Store information processing method, device, equipment, storage medium and program product

Abstract

This application discloses a method, apparatus, device, storage medium, and program product for processing store information. The method involves: acquiring store information of a target store, including the store name and address; determining the brand information and industry category of the target store based on the store name; searching a set of potential duplicate stores in an existing store information database whose brand information and industry category both match the target store; determining a target distance threshold corresponding to the industry category of the target store; and identifying stores in the potential duplicate store set whose addresses are less than the target store address as duplicate stores of the target store. According to the embodiments of this application, the accuracy of duplicate store identification can be improved.

Description

Technical Field

[0001] This application belongs to the field of data processing technology, and in particular relates to a method, apparatus, equipment, storage medium and program product for processing store information. Background Technology

[0002] For a service platform connecting consumers and brick-and-mortar merchants, collecting store information is necessary when conducting offline consumption subsidies and membership discounts for consumers. However, when store information is submitted from different sources, these sources may submit information for the same store, and the information for the same store submitted by different sources may differ. This can lead to the same store being mistakenly identified as two different stores based on the same information, meaning the same store is repeatedly counted. The store information collection process needs to accurately identify duplicate stores to ensure that each physical merchant corresponds to only one accurate store information record, allowing marketing resources to be precisely targeted to the actual physical merchants.

[0003] In related technologies, duplicate stores are mainly identified based on the similarity of store names and the distance between store locations. However, this method usually uses comparison with a fixed threshold to make the judgment, which has the problem of low accuracy. Summary of the Invention

[0004] This application provides a store information processing method, apparatus, equipment, storage medium, and program product that can improve the accuracy of duplicate store identification.

[0005] In a first aspect, embodiments of this application provide a store information processing method, including: Obtain the store information of the target store, including the store name and store address; Based on the store name, determine the brand information and industry category of the target store; Find a set of potential duplicate stores in the existing store information database whose brand information and industry category are consistent with the target store; Determine the target distance threshold corresponding to the industry category of the target store; Stores in the potential duplicate store set whose store address is less than the target store's store address are identified as duplicate stores of the target store.

[0006] Secondly, embodiments of this application provide a store information processing device, including: The information acquisition module is used to acquire the store information of the target store, including the store name and store address. The information extraction module is used to determine the brand information and industry category of a target store based on its store name. The potential store identification module is used to find a set of potential duplicate stores that match the target store in terms of brand information and industry category from the existing store information database. The threshold determination module is used to determine the target distance threshold corresponding to the industry category to which the target store belongs; The duplicate store identification module is used to identify stores in the potential duplicate store set whose store address is less than the target store address as duplicate stores of the target store.

[0007] Thirdly, embodiments of this application provide an electronic device, which includes: a processor and a memory storing computer program instructions; When the processor executes computer program instructions, it implements the store information processing method as described in the first aspect.

[0008] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer program instructions, which, when executed by a processor, implement the store information processing method as described in the first aspect.

[0009] Fifthly, embodiments of this application provide a computer program product in which instructions, when executed by the processor of an electronic device, cause the electronic device to perform the store information processing method as described in the first aspect.

[0010] In this embodiment, the store information of the target store is obtained, including the store name and store address. Based on the store name, the brand information and industry category of the target store are determined. A set of potential duplicate stores whose brand information and industry category are consistent with the target store are searched from the existing store information database. A target distance threshold corresponding to the industry category to which the target store belongs is determined. Stores in the potential duplicate store set whose store address is less than the target distance threshold are identified as duplicate stores of the target store. According to this embodiment, the brand and industry, two core elements of the store, are extracted based on the store name. The consistency judgment of the two core elements, brand and industry, is used instead of the traditional full-field similarity judgment of the store name, which improves the accuracy of duplicate store identification. In addition, different distance thresholds are used for different industries to determine whether stores are duplicates, which further improves the accuracy of duplicate store identification. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1This is a flowchart illustrating a store information processing method provided in some embodiments of this application; Figure 2 This is a schematic diagram of the model training process provided in some embodiments of this application; Figure 3 This is a schematic diagram of the process for identifying duplicate stores provided in some embodiments of this application; Figure 4 This is a flowchart illustrating a store information processing method provided in some embodiments of this application; Figure 5 This is a flowchart illustrating a store information processing method provided in some embodiments of this application; Figure 6 This is a schematic diagram of the process for merging store information provided in some embodiments of this application; Figure 7 This is a schematic diagram of the structure of a store information processing device provided in some embodiments of this application; Figure 8 These are schematic diagrams of the structure of electronic devices provided in some embodiments of this application. Detailed Implementation

[0013] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0014] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0015] Before providing a further detailed description of the embodiments of this application, the nouns and terms involved in the embodiments of this application will be explained, and the nouns and terms involved in the embodiments of this application shall be interpreted as follows.

