POI information mining method, device, equipment and storage medium

Abstract

The present disclosure provides a POI information mining method, device, equipment and storage medium, which relates to the field of artificial intelligence technology, specifically to technical fields such as image processing, text processing, and deep learning, and can be applied to scenarios such as POI information retrieval services and store sign verification. The specific implementation scheme includes: obtaining POI information, the POI information includes the sign image of the target store and text related to the target store; determining the first score of the first preset number of first brands with the highest similarity between the sign image and the sign image of the target store; determining the second score of the second preset number of second brands with the highest similarity between the text and the text related to the target store; and determining the target brand corresponding to the POI information from the first brand and the second brand based on the first score and the second score. The present disclosure can intelligently mine the POI information of a brand, achieve low cost and high timeliness, and can greatly improve the mining recall rate and accuracy of POI information.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, specifically to technical fields such as image processing, text processing, and deep learning, and can be applied to scenarios such as providing users with POI information retrieval services and store sign verification, and in particular relates to a POI information mining method, device, equipment, and storage medium. Background Art

[0002] Point of interest (POI) information (or POI data) generally refers to information related to a specific geographic location within geographic information. For example, a location could be a house, a shop or store, a mailbox, a bus stop, and so on. POI information can include fields such as the location's name, address, coordinates, phone number, business hours, and brand.

[0003] Brand is a key attribute of many stores in POI information. When users search for stores of a specific brand, search services (such as map services and food delivery services) need to address these chain search needs, identify stores in POI information that contain the target brand or whose POI information is related to the target brand, and then present these stores to users for selection.

[0004] Currently, there are two main methods for mining POI information. In one approach, search service providers can mine POI information for stores related to the target brand through merchant collaboration or official website crawlers. In the other approach, search service providers can cluster and mine POI information for stores related to the target brand using a storefront graph clustering algorithm (such as the Kmeans algorithm). Summary of the Invention

[0005] The present disclosure provides a POI information mining method, apparatus, device and storage medium, which can intelligently mine brand POI information with low cost and high timeliness, and can greatly improve the mining recall rate and accuracy of POI information.

[0006] According to a first aspect of the present disclosure, a POI information mining method is provided, the method comprising:

[0007] Obtain POI information of a target store, where the POI information includes a signboard image of the target store and text related to the target store; determine a first preset number of first brands having the highest similarity between their signboard images and the signboard images of the target store based on a preset signboard image library, and determine a first score for each first brand based on the similarity between the signboard images of the first brands and the signboard images of the target store, wherein the signboard image library includes signboard images of at least two brands; determine a second preset number of second brands having the highest similarity between their texts and text related to the target store based on a preset text library, and determine a second score for each second brand based on the similarity between the text related to the second brands and the text related to the target store, wherein the text library includes text related to at least two brands; determine a fusion score for each of the first brand and the second brand based on the first score and the second score; determine a target brand corresponding to the POI information of the target store from the first brand and the second brand based on the fusion score of each of the first brand and the second brand.

[0008] According to a second aspect of the present disclosure, a POI information mining device is provided, the device comprising:

[0009] An acquisition unit is used to acquire POI information of a target store, where the POI information includes a signboard image of the target store and text related to the target store; an image processing unit is used to determine, based on a preset signboard image library, a first preset number of first brands whose signboard images have the highest similarity with the signboard image of the target store, and determine a first score for each first brand based on the similarity between the signboard image of the first brand and the signboard image of the target store, wherein the signboard image library includes signboard images of at least two brands; a text processing unit is used to determine, based on a preset text library, a second preset number of second brands whose texts have the highest similarity with text related to the target store, and determine a second score for each second brand based on the similarity between text related to the second brand and text related to the target store, wherein the text library includes text related to at least two brands; a fusion unit is used to determine a fusion score of each of the first brand and the second brand based on the first score and the second score; an identification unit is used to determine, from the first brand and the second brand, a target brand corresponding to the POI information of the target store based on the fusion score of each of the first brand and the second brand.

[0010] According to a third aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method described in the first aspect.

[0011] According to a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause a computer to execute the method according to the first aspect.

[0012] According to a fifth aspect of the present disclosure, a computer program product is provided, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements the method according to the first aspect.

[0013] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The accompanying drawings are provided to facilitate a better understanding of the present invention and do not constitute a limitation of the present disclosure.

[0015] Figure 1 A flowchart of a POI information mining method according to an embodiment of the present disclosure;

[0016] Figure 2 Provided for the embodiments of the present disclosure Figure 1 A schematic diagram of an implementation flow of S102;

[0017] Figure 3 Provided for the embodiments of the present disclosure Figure 1 Another implementation flow diagram of S102;

[0018] Figure 4 Provided for the embodiments of the present disclosure Figure 1 A schematic diagram of an implementation flow of S103;

[0019] Figure 5 Provided for the embodiments of the present disclosure Figure 1 Another implementation flow diagram of S103;

[0020] Figure 6 A schematic diagram of the composition of a POI information mining device provided in an embodiment of the present disclosure;

[0021] Figure 7 A schematic block diagram of an example electronic device 700 is shown, which may be used to implement embodiments of the present disclosure. DETAILED DESCRIPTION

[0022] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0023] It should be understood that in the various embodiments of the present disclosure, the character " / " generally indicates an "or" relationship between the preceding and following related objects. The terms "first," "second," etc. are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features.

