment, for all embedding vectors of each address text, the first embedding vector is determined as the query vector, and the other embedding vectors are key vectors and value vectors. Then, the first key feature vector corresponding to its first embedding vector is retrieved. For each embedding vector, the attention operation is performed on it in turn with other embedding vectors as the first embedding vector. Thus, each embedding vector in the same address text can obtain the key feature vector determined with it as the query vector. Therefore, the same address text can obtain multiple key feature vectors. For example, for the binary embedding vector and the ternary embedding vector in the same address text, two key feature vectors can be obtained respectively.

[0135] In this embodiment, the corresponding key feature vector is determined for each embedding vector in the same address text in order to enable mutual query between different embedding vectors and ensure that the final set of key feature vectors does not omit key semantic information in the address text.

[0136] It is easy to understand that, according to this step, each address text can obtain its corresponding set of key feature vectors, that is, each address text has multiple key feature vectors corresponding to the number of its embedded vectors.

[0137] The second step involves performing pooling operations on all embedding vectors for each address text and then concatenating them to obtain the compressed feature vectors corresponding to each embedding vector:

[0138] Although the first step can uncover the deep semantics of the address text, the entire set of embedding vectors of the address text itself possesses original semantics and can be utilized when determining semantic similarity. Therefore, this step can be executed in parallel with the previous step, using the aforementioned pooling layer to perform pooling operations on all embedding vectors of each address text, thereby compressing each embedding vector and obtaining the corresponding compressed feature vector. Thus, each embedding vector of each address text undergoes feature compression to obtain its corresponding compressed feature vector. Although the compressed feature vector has fewer dimensions, it represents relatively original semantics. For the same address text, the compressed feature vector and the key feature vector of that address text complement each other perfectly.

[0139] The third step is to concatenate all key feature vectors and all compressed feature vectors of each address text into an address feature vector:

[0140] To simplify the feature representation of each address text, the concatenation layer can be used to concatenate all key feature vectors and all compressed feature vectors corresponding to each address text. The concatenation order should follow a uniform preset order, thereby obtaining the address feature vector corresponding to the address text.

[0141] Based on the above process, it is easy to understand that by using the attention layer to perform deep and comprehensive semantic mining on the embedding vector of each address text to obtain the corresponding key feature vector, and then combining the compressed feature vector obtained by compressing the embedding vector to carry the original semantics of the address text, the address feature vector of the address text is composed of the key feature vector and the compressed feature vector. This can achieve an effective, comprehensive and accurate feature representation of the address text, and provide a reliable information foundation for accurately calculating the semantic similarity between different address texts.

[0142] Step S3300: Apply a linear layer to calculate the similarity between the address feature vectors to obtain a similarity vector;

[0143] After the two branches obtain the address feature vectors of their respective input address texts, the two address feature vectors are input into the linear layer of the neural network model. The linear layer performs similarity calculations to calculate the semantic similarity between the two address feature vectors, which can be represented as a similarity vector.

[0144] Specifically, after determining the corresponding address feature vectors of the two address texts, a preset data distance algorithm can be applied through a linear layer to calculate their semantic similarity. The data distance algorithm can be any available algorithm, including but not limited to: cosine similarity algorithm, vector dot product algorithm, Euclidean distance algorithm, Pearson correlation coefficient, etc. By calculating the data distance between the two address texts using any data distance algorithm, it can be converted into the corresponding semantic similarity. Therefore, the semantic similarity corresponding to each standard address text represents the degree of semantic association between the standard address text and the entered address text. The higher the semantic similarity, the more consistent the text content of the two; conversely, the lower the semantic similarity, the more inconsistent the text content of the two.

[0145] Step S3400: Apply a classifier to map the similarity vector to a preset classification space, and obtain the classification probability corresponding to the positive category as the sorting score of the standard address text in the input information.

[0146] Finally, the similarity vector is input into the classifier of the neural network model. In the classifier, a fully connected layer is first used to fully connect the similarity vector and then classify and map it to the output layer. The output layer uses the Softmax function to calculate the classification probability of the similarity vector to each category in the preset classification space. Then, the classification probability of the positive category is directly used as the sorting score corresponding to the standard address text of the input information.

[0147] The positive category in the classification space refers to the category used to correspond to positive samples during the training of the neural network model. The classifier can be a multi-classifier or a binary classifier, but during the training phase, a category is designated as the positive category, and positive samples are used to supervise its results. During the inference phase, the classification probability of the positive category is used to represent the ranking score.

[0148] According to the above embodiments, the neural network model of this application is used to calculate the sorting score of the input address text and any standard address text, which is faster and can serve a large number of requests, thereby improving the service efficiency of address association.