[0016] BERT (Bidirectional Encoder Representations from Transformers) is a pre-trained language model that uses a bidirectional Transformer encoder structure. It can simultaneously utilize the left and right contextual information of the input sequence to more accurately understand the meaning of language. It is widely used in text classification, entity recognition, relation extraction, and other applications.

[0017] Before providing a more detailed description of the embodiments of this application, the relevant technologies will be introduced. As mentioned above, the relevant technologies mainly rely on fixed rules to determine duplicate stores. For example, if the geographical distance between two stores and the similarity of the full field of the store name both reach the threshold (e.g., the distance between two stores is within 300 meters and the name similarity is above 0.8), they are considered duplicate stores.

[0018] Store names are often composed of elements such as province, city, district, brand, industry, and business district, such as "City A, District B, Convenience Store C (D Park Store)". When identifying duplicate stores, the main focus is on whether different store information corresponds to the same entity, and the key is to identify whether their operating entities and main businesses are the same. However, the province, city, district, and business district in the store name are location attributes; they are variable and descriptive. The same location can be described in different ways. For example, "City A, District B, D Park, Convenience Store C" and "C Convenience Store (D Park Store)" are the same store, but their descriptions differ significantly. If full-field similarity of store names is used for duplicate store identification, they will be judged as different stores, leading to misjudgment. Therefore, the above method of identifying duplicate stores using full-field similarity of store names has low accuracy.

[0019] In addition, the above identification methods usually use a uniform distance threshold for all stores. However, the density of stores varies in different industries. Using the same distance threshold to determine whether a store is a duplicate will result in inaccurate identification results.

[0020] In view of this, in order to improve the accuracy of duplicate store identification, embodiments of this application provide a store information processing method, apparatus, equipment, medium, and program product. The aim is to extract core elements (brand, industry classification) for identifying duplicate stores from store names, replace full-field similarity judgment of store names with consistency judgment of brand and industry elements, and apply different distance thresholds for duplicate store judgment for stores in different industries, thereby improving the accuracy of duplicate store identification.

[0021] The store information processing method provided in this application can be applied to scenarios requiring the processing of physical merchant data, such as, but not limited to, marketing scenarios of financial service platforms. The store information processing method provided in this application will be described in detail below with reference to the accompanying drawings. It should be noted that the executing entity of the store information processing method provided in this application can be a store information processing device, such as a financial service platform. This application uses the execution of the store information processing method by a store information processing device as an example to illustrate the store information processing method provided in this application.

[0022] See Figure 1 The above is a flowchart illustrating a store information processing method provided in some embodiments of this application, such as... Figure 1 As shown, the method may include steps 110-150, which will be described in detail below.

[0023] Step 110. Obtain the store information of the target store, including the store name and store address.

[0024] In some embodiments of this application, the store information processing device can acquire store information provided by different channels (such as chain brands, service providers, banks, industry teams in the customer service department, branch offices, etc.). The store information of the target store can be the latest store information provided to the store information processing device by any channel. After acquiring the store information, the store information processing device processes the store information to determine whether the store represented by the store information, i.e., the target store, is a duplicate of an existing store.

[0025] In some embodiments of this application, the store information for each store may include the store name and store address.

[0026] Step 120. Based on the store name, determine the brand information and industry category of the target store.

[0027] In some embodiments of this application, the store name may include elements such as province, city, district, brand, industry, and business district. Among these, brand and industry are key elements for identifying duplicate stores, while province, city, district, and business district are non-key elements. Therefore, to avoid the influence of non-key elements on duplicate store identification, after obtaining the store information of the target store, the brand information and industry category of the target store are extracted from the store name.

[0028] In some embodiments of this application, in order to accurately and quickly extract brand information and industry category from store names, artificial intelligence technology can be used to pre-train a brand extraction model for extracting brand information from store names and an industry classification model for extracting the industry category of a store from its name. Thus, in step 120 above, brand information can be extracted from store names using the pre-trained brand extraction model; and the store names can be classified by industry using the pre-trained industry classification model to determine the industry category to which the target store belongs.

[0029] In some embodiments of this application, the BERT model has entity extraction capabilities. Based on this, a brand extraction model can be obtained by fine-tuning the BERT model to extract brand information from store names using its entity extraction capabilities. Specifically, the brand extraction model can be trained using steps 210-220 as follows.

[0030] Step 210. Obtain the first training dataset, which includes samples of store names with brand information.

[0031] In some embodiments of this application, multiple (e.g., 100,000) store name samples can be obtained from an external open-source database or an internal database, and the brand information of each store name sample can be labeled, thereby obtaining multiple store name samples with brand information labels. The dataset composed of multiple store name samples with brand information labels is used as the first training dataset.

