Method, device, electronic device and storage medium for comparing names of points of interest

Through the multi-task learning model combining semantic representation layer and associated semantic features, the accuracy of map interest point name similarity calculation in the prior art is solved, and the accuracy of identifying interest point name similarity is improved, especially in different industry situations calculations below.

CN113157829BActive Publication Date: 2024-11-26BAIDU ONLINE NETWORK TECH (BEIJIBG) CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202010074812.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-01-22
Publication Date
2024-11-26
Estimated Expiration
2040-01-22

AI Technical Summary

Technical Problem

The existing deep semantic matching method based on bidirectional LSTM and Attention has positive and reverse problems when calculating the similarity of interest points names in maps and inaccuracy in different industries, which affects the accuracy of interest points name similarity.

Method used

Using a multi-task learning model, the similarity of interest point names is calculated by combining the semantic similarity characteristics of point-of-interest name and associated semantic features (such as industry attributes and word attributes), and the similarity of point-of-interest names is calculated using pre-trained models and training samples. Parameter sharing and feature extraction improve the accuracy of similarity recognition.

Benefits of technology

It improves the accuracy of the similarity calculation of point-of-interest name, and can more accurately distinguish different points of interest names, especially names with different industry attributes, which enhances the accuracy of the calculation results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113157829B_ABST
    Figure CN113157829B_ABST
Patent Text Reader

Abstract

The present application discloses a method, device, electronic device and storage medium for comparing points of interest names, and relates to the field of artificial intelligence technology. The specific implementation scheme is: obtaining the first point of interest name and the second point of interest name to be compared; calculating the similarity between the first point of interest name and the second point of interest name according to the semantic representation features corresponding to the first point of interest name and the second point of interest name respectively; wherein the semantic representation features include: a semantic similarity feature of the point of interest name, and at least one semantic feature associated with the point of interest name. The embodiment of the present application calculates the similarity of two point of interest names according to the semantic similarity feature of the point of interest name and at least one semantic feature associated with the point of interest name, thereby improving the accuracy of calculating the similarity of the point of interest names.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The embodiments of the present application relate to the field of Internet technology, in particular to the field of artificial intelligence technology, and specifically to a method, device, electronic device and storage medium for comparing names of points of interest. Background Art

[0002] When processing massive amounts of map POI intelligence data, it is necessary to determine whether different intelligence data describe the same POI. The name similarity between different POIs is the most important basis for judgment. The accuracy of name similarity calculation directly affects the accuracy of subsequent calculation processes.

[0003] The following two schemes are currently commonly used to calculate the similarity of map POI names:

[0004] (1) Deep semantic matching method based on the bidirectional LSTM (Long Short-Term Memory) model: This solution uses a deep neural network based on the bidirectional LSTM model to calculate the name similarity and obtain a similarity probability (0-1 continuous value, 0 represents dissimilarity and 1 represents similarity). (2) Deep semantic matching method based on attention: This solution uses the attention idea of ​​the Transformer framework to construct a deep neural network to calculate the name similarity and obtain a similarity probability (0-1 continuous value, 0 represents dissimilarity and 1 represents similarity).

[0005] However, these two solutions still have certain shortcomings: Solution (1) has a forward and reverse problem when calculating the similarity of the names of two map POIs, that is, when the order of the two names being compared is swapped, the name similarity calculated by the model is very different; Solution (2) can solve the problem of solution (1), but for the case where the POI names have a certain similarity but belong to different industries, the similarity calculation is inaccurate. Summary of the invention

[0006] The embodiments of the present application disclose a method, device, electronic device and medium for comparing names of points of interest, so as to achieve the purpose of accurately calculating the similarity of names of points of interest.

[0007] In a first aspect, an embodiment of the present application discloses a method for comparing names of points of interest, including:

[0008] Obtain the name of the first point of interest and the name of the second point of interest to be compared;

[0009] Calculating the similarity between the first point of interest name and the second point of interest name according to the semantic representation features corresponding to the first point of interest name and the second point of interest name respectively;

[0010] The semantic representation features include: a semantic similarity feature of a point of interest name, and at least one semantic feature associated with a point of interest name;

[0011] The semantic similarity feature of the point of interest name is a differentiated semantic feature between the point of interest name and other point of interest names, and the associated semantic feature of the point of interest name is a semantic feature of other attributes associated with the point of interest name.

[0012] One embodiment in the above application has the following advantages or beneficial effects: the similarity of two point of interest names is calculated based on the semantic similarity feature of the point of interest name and at least one semantic feature associated with the point of interest name, thereby improving the accuracy of calculating the similarity of the point of interest names compared to calculating the similarity of the two point of interest names only based on the feature representation of the point of interest names.

[0013] In addition, the method for comparing POI names according to the above embodiment of the present application may also have the following additional technical features:

[0014] Optionally, the semantic features associated with the POI name include: industry attribute features of the POI name, and / or word attribute features of each participle in the POI name;

[0015] The industry attribute features of the POI name are semantic features of the industry attributes of the industry to which the POI name belongs, and the word attribute features of each participle in the POI name are semantic features of the word attributes to which each participle in the POI name belongs.