[0149] Furthermore, the neural network model of this application uses word segmentation methods of different granularities to segment the address text during the encoding stage, which can obtain richer original semantics. Subsequently, address feature vectors are obtained through deep semantic mining. The rich semantic representation capability can effectively prevent overfitting, make the entire neural network model easier to train to convergence, save training costs, and improve training efficiency.

[0150] Please see Figure 7 This application provides a communication address resolution service update device, which is a functional embodiment of the communication address resolution service update method of this application. The device includes: a self-built service invocation module 1100, a third-party service invocation module 1200, a result push and storage module 1300, and a timed training and update module 1400. The self-built service invocation module 1100 is configured to respond to a user-submitted communication address resolution request by invoking a predefined first resolution interface of the self-built communication address resolution service to process the request, thereby obtaining a first resolution result obtained after applying a neural network model to the self-built communication address resolution service. The neural network model is used to determine the standard address text corresponding to the input address text carried in the request from a standard database, which stores multiple standard address texts. The third-party service invocation module 1200 is configured to, when the first resolution... When the result indicates that the parsing fails, the second parsing interface predefined by the third-party communication address parsing service is invoked to process the communication address parsing request, so as to obtain the second parsing result determined according to the input address text carried in the request; the result push storage module 1300 is configured to push the standard address text and its parsing information of the input address text to the user when the first parsing result or the second parsing result contains the standard address text and its parsing information, and store it as a data record in the standard database, wherein the parsing information includes the postal code and / or latitude and longitude corresponding to the standard address text; the timed training update module 1400 is configured to respond to the timed task trigger event, apply the neural network model to determine the standard address text in the standard database that constitutes the positive sample according to the given address sample, use the positive sample to perform iterative training on the neural network model, and restart the service after training it to the convergence state.

[0151] Based on any embodiment of this application, prior to the timed training update module 1400, a full update decision module is included, configured to determine whether the standard database has reached a predetermined period since the last full update. When the predetermined period is reached, the module iterates through each data record in the standard database, uses the standard address text in each data record as a parameter, and calls the second parsing interface to obtain and update the parsing information corresponding to the standard address text in the data record.

[0152] Based on any embodiment of this application, the self-built service invocation module 1100 includes: a request-response parsing unit, configured to respond to a communication address parsing request submitted by a user and obtain the input address text in the communication address parsing request; a first cache query unit, configured to query whether parsing information corresponding to the input address text exists in a first cache corresponding to the self-built communication address parsing service, and push the parsing information to the user when the parsing information exists; a self-built service execution unit, configured to call a predefined first parsing interface of the self-built communication address parsing service to process the communication address parsing request when the parsing information does not exist in the first cache, so as to obtain a first parsing result containing standard address text and its corresponding parsing information; and a first cache storage unit, configured to store the standard address text and the input address text corresponding to the parsing information in the first cache when the first parsing result contains the parsing information.

[0153] Based on any embodiment of this application, the third-party service invocation module 1200 includes: a second cache query unit, configured to query whether there is parsing information corresponding to the entered address text in the second cache area corresponding to the third-party communication address parsing service when the first parsing result indicates parsing failure, and push the parsing information to the user when the parsing information exists; a third-party service execution unit, configured to call the second parsing interface predefined by the third-party communication address parsing service to process the communication address parsing request to obtain a second parsing result when the second parsing result contains the parsing information; and a second cache storage unit, configured to store the standard address text and the entered address text corresponding to the parsing information in the second cache area when the second parsing result contains the parsing information.

[0154] Based on any embodiment of this application, the self-built communication address resolution service is constructed as follows: a data retrieval module, configured to retrieve multiple standard address texts matching the entered address text from the standard database to form a candidate list; a score reasoning module, configured to call the neural network model to construct input information for each standard address text in the candidate list and the entered address text, and to determine the ranking score corresponding to the standard address text by the neural network model; a correct text determination module, configured to select the standard address text with the highest ranking score and exceeding a preset threshold from the candidate list as the correct text corresponding to the entered address text; a resolution execution module, configured to determine the latitude and longitude and / or postal code of the geographical location pointed to by the correct text; and a result return module, configured to return the standard address text corresponding to the correct text and the resolution information as the resolution information, and return it as the first resolution result.

[0155] Based on any embodiment of this application, the neural network model constructs the following modules during runtime: an encoding processing module, configured to use an encoding layer to encode the input address text and standard address text in the input information into embedding vectors respectively; a feature extraction module, configured to use a feature extraction layer to extract deep semantic information from the embedding vectors to obtain corresponding address feature vectors; a similarity calculation module, configured to use a linear layer to calculate the similarity between the address feature vectors to obtain a similarity vector; and a classification mapping module, configured to use a classifier to map the similarity vectors to a preset classification space to obtain the classification probability corresponding to the positive category as the ranking score of the standard address text in the input information.