[0032] In some embodiments of this application, in order to enable the brand extraction model to extract brand information from store names in various formats, the store name samples in the first training dataset may include store name samples of various categories such as merchant business name samples and merchant registration name samples.

[0033] A merchant's registered name refers to the name that a physical business registers with the business registration authority (such as the Administration for Industry and Commerce / Market Supervision and Management Department). The merchant registered name samples in the first training dataset can be obtained from publicly available business registration information.

[0034] A business name refers to the name of a physical store that is presented to consumers and used for daily operations and external display. It is usually characterized by high recognizability and simplicity, rather than the full legal name of the business registered with the enterprise registration authority. It omits or simplifies legally redundant information such as administrative divisions and organizational forms (such as "limited company" or "individual business") commonly found in registered names.

[0035] In some embodiments of this application, the merchant's business name is typically composed of one or more business elements such as brand information, industry information, and business district name. For example, the merchant's business name may include at least one of the following formats: brand information, brand information + industry information, brand information + business district name, or brand information + industry information + business district name. For instance, if C is a brand name, D Park is a business district name, and convenience store represents the industry category to which the store belongs, then the merchant's business name could be "A", "A Convenience Store", "A (D Park Store)", or "A Convenience Store (D Park Store)".

[0036] In some embodiments of this application, the merchant business name samples in the first training dataset can be collected from publicly available merchant information such as electronic maps or local life service platforms, or the merchant business name samples can be generated based on the obtained merchant registration name samples, or they can be obtained through on-site collection.

[0037] In some embodiments of this application, when generating merchant business name samples based on obtained merchant registration name samples, redundant fields (such as administrative divisions, organizational forms, etc.) in the obtained merchant registration name samples can be automatically removed by an algorithm based on the composition of the merchant business names, thereby generating a corresponding candidate set of merchant business names. Then, the candidate set of merchant business names is manually verified, and the names that pass the verification in the candidate set of merchant business names are used as merchant registration name samples. Among them, manual verification is mainly used to verify whether the name conforms to the composition of a merchant business name.

[0038] In some embodiments of this application, the brand information annotation for each store name sample in the first training dataset can be manually added by engineers. This ensures the accuracy of the brand information annotation.

[0039] Step 220. Based on the first training dataset, fine-tune the bidirectional transformer BERT model to obtain the brand extraction model.

[0040] In some embodiments of this application, after obtaining the first training dataset, it is divided into a first training set and a first validation set according to a set ratio (e.g., 8:2). The BERT model is then trained iteratively multiple times using the first training set. The training objective is to minimize the difference between the brand information predicted by the BERT model and the actual brand information labels of the store name samples. After each iteration, the first loss value of the trained BERT model is calculated according to a set first loss function, and the parameters of the BERT model are updated based on the first loss value until the first convergence condition is met. Training then stops, and the trained BERT model is used as the brand extraction model. The brand extraction model can then be further validated using the first validation set. The final brand extraction model is deployed in a store information processing device, allowing the device to extract brand information from store names using the brand extraction model.

[0041] In some embodiments of this application, the first loss function may be the cross-entropy loss function. Optimizing the model parameters based on the cross-entropy loss function enables the final brand extraction model to achieve a statistical accuracy of 94.6% for test samples with a confidence score of 0.98 or higher. The confidence score is used to indicate the probability estimate of the correctness of the model's output prediction. The confidence score is typically presented as a numerical value between 0 and 1; the closer the value is to 1, the more confident the model is in the accuracy of its output prediction.

[0042] In some embodiments of this application, the first convergence condition may include at least one of the following: The first loss value of the BERT model is greater than the set first loss threshold; The number of iterations has reached the first iteration count.

[0043] In some embodiments of this application, the BERT model has text classification capabilities. Based on this, an industry classification model can be obtained by fine-tuning the BERT model to determine the industry category of a store based on its name, utilizing the text classification capabilities of the BERT model. Specifically, the industry classification model can be trained using steps 310-320 as follows.

[0044] Step 310. Obtain the second training dataset, which includes samples of store names labeled with industry categories.

[0045] In some embodiments of this application, multiple (e.g., 500,000) store name samples can be obtained from external open-source databases or internal databases, and the industry category (e.g., food, shopping, accommodation, tourism, education, medical care, etc.) of each store name sample can be labeled, thereby obtaining multiple store name samples with industry category labels. The dataset composed of multiple store name samples with industry category labels is used as the second training dataset.

[0046] In some embodiments of this application, similar to the first training dataset, in order to enable the industry classification model to determine the industry category to which a store belongs based on store names in various different formats, the store name samples in the second training dataset may include store name samples of various categories such as merchant registration name samples and merchant business name samples.