[0024] Point of interest (POI) information (or POI data) generally refers to information related to a specific geographic location within geographic information. For example, a location could be a house, a shop or store, a mailbox, a bus stop, and so on. POI information can include fields such as the location's name, address, coordinates, phone number, business hours, and brand.

[0025] Brand is a key attribute of many stores in POI information. When users search for stores of a specific brand, search services (such as map services and food delivery services) need to address these chain search needs, identify stores in POI information that contain the target brand or whose POI information is related to the target brand, and then present these stores to users for selection.

[0026] Currently, there are two main methods for mining POI information. In one approach, search service providers can pre-acquire POI information for stores related to the target brand through merchant collaboration or official website crawlers. When providing search services, these pre-acquired POI information can be used to display information related to the target brand to users. However, this approach, which relies heavily on objective factors and lacks initiative, requires extensive collaboration, is costly, and lacks effectiveness.

[0027] Alternatively, search service providers can use a storefront clustering algorithm (such as the Kmeans algorithm) to cluster and mine POI information for stores related to the target brand. However, this method, which uses a storefront clustering algorithm, has low accuracy and recall rates for mining brand POI information.

[0028] For example, the store signs of a target brand may include Chinese text signs, pattern signs, and English text signs. However, when mining POI information of stores related to the target brand, only POI information of stores corresponding to one or two of the Chinese text signs, pattern signs, and English text signs may be mined, resulting in a low mining recall rate.

[0029] For example, when mining POI information of stores related to a target brand, POI information of stores corresponding to similar signs of other brands may be mistakenly mined, resulting in low mining accuracy.

[0030] The present disclosure provides a POI information mining method that can intelligently mine brand POI information, achieve low cost and high timeliness, and can greatly improve the mining recall rate and accuracy of POI information.

[0031] The execution subject of the method can be a computer or a server, or can also be other equipment with data processing capabilities. There is no limitation on the execution subject of the method.

[0032] For example, the server can be a backend server that provides users with POI information retrieval services (or retrieval services). For example, a map application can provide users with a retrieval service for searching for stores of a certain target brand, and the server can be the backend server of the map application. For another example, a food delivery application can also provide users with a retrieval service for searching for stores of a certain target brand, and the server can be the backend server of the food delivery application.

[0033] In some embodiments, the server may be a single server, or a server cluster consisting of multiple servers. In some implementations, the server cluster may also be a distributed cluster. The present disclosure does not limit the specific implementation of the server.

[0034] The POI information mining method is exemplified below.

[0035] Figure 1 Schematic diagram of the process of POI information mining method provided by the embodiment of the present disclosure. Figure 1 As shown, the method may include:

[0036] S101. Obtain POI information of a target store, where the POI information includes a signboard image of the target store and text related to the target store.

[0037] For example, the target store may be a commercial store, such as a restaurant, convenience store, etc. The target store's sign image may be a picture or video taken of the target store. The target store-related text may include user comments about the target store.

[0038] S102. Determine, based on a preset sign image library, a first preset number of first brands whose sign images have the highest similarity with the sign images of the target store, and determine a first score for each first brand based on the similarity between the sign image of the first brand and the sign image of the target store, wherein the sign image library includes sign images of at least two brands.

[0039] Exemplarily, the preset sign image library may include sign images of at least two brands. In some implementations, the sign image library may be constructed by clustering storefront images + manual annotation. For example, sign images of multiple brands may be obtained (sign images refer to images obtained by photographing a brand's store), and each sign image may be labeled to indicate which brand it belongs to. A clustering algorithm (such as the Kmeans algorithm, the BDSCAN clustering algorithm, etc.) may be used to cluster these sign images, with each cluster corresponding to a brand. These clusters obtained by clustering may be used to construct a sign image library.

[0040] For the target store's sign image in S101, the similarity between the target store's sign image and each cluster in the sign image library can be calculated, and this similarity can be used as the similarity between the sign image of the brand corresponding to each cluster and the target store's sign image. Based on the similarity between the sign image of the brand corresponding to each cluster and the target store's sign image, a first preset number of brands with the highest similarity can be selected as first brands. In other words, the number of first brands is the first preset number.

[0041] Exemplarily, the first preset number may be 10, 20, etc., and there is no limitation on the size of the first preset number.

[0042] Optionally, when calculating the similarity between the sign image of the target store and each cluster in the sign image library, the image features of the cluster center of the sign image in each cluster (or the average of the image features of all sign images) and the image features of the sign image of the target store can be calculated, and then the image features of the cluster center of the sign image in each cluster and the image features of the sign image of the target store are encoded using the resnet+arcface method. Feature index calculation is performed based on the encoded image features to obtain the similarity between the sign image of the target store and each cluster in the sign image library.

[0043] Exemplarily, the step of determining the first score of each first brand based on the similarity between the signboard image of the first brand and the signboard image of the target store in S102 may include: taking the similarity between the signboard image of the first brand and the signboard image of the target store as the first score of the first brand.