[0016] One embodiment of the above application has the following advantages or beneficial effects: when determining the similarity between two POI names, the accuracy of similarity recognition can be further improved by combining the industry attribute characteristics of the POI names and / or the word attribute characteristics of each word in the POI names.

[0017] Optionally, calculating the similarity between the first point of interest name and the second point of interest name according to the semantic representation features corresponding to the first point of interest name and the second point of interest name respectively includes:

[0018] Inputting the first point of interest name and the second point of interest name into a pre-trained multi-task learning model respectively;

[0019] Obtaining a similarity between the first point of interest name and the second point of interest name output by the multi-task learning model;

[0020] The multi-task learning model includes: a point of interest name similarity recognition sub-model and at least one point of interest name associated feature recognition sub-model set in parallel; the point of interest name similarity recognition sub-model and each point of interest name associated feature recognition sub-model share the same semantic representation layer;

[0021] The semantic representation layer is used to obtain the semantic representation features corresponding to the first point of interest name and the second point of interest name respectively; the point of interest name similarity recognition sub-model is used to output the similarity between the first point of interest name and the second point of interest name.

[0022] One embodiment of the above application has the following advantages or beneficial effects: by sharing parameters of the semantic representation layers in multiple single-task deep neural networks, multiple single-task neural networks are merged into a multi-task neural network learning model. Since the parameter-shared semantic representation layer can learn a deeper semantic expression of the point of interest, the accuracy of the similarity recognition of the point of interest can be guaranteed.

[0023] Optionally, before obtaining the name of the first point of interest and the name of the second point of interest to be compared, the method further includes:

[0024] Obtain training samples that match the POI name similarity recognition sub-model and the POI name associated feature recognition sub-model respectively;

[0025] The multi-task learning model is trained using the training samples, and when a training end condition is met, the trained multi-task learning model is obtained.

[0026] One embodiment of the above application has the following advantages or beneficial effects: using different training samples to train the multi-task model separately, so that the shared semantic representation layer can learn the deeper semantic expression of the interest point, thereby ensuring the accuracy of the interest point similarity recognition of the trained multi-task model.

[0027] Optionally, the POI name similarity recognition sub-model specifically includes: the semantic representation layer and the cross-attention comparison layer;

[0028] The semantic representation layer is used to obtain semantic representation features corresponding to the two inputted POI names, and input each of the semantic representation features into the cross attention comparison layer for similarity comparison;

[0029] The cross attention comparison layer is used to output the similarity between the names of two points of interest.

[0030] Optionally, the interest point name associated feature recognition sub-model includes an interest point name industry classification sub-model; the interest point name associated feature recognition sub-model includes: the semantic representation layer and the fully connected classification layer;

[0031] The semantic representation layer is used to obtain semantic representation features corresponding to the inputted POI names, and input the semantic representation features into the fully connected classification layer for industry classification;

[0032] The fully connected classification layer is used to output the industry classification result of the point of interest name.

[0033] Optionally, the interest point name associated feature recognition sub-model further includes an interest point name word segmentation and parsing sub-model; the interest point name word segmentation and parsing sub-model includes: the semantic representation layer and the conditional random field layer;

[0034] The semantic representation layer is used to obtain semantic representation features corresponding to the inputted POI name, and input the semantic representation features into the conditional random field layer to recognize word attributes;

[0035] The conditional random field layer is used to output the word attributes of each word segment in the name of the point of interest.

[0036] Optionally, using the training samples to train the multi-task learning model, and when a training end condition is met, obtaining the trained multi-task learning model, includes:

[0037] Acquire a first training sample set matching the interest point name similarity recognition sub-model, and a second training sample set corresponding to each of the interest point name associated feature recognition sub-models;

[0038] Respectively obtaining a set number of training samples from the first training sample set and each of the second training sample sets, and sequentially adding them to the newly created multi-task training sample set;

[0039] After training the multi-task learning model using the multi-task training sample set, determining whether the multi-task learning model meets a training end condition;

[0040] If not, return to execute the operation of obtaining a set number of training samples from the first training sample set and each of the second training sample sets respectively, until the trained multi-task learning model is obtained when the training end condition is met.

[0041] Optionally, determining whether the multi-task learning model meets the training end condition includes:

[0042] Using a test sample set to perform a recognition accuracy test on the interest point name similarity recognition sub-model;

[0043] If the recognition accuracy of the interest point name similarity recognition sub-model exceeds a set threshold, it is determined that the training end condition is met.

[0044] One embodiment of the above application has the following advantages or beneficial effects: by verifying with verification samples, the accuracy of similarity recognition of the trained multi-task model can be ensured.

[0045] In a second aspect, an embodiment of the present application discloses a device for comparing names of points of interest, including:

[0046] A point of interest name acquisition module, used to acquire the first point of interest name and the second point of interest name to be compared;

[0047] a similarity calculation module, configured to calculate the similarity between the first point of interest name and the second point of interest name according to the semantic representation features corresponding to the first point of interest name and the second point of interest name respectively;

[0048] The semantic representation features include: a semantic similarity feature of a point of interest name, and at least one semantic feature associated with a point of interest name;

[0049] The industry attribute features of the POI name are semantic features of the industry attributes of the industry to which the POI name belongs, and the word attribute features of each participle in the POI name are semantic features of the word attributes to which each participle in the POI name belongs.