[0047] In some embodiments of this application, the industry category corresponding to each store name sample in the second training dataset is the industry category to which the store corresponding to that store name sample belongs. Engineers can manually add the corresponding industry category label to each store name sample. This ensures the accuracy of the industry category labeling.

[0048] Step 320. Based on the second training dataset, fine-tune the BERT model to obtain the industry classification model.

[0049] In some embodiments of this application, after obtaining the second training dataset, it can be divided into a second training set and a second validation set according to a set division ratio. The BERT model is then trained iteratively multiple times using the second training set. The training objective is to minimize the difference between the industry category predicted by the BERT model and the actual industry category labeling of the store name samples. After each iteration, the second loss value of the trained BERT model is calculated according to a set second loss function, and the parameters of the BERT model are updated based on the second loss value until the second convergence condition is met. Training then stops, and the trained BERT model is used as the industry classification model. Subsequently, the industry classification model can be further validated based on the second validation set. Finally, the trained industry classification model is deployed on a store information processing device, allowing the device to determine the industry category of a store based on its name using this model.

[0050] In some embodiments of this application, the second loss function may be the cross-entropy loss function. Optimizing the model parameters based on the cross-entropy loss function enables the final industry classification model to achieve a statistical accuracy of 99.3% for test samples with a confidence score of 0.98 or higher.

[0051] In some embodiments of this application, the second convergence condition may include at least one of the following: The second loss value of the BERT model is greater than the set second loss threshold; The number of iterations has reached the second iteration count.

[0052] In some embodiments of this application, the trained brand extraction model and industry classification model can be encapsulated as independent HTTP services and deployed on the store information processing device. In this way, the store information processing device can extract brand information and industry category from the store name by calling the HTTP service interface of the aforementioned model, thereby completing the identification of duplicate stores. Here, HTTP service refers to a remote service that can be invoked via a network address (URL) using a standard HTTP request.

[0053] See Figure 2 This is a schematic diagram illustrating the training process of models such as the brand extraction model and industry classification model provided in the embodiments of this application. The result file refers to the result file generated during the model fine-tuning process.

[0054] Step 130. Find a set of potential duplicate stores in the existing store information database that have the same brand information and industry category as the target store.

[0055] In some embodiments of this application, the store information processing device can store the processed store information in a store information database. This database can include processed store information corresponding to one or more stores. Each store's information can include its store name and address, as well as brand information extracted from the store name and the store's industry category. Based on this, after obtaining the target store's brand information and industry category, the store information database can be traversed based on these information to find one or more stores whose brand information and industry category match the target store. Since stores whose brand information and industry category match the target store may be duplicates, the set of one or more stores found based on the target store's brand information and industry category is defined as a potential duplicate store set.

[0056] Step 140. Determine the target distance threshold corresponding to the industry category to which the target store belongs.

[0057] In some embodiments of this application, distance thresholds corresponding to different industry categories are preset, and a correspondence between industry categories and distance thresholds is established. Based on this, after determining the industry category to which the target store belongs, the distance threshold corresponding to the industry category to which the target store belongs can be determined based on the above correspondence, and the distance threshold is determined as the target distance threshold.

[0058] In some embodiments of this application, the distance threshold corresponding to each industry category can be determined by statistical analysis of the spatial distribution characteristics of stores within each industry, or the distance threshold corresponding to each industry category can be set based on industry experience, and the distance threshold corresponding to different industries can be different.

[0059] In some embodiments of this application, the method for determining the distance threshold may include: obtaining manually verified store addresses of the same brand within the same industry; calculating the geographical distance between all pairs of stores identified as the same entity; and using a specific high quantile (such as the 95th or 99th percentile) of this distance distribution as the distance threshold for that industry. For example, if statistics for the 'convenience store' industry show that 95% of stores of the same entity are less than 150 meters apart, then the distance threshold for that industry can be set to 150 meters. For industries such as 'gas stations', a larger threshold (such as 500 meters) can be set based on their service radius characteristics, using similar statistical methods or combining expert experience.

[0060] By setting corresponding distance thresholds for each industry, different distance thresholds can be used to determine whether stores are duplicates for different industries, resulting in higher accuracy.

[0061] Step 150. Identify stores in the potential duplicate store set whose store addresses are less than the target store address by a distance less than the target distance threshold as duplicate stores of the target store.

[0062] Duplicate stores refer to stores that are identical to the target store. After obtaining the set of potential duplicate stores and the target distance threshold, the system further determines, based on the target distance threshold, whether the stores in the set of potential duplicate stores are indeed duplicates of the target store.

[0063] In some embodiments of this application, the distance between the store address of each store in the potential duplicate store set and the store address of the target store can be calculated separately. Each calculated distance is compared with the target distance threshold to obtain the comparison result. Based on the comparison result, stores in the potential duplicate store set whose address location distance from the target store is less than the target distance threshold are identified as duplicate stores of the target store.