[0156] To address the aforementioned technical problems, embodiments of this application also provide computer equipment. For example... Figure 8 The diagram shows the internal structure of a computer device. The computer device includes a processor, a computer-readable storage medium, a memory, and a network interface connected via a system bus. The computer-readable storage medium stores an operating system, a database, and computer-readable instructions. The database may store control information vectors. When the computer-readable instructions are executed by the processor, they enable the processor to implement a communication address resolution service update method. The processor of the computer device provides computing and control capabilities, supporting the operation of the entire computer device. The memory of the computer device may store computer-readable instructions. When these computer-readable instructions are executed by the processor, they enable the processor to execute the communication address resolution service update method of this application. The network interface of the computer device is used for communication with a terminal. Those skilled in the art will understand that… Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0157] In this embodiment, the processor is used to execute... Figure 7 The specific functions of each module and its submodules are defined within the device. The memory stores the program code and various data required to execute these modules or submodules. The network interface is used for data transmission between the user terminal and the server. In this embodiment, the memory stores the program code and data required to execute all modules / submodules in the communication address resolution service update device of this application. The server can call the server's program code and data to execute the functions of all submodules.

[0158] This application also provides a storage medium storing computer-readable instructions, which, when executed by one or more processors, cause the one or more processors to perform the steps of the communication address resolution service update method of any embodiment of this application.

[0159] This application also provides a computer program product, including a computer program / instructions that, when executed by one or more processors, implement the steps of the method described in any embodiment of this application.

[0160] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. This computer program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. The aforementioned storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0161] In summary, this application reduces the frequency of using third-party communication address resolution services by building a self-built communication address resolution service, and enables the reuse of historical data generated by third-party communication address resolution services. This allows for the provision of efficient and accurate communication address resolution services to e-commerce platforms at a relatively low cost.

[0162] Those skilled in the art will understand that the steps, measures, and solutions in the various operations, methods, and processes discussed in this application can be alternated, modified, combined, or deleted. Furthermore, other steps, measures, and solutions in the various operations, methods, and processes discussed in this application can also be alternated, modified, rearranged, decomposed, combined, or deleted. Furthermore, steps, measures, and solutions in the prior art that are similar to those disclosed in this application can also be alternated, modified, rearranged, decomposed, combined, or deleted.

[0163] The above description is only a partial embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for updating a communication address resolution service, characterized in that, This is applied to a server and is used to centrally process communication address resolution requests forwarded by various independent websites of an e-commerce platform. These independent websites are used to deploy online stores, and the communication address resolution requests are triggered within these online stores. The process includes the following steps: In response to a user-submitted communication address resolution request, the system calls a predefined first resolution interface of its self-built communication address resolution service to process the request. This allows the system to obtain a first resolution result obtained by applying a neural network model to the self-built communication address resolution service. The neural network model is used to determine the standard address text corresponding to the input address text carried in the request from a standard database. The standard database stores multiple standard address texts. The self-built communication address resolution service is implemented by performing the following steps: recalling multiple standard address texts matching the input address text from the standard database to form a candidate list; calling the neural network model to construct input information for each standard address text in the candidate list and the input address text, and determining the ranking score corresponding to the standard address text; selecting the standard address text with the highest ranking score exceeding a preset threshold from the candidate list as the correct text corresponding to the input address text; determining the latitude and longitude and / or postal code of the geographical location pointed to by the correct text; using the postal code and / or latitude and longitude as resolution information, returning the standard address text corresponding to the correct text and the resolution information as the first resolution result. When the first parsing result indicates parsing failure, the second parsing interface predefined by the third-party communication address parsing service is invoked to process the communication address parsing request, so as to obtain the second parsing result determined according to the input address text carried in the request; When the first parsing result or the second parsing result contains the standard address text and its parsing information of the entered address text, it is pushed to the user and stored as a data record in the standard database. The parsing information includes the postal code and / or latitude and longitude corresponding to the standard address text. In response to a scheduled task trigger event, the neural network model is applied to determine the standard address text in the standard database that constitutes a positive sample based on the given address sample. The neural network model is then iteratively trained using the positive sample until it reaches a convergent state, after which the service is restarted.

2. The communication address resolution service update method according to claim 1, characterized in that, Before responding to a scheduled task trigger event, the following should be included: Determine whether the standard database has reached a predetermined period since the last full update. When the predetermined period is reached, iterate through each data record in the standard database, and use the standard address text in each data record as a parameter to call the second parsing interface to obtain and update the parsing information corresponding to the standard address text in the data record.