[0050] In a third aspect, an embodiment of the present application further provides an electronic device, including:

[0051] at least one processor; and

[0052] a memory communicatively connected to the at least one processor; wherein,

[0053] 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 so that the at least one processor can execute the point of interest name comparison method described in any embodiment of the present application.

[0054] In a fourth aspect, an embodiment of the present application further provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable the computer to execute the point of interest name comparison method described in any embodiment of the present application.

[0055] Other effects of the above optional manner will be described below in conjunction with specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present application.

[0057] Figure 1 is a flowchart of a method for comparing POI names according to the first embodiment of the present application;

[0058] Figure 2 is a flowchart of a method for comparing POI names according to a second embodiment of the present application;

[0059] Figure 3ais a flowchart of a multi-task learning model training method according to the third embodiment of the present application;

[0060] Figure 3b is a schematic diagram of the structure of a multi-task learning model according to the third embodiment of the present application;

[0061] Figure 4 is a schematic diagram of the structure of a device for comparing points of interest names according to a fourth embodiment of the present application;

[0062] Figure 5 It is a block diagram of an electronic device used to implement the method for comparing POI names in an embodiment of the present application. DETAILED DESCRIPTION

[0063] The following is a description of exemplary embodiments of the present application in conjunction with the accompanying drawings, including various details of the embodiments of the present application to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0064] Figure 1 1 is a flowchart of a method for comparing POI names in an embodiment of the present application. This embodiment can be applied to processing massive amounts of POI intelligence data on maps, such as determining whether different intelligence data describe the same POI. The method can be performed by a POI name comparison device, which is implemented in software and / or hardware, and is preferably configured in an electronic device, such as a server or a computer device. Figure 1 As shown, the method specifically includes the following:

[0065] S101: Obtain a first point of interest name and a second point of interest name to be compared.

[0066] Point of interest (POI) is a form of geographic information, which can be a building, a business or a bus stop. The attribute information of each POI generally includes name, address and coordinates. Therefore, when processing massive map POI intelligence data, the name of the first POI and the name of the second POI to be compared can be obtained based on the attribute information of the POI to be compared.

[0067] S102: Calculate the similarity between the first point of interest name and the second point of interest name according to the semantic representation features corresponding to the first point of interest name and the second point of interest name respectively.

[0068] Optionally, the semantic representation features corresponding to each other can be extracted from the first POI name and the second POI name respectively by natural language processing technology or based on a pre-trained model (e.g., a multi-task model), wherein the semantic representation features are exemplarily represented in the form of feature vectors or tensors, and the semantic representation features are digital representations of the POI names. When calculating the similarity between the first POI name and the second POI name, the similarity between the first POI name and the second POI name can be determined exemplarily by the cosine distance or Euclidean distance between the semantic representation features of the first POI name and the semantic representation features of the second POI name.

[0069] In the embodiment of the present application, the semantic representation features include: a semantic similarity feature of a point of interest name, and at least one semantic feature associated with a point of interest name. The semantic similarity feature of a point of interest name is a differentiated semantic feature between a point of interest name and other point of interest names, and the semantic feature associated with a point of interest name is a semantic feature of other attributes associated with a point of interest name.

[0070] Among them, the semantic similarity feature of the POI name can be optionally a feature representation corresponding to the name, for example, it can be a feature vector representation of the name; in order to make the calculation result accurate when calculating the name similarity between two POI names, it is necessary to maximize the difference between the semantic similarity features of the POI names of different POI names, and accordingly, when the similarity distance is finally calculated, the calculation result is more accurate. Therefore, the calculated semantic similarity feature of the POI name is the differentiated semantic feature between the POI name and the other POI names.

[0071] Among them, the method for obtaining the differentiated semantic feature can be to obtain a point of interest name similarity recognition model through training samples in advance, wherein the training samples include: two point of interest names pre-labeled with similarity calculation results. After the point of interest name similarity recognition model is successfully trained, the semantic representation layer in the point of interest name similarity recognition model can be used to obtain the point of interest name semantic similarity features corresponding to the point of interest names.

[0072] Through the above settings, when calculating the similarity between two POI names, POI names with similar names but different industry attributes can also be distinguished to improve calculation accuracy.

[0073] In this embodiment, in order to make the calculation result more accurate when calculating the name similarity between two points of interest names, in addition to considering the use of the semantic similarity features of the points of interest names, the point of interest name associated semantic features are further used. The point of interest name associated semantic features are semantic features of other attributes associated with the point of interest names, so as to further distinguish the difference values ​​of the semantic features of different point of interest names.

[0074] Optionally, the semantic features associated with the POI name may include: industry attribute features of the POI name, and / or word attribute features of each word segment in the POI name. For example, the POI name is "Shaxian Snacks", and its industry attribute is "food". Another example is the POI name "Peking University", which includes the word segments "Beijing" and "University". The word attribute of the word segment "Beijing" is geographical location, and the word attribute of the word segment "University" is "school".