[0064] See Figure 3 This is a flowchart illustrating the process of identifying duplicate stores of a target store based on a brand extraction model and an industry classification model.

[0065] In this embodiment, the store information of the target store is obtained, including the store name and store address. Based on the store name, the brand information and industry category of the target store are determined. A set of potential duplicate stores whose brand information and industry category are consistent with the target store are searched from the existing store information database. A target distance threshold corresponding to the industry category to which the target store belongs is determined. Stores in the potential duplicate store set whose store address is less than the target distance threshold are identified as duplicate stores of the target store. According to this embodiment, the brand and industry, two core elements of the store, are extracted based on the store name. The consistency judgment of the two core elements, brand and industry, is used instead of the traditional full-field similarity judgment of the store name, which improves the accuracy of duplicate store identification. In addition, different distance thresholds are used for different industries to determine whether stores are duplicates, which further improves the accuracy of duplicate store identification.

[0066] In some embodiments of this application, if duplicate stores of the target store are found, the store information can be further merged to ensure that a physical store corresponds to unique store information, thereby supporting the technical implementation of specific businesses such as precision marketing and business analysis.

[0067] In some embodiments of this application, store information merging includes store name merging. See also Figure 4 The store information processing method provided in this application embodiment may further include the following steps 160-170.

[0068] Step 160. If at least one duplicate store is identified, classify the store name of the target store and the store name of each duplicate store using a pre-trained store name classification model to obtain classification results. The classification results are used to indicate whether the store name belongs to the category of merchant business name or merchant registration name.

[0069] In some embodiments of this application, when at least one duplicate store that is the same as the target store is identified, the store name of the target store and the store name of the duplicate store can be merged to obtain the store name of the physical store corresponding to the target store, i.e., the target store name.

[0070] In some embodiments of this application, store names can generally be divided into two categories: merchant operating names and non-merchant operating names (such as merchant registration names). Merchant operating names typically use concise and commonly used words, which are highly recognizable and have the characteristics of being more intuitive, easier to understand, and easier to spread compared to other types of store names. Therefore, in order to obtain intuitive, easy-to-understand, and easy-to-spread target store names, before merging the store names of the target stores and each duplicate store, the category of each store name is determined separately so as to determine the target store name that conforms to the user's cognitive habits based on the category of the store name.

[0071] In some embodiments of this application, a store name classification model is used to classify the store name of the target store and the store name of each duplicate store separately, thereby obtaining classification results. The store name classification model is a pre-trained model used to determine the category of the store name.

[0072] In some embodiments of this application, the BERT model possesses text classification capabilities. Based on this, the BERT model can be fine-tuned and trained to obtain a store name classification model that utilizes the text classification capabilities of the BERT model to identify the category to which the store name belongs. The store name classification model can also employ methods such as... Figure 2 The training process shown is followed. Specifically, steps 410-420 can be used to train the store name classification model.

[0073] Step 410. Obtain the third training dataset, which includes sample data pairs. Each sample data pair includes the merchant's business name and non-merchant's business name of a store sample.

[0074] In some embodiments of this application, the third training dataset includes multiple (e.g., 100,000) sample data pairs, each including the merchant's business name and the merchant's registered name for a store sample. A store sample represents a physical store, and the merchant's business name is used as a positive sample, while the non-business name is used as a negative sample, thus constructing sample data pairs containing positive and negative samples. In this way, the model can accurately learn the essential differences in text structure and semantics between names of the same business entity but for different purposes.

[0075] In some embodiments of this application, the 'merchant business name' of the store sample can be obtained from the public consumer platform as a positive sample, and the 'merchant registration name' of the store sample filing can be obtained from the market entity registration authority (such as the market supervision bureau) as its corresponding negative sample.

[0076] Step 420. Fine-tune the BERT model based on the third training dataset to obtain the store name classification model.

[0077] In some embodiments of this application, after obtaining the third training dataset, it can be divided into a third training set and a third validation set according to a set partitioning ratio. The BERT model is then trained iteratively multiple times using the third training set for text classification fine-tuning. The training objective is to enable the BERT model to accurately distinguish whether an input store name is a business name. After each iteration, the third loss value of the trained BERT model is calculated according to a set third loss function, and the parameters of the BERT model are updated based on this third loss value until the third convergence condition is met. Training then stops, and the trained BERT model is used as the store name classification model. Subsequently, the store name classification model can be further validated using the third validation set. The final store name classification model is then deployed in a store information processing device, allowing the device to determine the category of store names using the model.