[0044] S103. Determine, based on a preset text library, a second preset number of second brands having the highest similarity between their texts and texts related to the target store, and determine a second score for each second brand based on the similarity between the texts related to the second brands and the texts related to the target store, wherein the text library includes texts related to at least two brands.

[0045] Exemplarily, the preset text library may include texts related to at least two brands. In some implementations, the text library may also be constructed through clustering and manual annotation. For example, multiple brand-related texts may be obtained, and each (or each group of) texts may be labeled with the brand to which they belong. A clustering algorithm may be used to cluster these texts, with each cluster corresponding to a brand. These clusters may be used to construct a text library.

[0046] For the target store-related text in S101, the similarity between the target store-related text and each cluster in the text library can be calculated, and the similarity between the brand text corresponding to each cluster and the target store-related text can be calculated. Based on the similarity between the brand text corresponding to each cluster and the target store-related text, a second preset number of brands with the highest similarity can be selected as second brands. In other words, the number of second brands is the second preset number.

[0047] For example, the second preset number may be 10, 20, etc. The second preset number may be the same as or different from the first preset number. There is no limitation on the size of the second preset number.

[0048] Optionally, when calculating the similarity between the text related to the target store and each cluster in the text library, the text features of the cluster center of the text in each cluster (or the average of the text features of all texts) and the text features of the text related to the target store can be calculated, and then the text features of the cluster center of the text in each cluster and the text features of the text related to the target store are encoded using the Bert+cycleloss method. Feature index calculation is performed based on the encoded text features to obtain the similarity between the text related to the target store and each cluster in the text library.

[0049] For example, the step of determining the second score of each second brand based on the similarity between the text related to the second brand and the text related to the target store in S103 may include: using the similarity between the text related to the second brand and the text related to the target store as the second score of the second brand.

[0050] S104: Determine a fusion score for each of the first brand and the second brand based on the first score and the second score.

[0051] Illustratively, determining the fusion score of each of the first brand and the second brand based on the first score and the second score may include: for each of the first brand and the second brand, summing the first score and the second score of each brand as the fusion score of the brand.

[0052] For example, assuming that the first preset number and the second preset number are both 5, the first brand includes: Brand 1, Brand 2, Brand 3, Brand 4, Brand 5, the first score corresponding to Brand 1 is 0.8, the first score corresponding to Brand 2 is 0.7, the first score corresponding to Brand 3 is 0.7, the first score corresponding to Brand 4 is 0.6, and the first score corresponding to Brand 5 is 0.5; the second brands include: Brand 1, Brand 3, Brand 4, Brand 6, Brand 7, the second score corresponding to Brand 1 is 0.9, the second score corresponding to Brand 3 is 0.8, the second score corresponding to Brand 4 is 0.7, the second score corresponding to Brand 6 is 0.6, and the second score corresponding to Brand 7 is 0.6. Then, for brand 1, the first score corresponding to brand 1 is 0.8 and the second score is 0.9, and the corresponding fusion score of brand 1 can be 0.8+0.9=1.7; for brand 2, the first score corresponding to brand 2 is 0.7 and the second score is 0 (that is, there is no second score), and the corresponding fusion score of brand 2 can be 0.7+0=0.7; similarly, the fusion score of each other brand can be obtained.

[0053] S105 : Determine, from the first brand and the second brand, a target brand corresponding to the POI information of the target store based on the fusion score of each of the first brand and the second brand.

[0054] In some implementations, based on the fusion score of each of the first brand and the second brand, one or more brands with the highest fusion score may be selected from the first brand and the second brand as the target brand corresponding to the POI information of the target store.

[0055] For example, taking the example given in S104, all brands appearing in the first brand and the second brand include brand 1, brand 2, brand 3, brand 4, brand 5, brand 6, and brand 7. Brand 1 with the highest fusion score can be selected from brand 1, brand 2, brand 3, brand 4, brand 5, brand 6, and brand 7 as the target brand corresponding to the POI information of the target store.

[0056] In some other implementations, the brand with a fusion score greater than a preset score from the first brand and the second brand may be selected as the target brand corresponding to the POI information of the target store. For example, the preset score may be 0.9, 0.8, etc., without limitation.

[0057] In an embodiment of the present disclosure, by obtaining POI information of a target store, the POI information includes a signboard image of the target store and text related to the target store, and according to a preset signboard image library, determining a first preset number of first brands with the highest similarity between the signboard image and the signboard image of the target store, and determining a first score for each first brand based on the similarity between the signboard image of the first brand and the signboard image of the target store, and according to a preset text library, determining a second preset number of second brands with the highest similarity between the text and the text related to the target store, and determining a second score for each second brand based on the similarity between the text related to the second brand and the text related to the target store, determining a fusion score of each of the first brand and the second brand based on the first score and the second score, and determining the target brand corresponding to the POI information of the target store from the first brand and the second brand based on the fusion score of each of the first brand and the second brand, thereby realizing intelligent mining of the brand's POI information with low cost and high timeliness.