[0075] Among them, a method for obtaining semantic features associated with one or more POI names can be to pre-train a POI name associated feature recognition model through training samples, for example, a POI name industry classification model, and train the POI name industry classification model by using POI names that are pre-labeled with POI industry classification results as training samples. After the POI name industry classification model is successfully trained, the semantic representation layer in the POI name industry classification model can be used to obtain the industry attribute features of the POI names corresponding to the POI names.

[0076] In an optional implementation of the present embodiment, a machine learning model corresponding to the semantic similarity features of the points of interest names and at least one semantic feature associated with the points of interest name may be pre-trained, and a semantic representation layer in the machine learning model may be obtained through training, and the semantic similarity features of the points of interest names corresponding to the first point of interest and the second point of interest, respectively, and at least one semantic feature associated with the point of interest name may be calculated, and the semantic representation features corresponding to the first point of interest name and the second point of interest name may be obtained by combining the above-mentioned semantic similarity features of the points of interest names and at least one semantic feature associated with the point of interest name.

[0077] In another optional implementation of the present embodiment, a multi-task learning model can be trained as a whole, in which different learning branches of the multi-task learning model correspond to different machine learning models, and the above machine learning models all share the same semantic representation layer. After the multi-task learning model is trained and generated, the semantic representation features obtained through the semantic representation layer can simultaneously learn the semantic similarity features of the points of interest names and at least one semantic feature associated with the point of interest name.

[0078] Correspondingly, if the semantic similarity feature of the POI name in the semantic representation features of the first POI name and the second POI name, and at least one POI name-associated semantic feature are independently separable, when calculating the similarity between the first POI name and the second POI name, optionally, the similarity result between the semantic similarity feature of the first POI name and the semantic similarity feature of the second POI name, and the similarity result between the associated semantic feature of the first POI name and the associated semantic feature of the second POI name are calculated, and the obtained similarity results are weighted summed to obtain the final similarity between the first POI name and the second POI name. Alternatively, if the semantic similarity feature of the POI name and at least one POI name-associated semantic feature are simultaneously integrated into the semantic representation features of the first POI name and the second POI name, the similarity result between the semantic representation features of the first POI name and the semantic representation features of the second POI name can be directly calculated as the similarity between the first POI name and the second POI name.

[0079] In an embodiment of the present application, the similarity of two point of interest names is calculated based on the semantic similarity feature of the point of interest name and at least one semantic feature associated with the point of interest name. Compared with calculating the similarity of the two point of interest names only based on the feature representation of the point of interest name, the accuracy of calculating the similarity of the point of interest names is improved.

[0080] Figure 2 is a flow chart of the method for comparing the names of points of interest of an embodiment of the present application. This embodiment is optimized on the basis of the above embodiment. Figure 2 , the method specifically comprises:

[0081] S201: Obtain a first point of interest name and a second point of interest name to be compared.

[0082] S202: Input the first point of interest name and the second point of interest name into a pre-trained multi-task learning model respectively.

[0083] In the embodiment of the present application, the pre-trained multi-task learning model includes: a sub-model for similarity recognition of interest points names set in parallel, and at least one sub-model for identifying features associated with interest points names. Among them, the sub-model for similarity recognition of interest points names is used to output the similarity between the first interest point name and the second interest point name, and the sub-model for similarity recognition of interest points names and the sub-models for identifying features associated with each interest point name share the same semantic representation layer.

[0084] It should be noted that at least one interest point name associated feature recognition submodel is only applicable to the training process of the multi-task learning model, and when the trained multi-task learning model is used to calculate the similarity of two interest point names, only the interest point name similarity recognition submodel in the multi-task model is actually used. Moreover, since the interest point name similarity recognition submodel and each interest point name associated feature recognition submodel share the same semantic representation layer, after the multi-task learning model is trained using training samples that match the interest point name similarity recognition submodel and each interest point name associated feature recognition submodel, the shared semantic representation layer can comprehensively and accurately extract the semantic representation features corresponding to each interest point name, for example, simultaneously obtain the interest point name semantic similarity features and at least one interest point name associated semantic feature, thereby ensuring the accuracy of subsequent recognition of the similarity between the two interest points.

[0085] Therefore, after obtaining the trained multi-task learning model, it is only necessary to input the first POI name and the second POI name to be compared into the multi-task learning model respectively, and the similarity of the two POI names can be determined according to the output of the multi-task model.

[0086] S203: Obtain a similarity between the first POI name and the second POI name output by the multi-task learning model.

[0087] After the first point of interest name and the second point of interest name are respectively input into the pre-trained multi-task learning model, the semantic representation layer of the multi-task learning model obtains the semantic representation features corresponding to the first point of interest name and the second point of interest name, respectively, and then calculates the similarity of the two point of interest names based on the respective semantic representation features of the two point of interest names.

[0088] The semantic similarity feature of the POI name is a differentiated semantic feature between the POI name and other POI names, and the associated semantic feature of the POI name is a semantic feature of other attributes associated with the POI name.

[0089] In this embodiment, by sharing parameters of the semantic representation layers in multiple single-task deep neural networks, multiple single-task neural networks are merged into a multi-task learning model. Since the parameter-shared semantic representation layer can identify deeper semantic expressions of points of interest, the accuracy of similarity recognition of points of interest can be guaranteed.