[0078] In some embodiments of this application, the training objective of the store name classification model is a multi-classification task; therefore, the third loss function can be the multivariate cross-entropy loss. Optimizing the parameters of the BERT model based on the multivariate cross-entropy loss can achieve a statistical accuracy of 91.6% for test samples with scores above 0.99 in the final trained store name classification model.

[0079] In some embodiments of this application, the third convergence condition may include at least one of the following: The third loss value of the BERT model is greater than the set third loss threshold; The number of iterations has reached the third iteration.

[0080] Step 170. Based on the classification results, merge the store name of the target store and the store name of at least one duplicate store into one target store name.

[0081] In some embodiments of this application, based on the classification results, a merchant's business name may be selected as the target store name from the store name of the target store and the store name of at least one duplicate store.

[0082] In some embodiments of this application, step 170 above may specifically determine the target store name based on the following rules: If only one of the target store's store name and at least one duplicate store's store name is classified as a merchant's business name, the store name classified as a merchant's business name shall be determined as the target store name. If at least two of the store names of the target store and at least one duplicate store are classified as merchant business names, the store name with the highest confidence score among the at least two store names classified as merchant business names shall be identified as the target store name, wherein the confidence score is used to indicate the correct probability of the classification result output by the store name classification model. If the store name of the target store and the store name of at least one duplicate store are both classified as merchant registration names, the store name with the lowest confidence score among the store name of the target store and the store name of at least one duplicate store shall be determined as the target store name.

[0083] By using the above-mentioned store name merging method, a target store name that is more in line with user perception, more intuitive, and has a better display effect can be determined as the store name of physical merchants.

[0084] In some embodiments of this application, the non-merchant business names in the third training dataset may include not only the merchant registration names of the store samples, but also incomplete names (such as those without brand or industry information) and substandard names (such as names that cannot identify the store and are composed of personal names, business districts, or other words). Thus, the trained store name classification model can identify not only merchant business names and merchant registration names, but also incomplete and substandard names. Based on this, before step 170, the store names of the target store and each duplicate store can be verified based on the classification results obtained in step 160 to determine incomplete or substandard names. Incomplete and substandard names are discarded, and step 170 is only performed on store names that pass the verification (i.e., are not "incomplete" or "substandard" names).

[0085] By adding the above verification steps, incomplete and substandard names can be filtered out, thus avoiding the identification of incomplete or substandard names as target store names and improving the quality of the final merged target store names.

[0086] In some embodiments of this application, store information merging also includes store address merging. See [link to relevant documentation]. Figure 5 The store information processing method provided in this application embodiment may further include the following steps 180-190.

[0087] Step 180. Determine the address precision of the target store's address and the address precision of each duplicate store's address.

[0088] In some embodiments of this application, the address precision of a store address is used to indicate how well the store address matches, is detailed, and accurate in relation to its actual geographical location. The more detailed the address, the higher the address precision. Generally, addresses containing complete hierarchical information such as 'province, city, district, street, and house number' have high precision, while addresses containing only high-level information such as 'province and city' have low precision. For example, the precision of the address "Province F, City N" is lower than that of the address "Province F, City N, District M, Street R".

[0089] In some embodiments of this application, for each store address, a map geocoding service can be used to determine the address level of the store address, and the address level is used to represent the address precision. A higher address level indicates higher address precision.

[0090] Step 190. Determine the store address with the highest address precision among the store address of the target store and the store address of at least one duplicate store as the store address corresponding to the target store name.

[0091] By using the above method, the addresses with the highest accuracy are prioritized and output to ensure the availability of store addresses after merging.

[0092] In some embodiments of this application, the target store name and the store address corresponding to the target store name can be stored in a store information database for easy use later.

[0093] See Figure 6 This is a schematic diagram illustrating the process of merging store information in an embodiment of this application.

[0094] Based on the store information processing method provided in the above embodiments, this application also provides specific implementations of the store information processing device. Please refer to the following embodiments.

[0095] See Figure 7 The store information processing device 700 provided in this application embodiment includes the following modules: The information acquisition module 701 is used to acquire the store information of the target store, including the store name and store address; The information extraction module 702 is used to determine the brand information and industry category of the target store based on the store name; The potential store identification module 703 is used to find a set of potential duplicate stores from the existing store information database whose brand information and industry category are consistent with the target store; The threshold determination module 704 is used to determine the target distance threshold corresponding to the industry category to which the target store belongs; The duplicate store determination module 705 is used to determine stores in the potential duplicate store set whose store address is less than the target distance threshold as duplicate stores of the target store.