[0058] Among them, the first score related to the image and the second score related to the text are fused, and the target brand corresponding to the POI information of the target store is identified based on the fused score, which can greatly improve the mining recall rate and accuracy of the POI information.

[0059] For example, the POI information mining method provided by the embodiments of the present disclosure can be applied to scenarios where the brand of a new store is verified. For example, if a new store has opened but the map data does not yet contain brand information, the POI information mining method provided by the embodiments of the present disclosure can be used to determine the true brand of the newly opened store.

[0060] In some embodiments, the above S104 may also include: taking the weight of the first score as the first weight and the weight of the second score as the second weight, summing the first score and the second score corresponding to each brand in the first brand and the second brand to obtain a fusion score for each brand in the first brand and the second brand.

[0061] For example, assuming that the first weight is α, the second weight is β, the first score is score1, and the second score is score2, the fusion score can be obtained by the following formula.

[0062] score 融合 =α*score1+β*score2

[0063] Among them, score 融合 represents the fusion score.

[0064] In this embodiment, by setting a first weight and a second weight for the first score and the second score respectively, it is possible to flexibly adjust the influence of the signboard image and text on the mining results during the POI information mining process.

[0065] Optionally, the sum of the first weight and the second weight may be 1 or may not be 1, which is not limited here.

[0066] In some embodiments, the text library includes at least two, and each text library includes a corresponding type of text.

[0067] The above-mentioned step of determining the second preset number of second brands with the highest similarity between the text and the text related to the target store based on the preset text library may include: determining the second preset number of second brands with the highest similarity between the text and the text related to the target store based on each text library respectively, to obtain at least two types of second preset number of second brands.

[0068] The above-mentioned step of determining the second score of each second brand based on the similarity between the text related to the second brand and the text related to the target store may include: for each type of second brand, determining the second score of each second brand based on the similarity between the text related to the second brand and the text related to the target store.

[0069] Exemplarily, the text types may include at least two of the following: comment type, recommendation type, and network type. For example, a comment type text may refer to a user's evaluation of a target store on a platform (such as a map, takeout, etc.). A recommendation type text may be a dish or product recommended by a user or a merchant of a target store on a platform (such as a map, takeout, etc.) for the target store. A network type text may be WIFI information such as the network IP address, WIFI name, or WIFI password of the target store. This disclosure does not limit the type of text.

[0070] In this embodiment, for each type of text, a text library corresponding to that type of text can be constructed. For each text library, a second preset number of second brands whose text has the highest similarity to text related to the target store can be determined, and the second preset number of second brands corresponding to the text type of the text library can be obtained. Thus, a second preset number of second brands of at least two types can be obtained.

[0071] For example, if the text library includes a review text library, a recommendation text library, and a network text library, then a second preset number of second brands with the highest similarity between the text in each text library and the text related to the target store can be determined, thereby obtaining a second preset number of second brands of the review type, a second preset number of second brands of the recommendation type, and a second preset number of second brands of the network type. In other words, a second preset number of second brands can be obtained for each text type.

[0072] For each type of second brand, a second score for each second brand may be determined based on the similarity between the text related to the second brand and the text related to the target store.

[0073] In this embodiment, when determining the fusion score for each of the first and second brands based on the first and second scores, the first and second scores of each brand can also be summed to form the fusion score for that brand. The difference from the previous embodiment is that in this embodiment, the second score for each brand includes second scores corresponding to multiple categories.

[0074] For example, assume that both the first and second preset numbers are 2, the first and second weights are 0.5, and the first brand includes Brand 1 and Brand 2, with Brand 1 having a first score of 0.8 and Brand 2 having a first score of 0.7. The second brand includes a comment-based second brand and a network-based second brand, where the comment-based second brand includes Brand 1 and Brand 3, with Brand 1 having a second score of 0.9 and Brand 3 having a second score of 0.7. The network-based second brand includes Brand 2 and Brand 3, with Brand 2 having a second score of 0.8 and Brand 3 having a second score of 0.7. Therefore, for Brand 1, the first score for Brand 1 is 0.8, the second score for the comment-based second brand is 0.9, and the second score for the network-based second brand is 0. The fusion score for Brand 1 can be 0.8*0.5+0.9*0.5+0*0.5=0.85. Similarly, the fusion score for each of the other brands can be obtained.

[0075] In this embodiment, the text library includes at least two, each text library corresponds to a type of text, and based on each text library, a second preset number of second brands with the highest similarity between the text and the text related to the target store are determined to obtain at least two types of second preset number of second brands. For each type of second brand, a second score of each second brand is determined based on the similarity between the text related to the second brand and the text related to the target store. This can realize the mining of POI information based on multiple text types, and can further improve the recall rate and accuracy.

[0076] In some implementations, when summing the first score and the second score corresponding to each of the first brand and the second brand, the second scores corresponding to different types of second brands are weighted by the second weight. That is, the second scores corresponding to different types of second brands are weighted the same.

[0077] In some other implementations, when summing the first score and the second score corresponding to each of the first brand and the second brand, the sum of the weights of the second scores corresponding to all types of second brands may be the second weight, and the weights of the second scores corresponding to different types of second brands may be different.