[0090] Figure 3a is a flowchart of the training method of the multi-task learning model provided in the embodiment of the present application, see Figure 3a , the method specifically comprises:

[0091] S301, obtaining training samples that match the interest point name similarity recognition sub-model and the interest point name associated feature recognition sub-models respectively.

[0092] S302: Use the training samples to train the multi-task learning model, and when a training end condition is met, obtain the trained multi-task learning model.

[0093] See also Figure 3b , which shows a schematic diagram of the structure of a multi-task learning model, which includes: a sub-model for similarity recognition of POI names, a sub-model for industry classification of POI names, and a sub-model for word segmentation and parsing of POI names, which are set in parallel, and the three sub-models share the same semantic representation layer. The sub-model for industry classification of POI names and the sub-model for word segmentation and parsing of POI names are sub-models for identifying associated features of POI names.

[0094] Among them, the interest point name similarity recognition sub-model specifically includes: the semantic representation layer and the cross-attention comparison layer; the training samples matched with the sub-model include two interest point names and the similarity results between the two interest points, such as "China Post University-China University of Political Science and Law-not similar". When training the interest point name similarity recognition sub-model, the semantic representation layer is used to obtain the semantic representation features corresponding to the two input interest point names, and input each of the semantic representation features into the cross-attention comparison layer for similarity comparison; the cross-attention comparison layer is used to output the similarity between the two interest point names.

[0095] The industry classification sub-model of the point of interest name includes: the semantic representation layer and the fully connected classification layer; the training samples matching the sub-model include the point of interest name, and the industry classification matching the point of interest name, such as "Shaxian snacks" - "food". When training the industry classification sub-model of the point of interest name, the semantic representation layer is used to obtain the semantic representation features corresponding to the input point of interest name, and input the semantic representation features to the fully connected classification layer for industry classification; the fully connected classification layer is used to output the industry classification results of the point of interest name.

[0096] The word segmentation and parsing submodel of the POI name includes: the semantic representation layer and the conditional random field (CRF) layer; the training samples matched with the submodel include the word segmentation results of the POI name, and the word attributes corresponding to each word segmentation. When training the word segmentation and parsing submodel of the POI name, the semantic representation layer is used to obtain the semantic representation features corresponding to the input POI name, and input the semantic representation features to the conditional random field layer for word attribute recognition; the conditional random field layer is used to output the word attributes of each word segmentation in the POI name. Since the POI name similarity recognition submodel, the POI name industry classification submodel, and the POI name word segmentation and parsing submodel share the same semantic representation layer, after the multi-task learning model is trained using the corresponding training samples, the shared semantic representation layer can comprehensively and accurately extract the semantic representation features corresponding to each POI name.

[0097] When training the multi-task model, first obtain a certain number of training samples that match the interest point name similarity recognition sub-model, and train the interest point name similarity recognition sub-model in the multi-task learning model. Then obtain a certain number of training samples that match the interest point name industry classification sub-model, and train the interest point name industry classification sub-model in the multi-task learning model. Finally, obtain a certain number of training samples corresponding to the interest point name word segmentation and parsing sub-model, and train the interest point name word segmentation and parsing sub-model in the multi-task learning model. The above training process is regarded as one round, and then the multi-task learning model is trained round by round according to the above process, and when the training end condition is met, the trained multi-task learning model is obtained.

[0098] In an optional implementation, the multi-task learning model is trained using the training samples, and when a training end condition is met, an operation of obtaining the trained multi-task learning model includes:

[0099] S1. Acquire a first training sample set matching the interest point name similarity recognition sub-model and a second training sample set corresponding to each of the interest point name associated feature recognition sub-models.

[0100] S2. Obtain a set number of training samples from the first training sample set and each of the second training sample sets respectively, and sequentially add them to the newly created multi-task training sample set.

[0101] S3. After training the multi-task learning model using the multi-task training sample set, determine whether the multi-task learning model meets the training end condition.

[0102] S4. If not, return to execute the operation of obtaining a set number of training samples from the first training sample set and each of the second training sample sets respectively, until the training end condition is met, and the trained multi-task learning model is obtained.

[0103] Among them, judging whether the multi-task learning model meets the training end condition includes: using a test sample set to test the recognition accuracy of the interest point name similarity recognition sub-model; if the recognition accuracy of the interest point name similarity recognition sub-model exceeds a set threshold, it is determined that the training end condition is met.

[0104] In an embodiment of the present application, by constructing a multi-task model with a common semantic representation layer and using sample data for training, the semantic representation layer of the obtained multi-task learning model can more accurately and comprehensively extract the semantic representation features of the POI names, thereby ensuring the accuracy of calculating the names of the two POIs.