[0096] In this embodiment, the store information of the target store is obtained, including the store name and store address. Based on the store name, the brand information and industry category of the target store are determined. A set of potential duplicate stores whose brand information and industry category are consistent with the target store are searched from the existing store information database. A target distance threshold corresponding to the industry category to which the target store belongs is determined. Stores in the potential duplicate store set whose store address is less than the target distance threshold are identified as duplicate stores of the target store. According to this embodiment, the brand and industry, two core elements of the store, are extracted based on the store name. The consistency judgment of the two core elements, brand and industry, is used instead of the traditional full-field similarity judgment of the store name, which improves the accuracy of duplicate store identification. In addition, different distance thresholds are used for different industries to determine whether stores are duplicates, which further improves the accuracy of duplicate store identification.

[0097] In some embodiments of this application, the information extraction module 702 is specifically used for: Brand information is extracted from the store names using a pre-trained brand extraction model; The store name is classified by industry using a pre-trained industry classification model to determine the industry category to which the target store belongs.

[0098] In some embodiments of this application, the training method of the brand extraction model includes: Obtain a first training dataset, which includes samples of store names with brand information; Based on the first training dataset, the bidirectional transformer BERT model is fine-tuned and trained to obtain the brand extraction model.

[0099] In some embodiments of this application, the training method of the industry classification model includes: Obtain a second training dataset, which includes samples of store names labeled with industry categories; Based on the second training dataset, the BERT model is fine-tuned to obtain the industry classification model.

[0100] In some embodiments of this application, the apparatus 700 may further include: a name merging module, used for: If at least one of the duplicate stores is identified, the store name of the target store and the store name of each of the duplicate stores are classified by a pre-trained store name classification model to obtain a classification result. The classification result is used to indicate whether the store name is a merchant business name or a non-merchant business name. Based on the classification results, the store name of the target store and the store name of the at least one duplicate store are merged into a single target store name.

[0101] In some embodiments of this application, the name merging module is specifically used for: If only one of the store names of the target store and the store names of the at least one duplicate store is classified as a merchant business name, the store name classified as a merchant business name shall be determined as the target store name. If at least two of the store names of the target store and the store names of the at least one duplicate store are classified as merchant business names, the store name with the highest confidence score among the at least two store names classified as merchant business names shall be determined as the target store name, wherein the confidence score is used to indicate the correct probability of the classification result output by the store name classification model. If both the target store name and the at least one duplicate store name are classified as non-merchant business names, the store name with the lowest confidence score among the target store name and the at least one duplicate store name shall be determined as the target store name.

[0102] In some embodiments of this application, the training method of the store name classification model includes: Obtain a third training dataset, which includes sample data pairs, each of which includes the merchant business name and non-merchant business name of a store sample; The BERT model was fine-tuned and trained based on the third training dataset to obtain the store name classification model.

[0103] In some embodiments of this application, the merchant business name includes at least one of the following components: brand information, brand information and industry information, brand information and business district name, and brand information, industry information and business district name.

[0104] In some embodiments of this application, the apparatus 700 may further include: an address merging module, used for: If at least one of the duplicate stores is identified, the address precision of the target store's address and the address precision of each of the duplicate stores' addresses are determined respectively. The store address with the highest address precision among the store address of the target store and the store addresses of at least one duplicate store is determined as the store address corresponding to the target store name.

[0105] The store information processing device provided in this application embodiment can achieve... Figures 1 to 6 The various processes implemented in the method implementation examples will not be described again here to avoid repetition.

[0106] Figure 8 A schematic diagram of the hardware structure of the electronic device provided in an embodiment of this application is shown.

[0107] Electronic device 800 may include processor 801 and memory 802 storing computer program instructions.

[0108] Specifically, the processor 801 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0109] Memory 802 may include a large-capacity memory for data or instructions. For example, and not limitingly, memory 802 may include a hard disk drive (HDD), a floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 802 may include removable or non-removable (or fixed) media. Where appropriate, memory 802 may be internal or external to electronic device 800. In a particular embodiment, memory 802 is a non-volatile solid-state memory. Memory 802 may include read-only memory (ROM), random access memory (RAM), disk storage media devices, optical storage media devices, flash memory devices, electrical, optical, or other physical / tangible memory storage devices. Thus, generally, memory 802 includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it performs the operations described in any of the store information processing methods in the above embodiments.

[0110] The processor 801 reads and executes computer program instructions stored in the memory 802 to implement any of the store information processing methods in the above embodiments.

[0111] In one example, the electronic device 800 may further include a communication interface 803 and a bus 810. For example, Figure 8 As shown, the processor 801, memory 802, and communication interface 803 are connected through bus 810 and complete communication with each other.

[0112] The communication interface 803 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0113] Bus 810 includes hardware, software, or both, that couples components of electronic device 800 together. For example, and not limitingly, the bus may include Accelerated Graphics Port (AGP) or other graphics buses, Enhanced Industry Standard Architecture (EISA) buses, Front Side Bus (FSB), HyperTransport (HT) interconnects, Industry Standard Architecture (ISA) buses, Infinite Bandwidth Interconnects, Low Pin Count (LPC) buses, memory buses, Microchannel Architecture (MCA) buses, Peripheral Component Interconnect (PCI) buses, PCI-Express (PCI-X) buses, Serial Advanced Technology Attachment (SATA) buses, Video Electronics Standards Association Local (VLB) buses, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 810 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, any suitable bus or interconnect is contemplated herein.