[0078] In this implementation, different types of second brands have different weights corresponding to the second scores, which can flexibly adjust the impact of different types of text on the mining results during the POI information mining process.

[0079] Figure 2 Provided for the embodiments of the present disclosure Figure 1 A schematic diagram of an implementation flow of S102 is shown in FIG. Figure 2 As shown, in some embodiments, the step of determining the first score of each first brand based on the similarity between the signboard image of the first brand and the signboard image of the target store in S102 may include:

[0080] S201: Determine a similarity score for each first brand based on the similarity between the first brand's signboard image and the target store's signboard image.

[0081] For example, assuming that the first brand includes Brand 1 and Brand 2, the similarity between the signboard image of Brand 1 and the signboard image of the target store is 0.8, and the similarity between the signboard image of Brand 2 and the signboard image of the target store is 0.7, then the similarity score of Brand 1 can be 0.8, and the similarity score of Brand 2 can be 0.7.

[0082] S202: Determine a consistency score for each first brand based on the number of first brands belonging to the same brand among the first preset number of first brands and the first preset number.

[0083] For example, in some scenarios, some stores may have different signs, but these stores may belong to the same brand. For example, a brand's store signs may include: pattern signs, Chinese signs, English signs, etc. Among the first preset number of first brands determined in the embodiments of this disclosure, it is possible that some of the first brands belong to the same brand.

[0084] For example, assuming the first preset number is 5, the first brand includes Brand 1, Brand 2, Brand 3, Brand 4, and Brand 5, among which Brand 1 and Brand 3 are both Brand A, but Brand 1 uses the Chinese signboard of Brand A, and Brand 3 uses the English signboard of Brand A.

[0085] S202 may include: for each first brand, taking a ratio of the number of first brands that are the same brand as the first brand in a first preset number of first brands to the first preset number as a consistency score of the first brand.

[0086] For example, if the first brands mentioned above include Brand 1, Brand 2, Brand 3, Brand 4, and Brand 5, where Brand 1 and Brand 3 are both Brand A, and Brand 1 and Brand 3 are the same brand, and the number of first brands that share the same brand with Brand 1 is 2, then Brand 1's consistency score can be 0.4; Brand 3's consistency score is also 0.4. The number of first brands that share the same brand with Brand 2 is 1, so Brand 2's consistency score can be 0.2. The number of first brands that share the same brand with Brand 4 is also 1, and Brand 4's consistency score is also 0.2.

[0087] S203: Determine a first score for the first brand based on the similarity score and consistency score of the first brand.

[0088] In some implementations, S203 may include: summing the similarity score and the consistency score of the first brand to obtain a first score of the first brand.

[0089] In some implementations, S203 may include: taking the weight of the first brand's similarity score as a third weight and the weight of the first brand's consistency score as a fourth weight, and summing the similarity score and consistency score of the first brand to obtain a first score for the first brand. The sum of the third weight and the fourth weight may or may not be 1.

[0090] In this embodiment, by introducing the consistency score, the score proportion of brands with good consistency can be increased, and the mining recall rate and accuracy of POI information can be further improved.

[0091] Figure 3 Provided for the embodiments of the present disclosure Figure 1 Another implementation flow chart of S102 is shown in FIG. Figure 3 As shown, in some embodiments, the step of determining, in S102, a first preset number of first brands having the highest similarity between their signboard images and the signboard images of the target store based on a preset signboard image library may include:

[0092] S301. Input the target store's signboard image into a preset image recognition model, and determine the candidate first brands whose signboard images are similar to the target store's signboard image and the confidence level of each candidate first brand through the image recognition model. The image recognition model is trained using a signboard image library.

[0093] For example, a neural network can be trained using a library of signboard images to generate an image recognition model. The type of neural network is not limited. The image recognition model can be configured to predict the brand of a target store's signboard image. After inputting the target store's signboard image into the image recognition model, the model can output candidate first brands whose signboard images are similar to the target store's signboard image, as well as a confidence score for each candidate first brand.

[0094] S302: Using the confidence of the candidate first brands as the similarity, a first preset number of first brands with the highest similarity are selected from the candidate first brands.

[0095] For example, assuming that the first preset number is 2, the candidate first brands include Brand 1, Brand 2, Brand 3, Brand 4, and Brand 5, the confidence of Brand 1 is 0.9, the confidence of Brand 2 is 0.8, the confidence of Brand 3 is 0.7, the confidence of Brand 4 is 0.6, and the confidence of Brand 5 is 0.5, then Brand 1 and Brand 2 with the highest confidence can be selected as the first brands.

[0096] In this embodiment, by introducing an image recognition model and selecting a first preset number of first brands in combination with the confidence level output by the image recognition model, the accuracy of the similarity calculation result can be improved, thereby improving the mining recall rate of POI information.

[0097] Figure 4 Provided for the embodiments of the present disclosure Figure 1 A schematic diagram of an implementation flow of S103 in FIG. Figure 4 As shown, in some embodiments, the step of determining the second score of each second brand based on the similarity between the text related to the second brand and the text related to the target store in S103 may include:

[0098] S401: Determine a similarity score for each second brand based on the similarity between the text related to the second brand and the text related to the target store.