[0105] Figure 4 400 is a schematic diagram of the structure of the device for comparing the names of points of interest of an embodiment of the present application. The present embodiment can be applied to the case of processing a large amount of map point of interest intelligence data. The device can implement the method for comparing the names of points of interest described in any embodiment of the present application. The device 400 specifically includes the following:

[0106] The POI name acquisition module 401 is used to acquire the first POI name and the second POI name to be compared;

[0107] A similarity calculation module 402, configured to calculate the similarity between the first interest point name and the second interest point name according to the semantic representation features corresponding to the first interest point name and the second interest point name respectively;

[0108] The semantic representation features include: a semantic similarity feature of a point of interest name, and at least one semantic feature associated with a point of interest name;

[0109] The semantic similarity feature of the point of interest name is a differentiated semantic feature between the point of interest name and other point of interest names, and the associated semantic feature of the point of interest name is a semantic feature of other attributes associated with the point of interest name.

[0110] Optionally, the semantic features associated with the POI name include: industry attribute features of the POI name, and / or word attribute features of each participle in the POI name;

[0111] The industry attribute features of the POI name are semantic features of the industry attributes of the industry to which the POI name belongs, and the word attribute features of each participle in the POI name are semantic features of the word attributes to which each participle in the POI name belongs.

[0112] Optionally, the similarity calculation module includes:

[0113] An input unit, used to input the first point of interest name and the second point of interest name into a pre-trained multi-task learning model respectively;

[0114] a similarity calculation unit, for obtaining a similarity between the first POI name and the second POI name output by the multi-task learning model;

[0115] The multi-task learning model includes: a point of interest name similarity recognition sub-model and at least one point of interest name associated feature recognition sub-model set in parallel; the point of interest name similarity recognition sub-model and each point of interest name associated feature recognition sub-model share the same semantic representation layer;

[0116] The semantic representation layer is used to obtain the semantic representation features corresponding to the first point of interest name and the second point of interest name respectively; the point of interest name similarity recognition sub-model is used to output the similarity between the first point of interest name and the second point of interest name.

[0117] Optionally, the device further comprises:

[0118] A sample acquisition module is used to acquire training samples that match the interest point name similarity recognition sub-model and the interest point name associated feature recognition sub-model respectively;

[0119] The training module is used to train the multi-task learning model using the training samples, and obtain the trained multi-task learning model when the training end condition is met.

[0120] Optionally, the POI name similarity recognition sub-model specifically includes: the semantic representation layer and the cross-attention comparison layer;

[0121] The semantic representation layer is used to obtain semantic representation features corresponding to the two inputted POI names, and input each of the semantic representation features into the cross attention comparison layer for similarity comparison;

[0122] The cross attention comparison layer is used to output the similarity between the names of two points of interest.

[0123] Optionally, the interest point name associated feature recognition sub-model includes an interest point name industry classification sub-model; the interest point name associated feature recognition sub-model includes: the semantic representation layer and the fully connected classification layer;

[0124] The semantic representation layer is used to obtain semantic representation features corresponding to the inputted POI names, and input the semantic representation features into the fully connected classification layer for industry classification;

[0125] The fully connected classification layer is used to output the industry classification result of the point of interest name.

[0126] Optionally, the interest point name associated feature recognition sub-model further includes an interest point name word segmentation and parsing sub-model; the interest point name word segmentation and parsing sub-model includes: the semantic representation layer and the conditional random field layer;

[0127] The semantic representation layer is used to obtain semantic representation features corresponding to the inputted POI name, and input the semantic representation features into the conditional random field layer to recognize word attributes;

[0128] The conditional random field layer is used to output the word attributes of each word segment in the name of the point of interest.

[0129] Optionally, the training module includes:

[0130] An acquisition unit, used to acquire a first training sample set matching the interest point name similarity recognition sub-model, and a second training sample set corresponding to each of the interest point name associated feature recognition sub-models;

[0131] A selection unit, used to obtain a set number of training samples from the first training sample set and each of the second training sample sets, and sequentially add them to the newly created multi-task training sample set;

[0132] A training unit, configured to train the multi-task learning model using the multi-task training sample set, and then determine whether the multi-task learning model meets a training end condition;

[0133] A judgment unit is used to return to execute the operation of obtaining a set number of training samples from the first training sample set and each of the second training sample sets respectively if it is judged that the training end condition is not met, until the trained multi-task learning model is obtained when the training end condition is met.

[0134] Optionally, determining whether the multi-task learning model meets the training end condition includes:

[0135] Using a test sample set to perform a recognition accuracy test on the interest point name similarity recognition sub-model;

[0136] If the recognition accuracy of the interest point name similarity recognition sub-model exceeds a set threshold, it is determined that the training end condition is met.

[0137] The POI name comparison device provided in the embodiment of the present application can execute the POI name comparison method provided in any embodiment of the present application, and has the corresponding functional modules and beneficial effects of the execution method. For the contents not described in detail in this embodiment, please refer to the description in any method embodiment of the present application.

[0138] According to an embodiment of the present application, the present application also provides an electronic device and a readable storage medium.

[0139] like Figure 5 , is a block diagram of an electronic device according to a method for comparing points of interest names according to an embodiment of the present application. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, 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 processing, cellular phones, smart phones, 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 application described and / or required herein.

[0140] like Figure 5 As shown, the electronic device includes: one or more processors 501, a memory 502, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. The various components are interconnected using different buses and can be installed on a common mainboard or installed in other ways as needed. The processor can process instructions executed in the electronic device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In other embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple electronic devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 5 A processor 501 is taken as an example.