3. The communication address resolution service update method according to claim 2, characterized in that, In response to a user-submitted communication address resolution request, the system calls a predefined first resolution interface of the self-built communication address resolution service to process the request, thereby obtaining a first resolution result obtained by applying a neural network model to the self-built communication address resolution service. The neural network model is used to determine, from a standard database, the standard address text corresponding to the input address text carried in the request. The standard database stores multiple standard address texts, including: In response to a user-submitted communication address resolution request, obtain the input address text from the communication address resolution request; The system queries the first cache corresponding to the self-built communication address resolution service to see if there is any resolution information corresponding to the entered address text. If the resolution information exists, the system pushes the resolution information to the user. When the first cache does not contain the parsing information, the first parsing interface predefined by the self-built communication address parsing service is called to process the communication address parsing request, so as to obtain a first parsing result containing standard address text and its corresponding parsing information. When the first parsing result contains the parsing information, the standard address text and the input address text corresponding to the parsing information are stored in the first cache area.

4. The communication address resolution service update method according to claim 1, characterized in that, When the first or second parsing result contains the standard address text and its parsing information of the entered address text, it is pushed to the user and stored as a data record in the standard database. The parsing information includes the postal code and / or latitude and longitude corresponding to the standard address text, including: When the first parsing result indicates parsing failure, the system queries the second cache corresponding to the third-party communication address parsing service to see if there is parsing information corresponding to the entered address text. If the parsing information exists, the system pushes the parsing information to the user. When the parsing information is not found in the second cache, the second parsing interface predefined by the third-party communication address parsing service is invoked to process the communication address parsing request in order to obtain the second parsing result. When the second parsing result contains the parsing information, the standard address text and the input address text corresponding to the parsing information are stored in the second cache area.

5. The communication address resolution service update method according to any one of claims 1 to 4, characterized in that, The neural network model performs the following steps: The application encoding layer encodes the input address text and standard address text in the input information into embedding vectors respectively; The application feature extraction layer extracts deep semantic information from the embedded vector to obtain the corresponding address feature vector; A linear layer is applied to calculate the similarity between the address feature vectors to obtain a similarity vector; The similarity vector is mapped to a preset classification space using a classifier, and the classification probability corresponding to the positive category is used as the ranking score of the standard address text in the input information.

6. A communication address resolution service update device, characterized in that, This is applied to a server for centrally processing communication address resolution requests forwarded by various independent websites of an e-commerce platform. These independent websites are used to deploy online stores, and the communication address resolution requests are triggered within these online stores, including: The self-built service invocation module is configured to respond to user-submitted communication address resolution requests. It invokes a predefined first resolution interface of the self-built communication address resolution service to process the request, obtaining a first resolution result after applying a neural network model. The neural network model is used to determine the standard address text corresponding to the input address text carried in the request from a standard database. The standard database stores multiple standard address texts. The self-built communication address resolution service is implemented by performing the following steps: recalling multiple standard address texts matching the input address text from the standard database to form a candidate list; invoking the neural network model to construct input information from each standard address text in the candidate list and the input address text, and determining the ranking score corresponding to the standard address text; selecting the standard address text with the highest ranking score exceeding a preset threshold from the candidate list as the correct text corresponding to the input address text; determining the latitude and longitude and / or postal code of the geographical location pointed to by the correct text; using the postal code and / or latitude and longitude as resolution information, returning the standard address text corresponding to the correct text and the resolution information as the first resolution result. The third-party service call module is configured to call the second parsing interface predefined by the third-party communication address parsing service to process the communication address parsing request when the first parsing result indicates that the parsing has failed, so as to obtain the second parsing result determined according to the input address text carried in the request; The result push storage module is configured to push the standard address text and its parsing information to the user when the first parsing result or the second parsing result contains the standard address text of the entered address text, and store it as a data record in the standard database. The parsing information includes the postal code and / or latitude and longitude corresponding to the standard address text. The scheduled training and update module is configured to respond to scheduled task trigger events. Based on the given address samples, the neural network model is used to determine the standard address text in the standard database, which together constitute positive samples. The neural network model is iteratively trained using the positive samples until it reaches a convergent state, and then the service is restarted.

7. The communication address resolution service update apparatus according to claim 6, characterized in that, The device also includes a full update decision module, configured to determine whether the standard database has reached a predetermined period since the last full update. When the predetermined period is reached, the module iterates through each data record in the standard database, uses the standard address text in each data record as a parameter, and calls the second parsing interface to obtain and update the parsing information corresponding to the standard address text in the data record.

8. A computer device comprising a central processing unit and a memory, characterized in that, The central processing unit is used to invoke and run a computer program stored in the memory to perform the steps of the method as described in any one of claims 1 to 5.

9. A computer-readable storage medium, characterized in that, It stores, in the form of computer-readable instructions, a computer program implemented according to any one of claims 1 to 5, which, when invoked by a computer, executes the steps included in the corresponding method.

10. A computer program product, characterized in that, Includes a computer program / instructions that, when executed by a processor, implement the steps of the method according to any one of claims 1 to 5.

Images