[0099] S401 can refer to S201 and will not be described in detail.

[0100] S402: Determine a consistency score for each second brand according to the number of second brands belonging to the same brand among the second preset number of second brands and the second preset number.

[0101] S402 can refer to S202 and will not be described in detail.

[0102] S403: Determine a second score for the second brand based on the similarity score and consistency score of the second brand.

[0103] In some implementations, S403 may include: summing the similarity score and the consistency score of the second brand to obtain a second score of the second brand.

[0104] In some implementations, S403 may include: using the weight of the second brand's similarity score as a fifth weight and the weight of the second brand's consistency score as a sixth weight, and summing the similarity score and consistency score of the second brand to obtain a second score for the second brand. The sum of the fifth weight and the sixth weight may or may not be 1.

[0105] In this embodiment, by introducing the consistency score, the score proportion of brands with good consistency can also be increased, further improving the mining recall rate and accuracy of POI information.

[0106] Figure 5 Provided for the embodiments of the present disclosure Figure 1 Another implementation flow diagram of S103 is shown in FIG. Figure 3 As shown, in some embodiments, the step of determining, in S103, a second preset number of second brands having the highest similarity between text and text related to the target store based on a preset text library may include:

[0107] S501. Input text related to the target store into a preset text recognition model, and determine candidate second brands whose text is similar to the text related to the target store and the confidence level of each candidate second brand through the text recognition model. The text recognition model is trained using a text library.

[0108] For example, a neural network can be trained using a text library to generate a text recognition model. The type of neural network is not limited herein. The text recognition model can predict which signboard the target store-related text belongs to. After inputting the target store-related text into the text recognition model, the text recognition model can output candidate second brands whose text is similar to the target store-related text, as well as a confidence score for each candidate second brand.

[0109] Optionally, in the embodiments of the present disclosure, for each type of text (ie, each text library) mentioned in the aforementioned embodiments, a text recognition model may be trained to recognize different types of text.

[0110] S502: Using the confidence of the candidate second brands as the similarity, a second preset number of second brands with the highest similarity are selected from the candidate second brands.

[0111] S502 can refer to S302 and will not be described in detail.

[0112] In this embodiment, by introducing a text recognition model and selecting a second preset number of second brands in combination with the confidence level output by the text recognition model, the accuracy of the similarity calculation result can also be improved, thereby improving the mining recall rate of POI information.

[0113] In an exemplary embodiment, the present disclosure further provides a POI information mining device, which can be used to implement the POI information mining method as described in the above embodiments. Figure 6 This is a schematic diagram of the composition of the POI information mining device provided in the embodiment of the present disclosure. Figure 6 As shown, the device may include: an acquisition unit 601, an image processing unit 602, a text processing unit 603, a fusion unit 604, and a recognition unit 605.

[0114] The acquisition unit 601 is configured to acquire POI information of a target store, where the POI information includes a signboard image of the target store and text related to the target store.

[0115] The image processing unit 602 is used to determine, based on a preset sign image library, a first preset number of first brands whose sign images have the highest similarity with the sign images of the target store, and determine a first score for each first brand based on the similarity between the sign images of the first brands and the sign images of the target store, wherein the sign image library includes sign images of at least two brands.

[0116] The text processing unit 603 is used to determine a second preset number of second brands whose texts have the highest similarity with texts related to the target store based on a preset text library, and determine a second score for each second brand based on the similarity between the texts related to the second brands and the texts related to the target store, where the text library includes texts related to at least two brands.

[0117] The fusion unit 604 is configured to determine a fusion score for each of the first brand and the second brand based on the first score and the second score.

[0118] The identification unit 605 is configured to determine a target brand corresponding to the POI information of the target store from the first brand and the second brand according to the fusion score of each of the first brand and the second brand.

[0119] Optionally, the fusion unit 604 is specifically configured to sum the first score and the second score corresponding to each brand in the first brand and the second brand, taking the weight of the first score as the first weight and the weight of the second score as the second weight, to obtain a fusion score for each brand in the first brand and the second brand.

[0120] Optionally, the text library includes at least two, and each text library includes text of one type.

[0121] The text processing unit 603 is specifically configured to determine, based on each text library, a second preset number of second brands having the highest similarity between the text and the text related to the target store, and obtain a second preset number of second brands of at least two types.

[0122] The text processing unit 603 is further configured to determine, for each type of second brand, a second score for each second brand according to the similarity between the text related to the second brand and the text related to the target store.

[0123] Optionally, when summing the first score and the second score corresponding to each of the first brand and the second brand, the second scores corresponding to different types of second brands have different weights.

[0124] Optionally, the type includes at least two of the following: comment type, recommendation type, and network type.

[0125] Optionally, the image processing unit 602 is specifically used to determine the similarity score of each first brand based on the similarity between the signboard image of the first brand and the signboard image of the target store; determine the consistency score of each first brand based on the number of first brands belonging to the same brand in the first preset number of first brands and the first preset number; determine the first score of the first brand based on the similarity score and consistency score of the first brand.