[0141] The memory 502 is a non-transitory computer-readable storage medium provided in the present application. The memory stores instructions executable by at least one processor to enable the at least one processor to perform the method for comparing the names of points of interest provided in the present application. The non-transitory computer-readable storage medium of the present application stores computer instructions, which are used to enable a computer to perform the method for comparing the names of points of interest provided in the present application.

[0142] The memory 502 is a non-transient computer-readable storage medium that can be used to store non-transient software programs, non-transient computer executable programs and modules, such as the program instructions / modules corresponding to the method for comparing the names of points of interest in the embodiment of the present application (for example, the attached Figure 4 The processor 501 executes various functional applications and data processing of the server by running the non-transient software programs, instructions and modules stored in the memory 502, that is, the POI name comparison method in the above method embodiment is implemented.

[0143] The memory 502 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and applications required for at least one function; the data storage area may store data created according to the use of an electronic device that implements the method for comparing the names of points of interest, etc. In addition, the memory 502 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory 502 may optionally include a memory remotely arranged relative to the processor 501, and these remote memories may be connected to the electronic device that implements the method for comparing the names of points of interest via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0144] The electronic device implementing the method for comparing the names of points of interest may further include: an input device 503 and an output device 504. The processor 501, the memory 502, the input device 503 and the output device 504 may be connected via a bus or other means. Figure 5 The example of connecting through bus is taken in the following.

[0145] The input device 503 can receive input digital or character information, and generate key signal input related to user settings and function control of the electronic device implementing the method for comparing the names of points of interest, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, an indicator rod, one or more mouse buttons, a trackball, a joystick and other input devices. The output device 504 may include a display device, an auxiliary lighting device (e.g., an LED) and a tactile feedback device (e.g., a vibration motor), etc. The display device may include, but is not limited to, a liquid crystal display (LCD), a light emitting diode (LED) display and a plasma display. In some embodiments, the display device may be a touch screen.

[0146] Various implementations of the systems and techniques described herein can be realized in digital electronic circuit systems, integrated circuit systems, dedicated ASICs (application specific integrated circuits), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including 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.

[0147] These computer programs (also referred to as programs, software, software applications, or code) include machine instructions for programmable processors and can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, device, and / or means (e.g., disk, optical disk, memory, programmable logic device (PLD)) for providing machine instructions and / or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor.

[0148] 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).

[0149] The systems and techniques described herein may be implemented in a computing system that includes backend 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 frontend components (e.g., a user computer with 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 backend components, middleware components, or frontend components. The components of the system may 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), the Internet, and a blockchain network.

[0150] A computer system may include clients and servers. Clients and servers are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship to each other.

[0151] According to the technical solution of the embodiment of the present application, the similarity of two point of interest names is calculated based on the semantic similarity feature of the point of interest name and at least one semantic feature associated with the point of interest name. Compared with calculating the similarity of the two point of interest names only based on the feature representation of the point of interest name, the accuracy of calculating the similarity of the point of interest names is improved.

[0152] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this application can be executed in parallel, sequentially or in different orders, as long as the expected results of the technical solution disclosed in this application can be achieved, and this document is not limited here.

[0153] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A method for comparing names of points of interest, characterized in that: include: Obtain the name of the first point of interest and the name of the second point of interest to be compared; Calculating the similarity between the first point of interest name and the second point of interest name according to the semantic representation features corresponding to the first point of interest name and the second point of interest name respectively; The semantic representation features include: a semantic similarity feature of a point of interest name, and at least one semantic feature associated with a point of interest name; The semantic similarity feature of the POI name is a differentiated semantic feature between the POI name and other POI names, and the associated semantic feature of the POI name is a semantic feature of other attributes associated with the POI name; The semantic features associated with the POI name include: industry attribute features of the POI name and word attribute features of each word in the POI name; The industry attribute feature of the POI name is the semantic feature of the industry attribute of the industry to which the POI name belongs, and the word attribute feature of each participle in the POI name is the semantic feature of the word attribute to which each participle in the POI name belongs; The method of calculating the similarity between the first point of interest name and the second point of interest name according to the semantic representation features corresponding to the first point of interest name and the second point of interest name respectively includes: Inputting the first point of interest name and the second point of interest name into a pre-trained multi-task learning model respectively; Obtaining a similarity between the first point of interest name and the second point of interest name output by the multi-task learning model; The multi-task learning model includes: a point of interest name similarity recognition sub-model set in parallel, and at least one point of interest name associated feature recognition sub-model; the point of interest name similarity recognition sub-model and each point of interest name associated feature recognition sub-model share the same semantic representation layer; the point of interest name similarity recognition sub-model specifically includes: the semantic representation layer and the cross-attention comparison layer; the point of interest name associated feature recognition sub-model includes a point of interest name industry classification sub-model; the point of interest name industry classification sub-model includes: the semantic representation layer and the fully connected classification layer; the point of interest name associated feature recognition sub-model also includes a point of interest name word segmentation and parsing sub-model; the point of interest name word segmentation and parsing sub-model includes: the semantic representation layer and the conditional random field layer; The semantic representation layer is used to obtain the semantic representation features corresponding to the first point of interest name and the second point of interest name respectively; the point of interest name similarity recognition sub-model is used to output the similarity between the first point of interest name and the second point of interest name.