[0114] Furthermore, in conjunction with the store information processing methods in the above embodiments, this application embodiment can provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the store information processing methods in the above embodiments.

[0115] This application also provides a computer program product, including a computer program, which, when executed, implements any of the store information processing methods described in the above embodiments.

[0116] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0117] The functional blocks shown in the above-described structural diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0118] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0119] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0120] The above description is merely a specific implementation of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A method for processing store information, characterized in that, include: Obtain the store information of the target store, including the store name and store address; Based on the store name, determine the brand information and industry category of the target store; Find a set of potential duplicate stores in the existing store information database whose brand information and industry category are consistent with the target store; Determine the target distance threshold corresponding to the industry category to which the target store belongs; Stores in the potential duplicate store set whose store address is less than the target distance threshold from the store address of the target store are identified as duplicate stores of the target store.

2. The method according to claim 1, characterized in that, The process of determining the brand information and industry category of the target store based on the store address includes: Brand information is extracted from the store names using a pre-trained brand extraction model; The store name is classified by industry using a pre-trained industry classification model to determine the industry category to which the target store belongs.

3. The method according to claim 2, characterized in that, The training methods for the brand extraction model include: Obtain a first training dataset, which includes samples of store names with brand information; Based on the first training dataset, the bidirectional transformer BERT model is fine-tuned and trained to obtain the brand extraction model.

4. The method according to claim 2, characterized in that, The training methods for the industry classification model include: Obtain a second training dataset, which includes samples of store names labeled with industry categories; Based on the second training dataset, the BERT model is fine-tuned to obtain the industry classification model.

5. The method according to any one of claims 1-4, characterized in that, The method further includes: If at least one of the duplicate stores is identified, the store name of the target store and the store name of each of the duplicate stores are classified by a pre-trained store name classification model to obtain a classification result. The classification result is used to indicate whether the store name is a merchant business name or a non-merchant business name. Based on the classification results, the store name of the target store and the store name of the at least one duplicate store are merged into a single target store name.

6. The method according to claim 5, characterized in that, The step of merging the store name of the target store and the store names of the at least one duplicate store into a single target store name based on the classification result includes: If only one of the store names of the target store and the store names of the at least one duplicate store is classified as a merchant business name, the store name classified as a merchant business name shall be determined as the target store name. If at least two of the store names of the target store and the store names of the at least one duplicate store are classified as merchant business names, the store name with the highest confidence score among the at least two store names classified as merchant business names shall be determined as the target store name, wherein the confidence score is used to indicate the correct probability of the classification result output by the store name classification model. If both the target store name and the at least one duplicate store name are classified as non-merchant business names, the store name with the lowest confidence score among the target store name and the at least one duplicate store name shall be determined as the target store name.

7. The method according to claim 5, characterized in that, The training methods for the store name classification model include: Obtain a third training dataset, which includes sample data pairs, each of which includes the merchant business name and non-merchant business name of a store sample; The BERT model was fine-tuned and trained based on the third training dataset to obtain the store name classification model.

8. The method according to claim 7, characterized in that, The merchant's business name includes at least one of the following components: brand information, brand information and industry information, brand information and business district name, or brand information, industry information, and business district name.

9. The method according to claim 5, characterized in that, The method further includes: If at least one of the duplicate stores is identified, the address precision of the target store's address and the address precision of each of the duplicate stores' addresses are determined respectively. The store address with the highest address precision among the store address of the target store and the store addresses of at least one duplicate store is determined as the store address corresponding to the target store name.

10. A store information processing device, characterized in that, include: The information acquisition module is used to acquire the store information of the target store, including the store name and store address; The information extraction module is used to determine the brand information and industry category of the target store based on the store name; The potential store identification module is used to find a set of potential duplicate stores from the existing store information database whose brand information and industry category are consistent with the target store; The threshold determination module is used to determine the target distance threshold corresponding to the industry category to which the target store belongs; The duplicate store identification module is used to identify stores in the potential duplicate store set whose store address is less than the target distance threshold as duplicate stores of the target store.

11. An electronic device, characterized in that, The electronic device includes: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, it implements the store information processing method as described in any one of claims 1-9.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement the store information processing method as described in any one of claims 1-9.

13. A computer program product, characterized in that, When the instructions in the computer program product are executed by the processor of the electronic device, the electronic device performs the store information processing method as described in any one of claims 1-9.

Images