[0126] Optionally, the image processing unit 602 is specifically used to input the signboard image of the target store into a preset image recognition model, and determine the candidate first brands whose signboard images are similar to the signboard image of the target store, as well as the confidence of each candidate first brand through the image recognition model. The image recognition model is trained using a signboard image library; the confidence of the candidate first brand is used as the similarity, and a first preset number of first brands with the highest similarity are selected from the candidate first brands.

[0127] Optionally, the text processing unit 603 is specifically used to determine the similarity score of each second brand based on the similarity between the text related to the second brand and the text related to the target store; determine the consistency score of each second brand based on the number of second brands belonging to the same brand in the second preset number of second brands and the second preset number; determine the second score of the second brand based on the similarity score and consistency score of the second brand.

[0128] Optionally, the text processing unit 603 is specifically used to input the text related to the target store into a preset text recognition model, and determine the candidate second brands whose text is similar to the text related to the target store, as well as the confidence of each candidate second brand through the text recognition model. The text recognition model is obtained by training using a text library; the confidence of the candidate second brand is used as the similarity, and a second preset number of second brands with the highest similarity are selected from the candidate second brands.

[0129] In the technical solutions disclosed herein, the acquisition, storage, and application of user personal information involved comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0130] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0131] In an exemplary embodiment, an electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method described in the above embodiments. The electronic device may be the computer or server described above.

[0132] In an exemplary embodiment, the readable storage medium may be a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause a computer to execute the method according to the above embodiments.

[0133] In an exemplary embodiment, a computer program product includes a computer program, which implements the method according to the above embodiments when executed by a processor.

[0134] Figure 7A schematic block diagram of an example electronic device 700 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0135] like Figure 7 As shown, electronic device 700 includes a computing unit 701, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 702 or a computer program loaded from a storage unit 708 into a random access memory (RAM) 703. Various programs and data required for the operation of device 700 can also be stored in RAM 703. Computing unit 701, ROM 702, and RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to bus 704.

[0136] Multiple components in the electronic device 700 are connected to the I / O interface 705, including an input unit 706, such as a keyboard, a mouse, etc.; an output unit 707, such as various types of displays, speakers, etc.; a storage unit 708, such as a magnetic disk, an optical disk, etc.; and a communication unit 709, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 709 allows the electronic device 700 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0137] The computing unit 701 can be any general-purpose and / or specialized processing component with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 701 performs the various methods and processes described above, such as the POI information mining method. For example, in some embodiments, the POI information mining method can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 708. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 700 via the ROM 702 and / or the communication unit 709. When the computer program is loaded into the RAM 703 and executed by the computing unit 701, one or more steps of the POI information mining method described above can be performed. Alternatively, in other embodiments, the computing unit 701 can be configured to perform the POI information mining method by any other suitable means (e.g., via firmware).

[0138] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system comprising at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0139] The program code for implementing the method of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0140] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0141] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0142] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0143] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact through a communication network. The client-server relationship arises through computer programs running on the respective computers and having a client-server relationship with each other. The server may be a cloud server, a server in a distributed system, or a server integrated with a blockchain.

[0144] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not limited herein.

[0145] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A POI information mining method, comprising: Obtaining POI information of a target store, wherein the POI information includes a signboard image of the target store and text related to the target store; Determining, based on a preset sign image library, a first preset number of first brands having the highest similarity between their sign images and the sign images of the target store, and determining a first score for each of the first brands based on the similarity between the sign images of the first brands and the sign images of the target store, wherein the sign image library includes sign images of at least two brands; determining, based on a preset text library, a second preset number of second brands having the highest similarity between their text and text related to the target store, and determining a second score for each of the second brands based on the similarity between the text related to the second brands and the text related to the target store, wherein the text library includes text related to at least two brands; determining a fusion score for each of the first brand and the second brand based on the first score and the second score; determining, from the first brand and the second brand, a target brand corresponding to the POI information of the target store; The step of determining the first score of each first brand according to the similarity between the signboard image of the first brand and the signboard image of the target store includes: determining a similarity score for each of the first brands based on a similarity between the signboard image of the first brand and the signboard image of the target store; determining a consistency score for each of the first brands according to the number of first brands belonging to the same brand among the first preset number of first brands and the first preset number; determining a first score for the first brand based on the similarity score and the consistency score of the first brand; The step of determining, based on a preset signboard image library, a first preset number of first brands having the highest similarity between their signboard images and the signboard images of the target store includes: Inputting the target store's signboard image into a preset image recognition model, and determining, using the image recognition model, candidate first brands whose signboard images are similar to the target store's signboard image, and a confidence score for each candidate first brand, wherein the image recognition model is trained using the signboard image library; The confidence levels of the candidate first brands are used as similarities, and a first preset number of first brands with the highest similarities are selected from the candidate first brands.

2. The method according to claim 1, wherein determining a fusion score of each of the first brand and the second brand based on the first score and the second score comprises: The weight of the first score is taken as a first weight, and the weight of the second score is taken as a second weight. The first score and the second score corresponding to each of the first brand and the second brand are summed to obtain a fusion score of each of the first brand and the second brand.

3. The method according to claim 2, wherein the text libraries include at least two, each of which includes a type of text; The step of determining, based on a preset text library, a second preset number of second brands having the highest similarity between texts and texts related to the target store, includes: Determining, based on each of the text libraries, a second preset number of second brands having the highest similarity between their texts and the texts related to the target store, to obtain a second preset number of second brands of at least two types; Determining a second score for each second brand based on the similarity between the text related to the second brand and the text related to the target store includes: For each type of the second brand, a second score for each second brand is determined according to the similarity between the text related to the second brand and the text related to the target store.

4. The method according to claim 3, wherein when the first score and the second score corresponding to each of the first brand and the second brand are summed, the second scores corresponding to different types of the second brands have different weights.

5. The method according to claim 3 or 4, wherein the types include at least two of the following: comment type, recommendation type, and network type.

6. The method according to any one of claims 1 to 5, wherein determining the second score for each second brand based on the similarity between the text related to the second brand and the text related to the target store comprises: determining a similarity score for each of the second brands based on a similarity between the text related to the second brand and the text related to the target store; determining a consistency score for each of the second brands according to the number of second brands belonging to the same brand among the second preset number of second brands and the second preset number; A second score of the second brand is determined according to the similarity score and the consistency score of the second brand.

7. The method according to claim 6, wherein determining, based on a preset text library, a second preset number of second brands having the highest similarity between their texts and the texts related to the target store comprises: inputting the target store-related text into a preset text recognition model, and determining, using the text recognition model, candidate second brands whose text is similar to the target store-related text, and a confidence score for each candidate second brand, wherein the text recognition model is trained using the text library; The confidence levels of the candidate second brands are used as similarities, and a second preset number of second brands with the highest similarities are selected from the candidate second brands.

8. A POI information mining device, comprising: an acquiring unit, configured to acquire POI information of a target store, wherein the POI information includes a signboard image of the target store and text related to the target store; an image processing unit, configured to determine, based on a preset sign image library, a first preset number of first brands having the highest similarity between their sign images and the sign image of the target store, and determine a first score for each of the first brands based on the similarity between the sign image of the first brand and the sign image of the target store, wherein the sign image library includes sign images of at least two brands; a text processing unit, configured to determine, based on a preset text library, a second preset number of second brands having texts with the highest similarity to texts related to the target store, and determine a second score for each of the second brands based on the similarity between the texts related to the second brands and the texts related to the target store, wherein the text library includes texts related to at least two brands; a fusion unit, configured to determine a fusion score of each of the first brand and the second brand based on the first score and the second score; an identification unit, configured to determine, from the first brand and the second brand, a target brand corresponding to the POI information of the target store according to a fusion score of each of the first brand and the second brand; The image processing unit is specifically configured to determine a similarity score for each first brand based on the similarity between the signboard image of the first brand and the signboard image of the target store; determine a consistency score for each first brand based on the number of first brands belonging to the same brand in the first preset number of first brands and the first preset number; and determine a first score for the first brand based on the similarity score and consistency score of the first brand; Among them, the image processing unit is specifically used to input the sign image of the target store into a preset image recognition model, and determine the candidate first brands whose sign images are similar to the sign image of the target store, as well as the confidence of each candidate first brand through the image recognition model. The image recognition model is trained using the sign image library; the confidence of the candidate first brand is used as the similarity, and the first preset number of first brands with the highest similarity are selected from the candidate first brands.

9. The device according to claim 8, wherein the fusion unit is specifically configured to sum the first score and the second score corresponding to each of the first brand and the second brand, taking the weight of the first score as a first weight and the weight of the second score as a second weight, to obtain a fusion score for each of the first brand and the second brand.

10. The apparatus according to claim 9, wherein the text libraries include at least two, each of which includes a type of text; The text processing unit is specifically configured to determine, based on each of the text libraries, a second preset number of second brands having the highest similarity between the text and the text related to the target store, and obtain a second preset number of second brands of at least two types; The text processing unit is further configured to determine, for each type of the second brand, a second score for each second brand according to a similarity between text related to the second brand and text related to the target store.

11. The apparatus according to claim 10, wherein when the first score and the second score corresponding to each of the first brand and the second brand are summed, the second scores corresponding to different types of the second brands have different weights.

12. In the device according to any one of claims 8 to 11, the text processing unit is specifically used to determine the similarity score of each second brand based on the similarity between the text related to the second brand and the text related to the target store; determine the consistency score of each second brand based on the number of second brands belonging to the same brand in the second preset number of second brands and the second preset number; and determine the second score of the second brand based on the similarity score and consistency score of the second brand.

13. The device according to claim 12, wherein the text processing unit is specifically used to input the text related to the target store into a preset text recognition model, determine the candidate second brands whose text is similar to the text related to the target store, and the confidence of each candidate second brand through the text recognition model, and the text recognition model is trained using the text library; use the confidence of the candidate second brand as the similarity, and select a second preset number of second brands with the highest similarity from the candidate second brands.

14. An electronic device comprising: at least one processor; and a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method according to any one of claims 1 to 7.

15. A non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause a computer to execute the method according to any one of claims 1 to 7.

16. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 7.

Images