2. The method according to claim 1, characterized in that Before obtaining the first POI name and the second POI name to be compared, the method further includes: Obtain training samples that match the POI name similarity recognition sub-model and the POI name associated feature recognition sub-model respectively; The multi-task learning model is trained using the training samples, and when a training end condition is met, the trained multi-task learning model is obtained.

3. The method according to claim 2, characterized in that The semantic representation layer is used to obtain semantic representation features corresponding to the two inputted POI names, and input each of the semantic representation features into the cross attention comparison layer for similarity comparison; The cross attention comparison layer is used to output the similarity between the names of two points of interest.

4. The method according to claim 2, characterized in that: The semantic representation layer is used to obtain semantic representation features corresponding to the inputted POI names, and input the semantic representation features into the fully connected classification layer for industry classification; The fully connected classification layer is used to output the industry classification result of the point of interest name.

5. The method according to claim 2, characterized in that: The semantic representation layer is used to obtain semantic representation features corresponding to the inputted POI name, and input the semantic representation features into the conditional random field layer for word attribute recognition; The conditional random field layer is used to output the word attributes of each word segment in the name of the point of interest.

6. The method according to claim 2, characterized in that The multi-task learning model is trained using the training samples, and when a training end condition is met, the trained multi-task learning model is obtained, including: Acquire a first training sample set matching the interest point name similarity recognition sub-model, and a second training sample set corresponding to each of the interest point name associated feature recognition sub-models; Respectively obtaining a set number of training samples from the first training sample set and each of the second training sample sets, and sequentially adding them to the newly created multi-task training sample set; After training the multi-task learning model using the multi-task training sample set, determining whether the multi-task learning model meets a training end condition; If not, return to execute the operation of obtaining a set number of training samples from the first training sample set and each of the second training sample sets respectively, until the trained multi-task learning model is obtained when the training end condition is met.

7. The method according to claim 6, characterized in that Determining whether the multi-task learning model meets the training end condition includes: Using a test sample set to perform a recognition accuracy test on the interest point name similarity recognition sub-model; If the recognition accuracy of the interest point name similarity recognition sub-model exceeds a set threshold, it is determined that the training end condition is met.

8. A device for comparing names of points of interest, characterized in that: include: A point of interest name acquisition module, used to acquire the first point of interest name and the second point of interest name to be compared; a similarity calculation module, configured to calculate the similarity between the first point of interest name and the second point of interest name according to the semantic representation features corresponding to the first point of interest name and the second point of interest name respectively; The semantic representation features include: a semantic similarity feature of a point of interest name, and at least one semantic feature associated with a point of interest name; The semantic similarity feature of the POI name is a differentiated semantic feature between the POI name and other POI names, and the associated semantic feature of the POI name is a semantic feature of other attributes associated with the POI name; The semantic features associated with the POI name include: industry attribute features of the POI name and word attribute features of each word in the POI name; The industry attribute feature of the POI name is the semantic feature of the industry attribute of the industry to which the POI name belongs, and the word attribute feature of each participle in the POI name is the semantic feature of the word attribute to which each participle in the POI name belongs; Wherein, the similarity calculation module includes: An input unit, used to input the first point of interest name and the second point of interest name into a pre-trained multi-task learning model respectively; a similarity calculation unit, for obtaining a similarity between the first POI name and the second POI name output by the multi-task learning model; The multi-task learning model includes: a point of interest name similarity recognition sub-model set in parallel, and at least one point of interest name associated feature recognition sub-model; the point of interest name similarity recognition sub-model and each point of interest name associated feature recognition sub-model share the same semantic representation layer; the point of interest name similarity recognition sub-model specifically includes: the semantic representation layer and the cross-attention comparison layer; the point of interest name associated feature recognition sub-model includes a point of interest name industry classification sub-model; the point of interest name industry classification sub-model includes: the semantic representation layer and the fully connected classification layer; the point of interest name associated feature recognition sub-model also includes a point of interest name word segmentation and parsing sub-model; the point of interest name word segmentation and parsing sub-model includes: the semantic representation layer and the conditional random field layer; The semantic representation layer is used to obtain the semantic representation features corresponding to the first point of interest name and the second point of interest name respectively; the point of interest name similarity recognition sub-model is used to output the similarity between the first point of interest name and the second point of interest name.

9. The device according to claim 8, characterized in that The device also includes: A sample acquisition module is used to acquire training samples that match the interest point name similarity recognition sub-model and the interest point name associated feature recognition sub-model respectively; The training module is used to train the multi-task learning model using the training samples, and obtain the trained multi-task learning model when the training end condition is met.

10. The device according to claim 9, characterized in that The semantic representation layer is used to obtain semantic representation features corresponding to the two inputted POI names, and input each of the semantic representation features into the cross attention comparison layer for similarity comparison; The cross attention comparison layer is used to output the similarity between the names of two points of interest.

11. The device according to claim 9, characterized in that The semantic representation layer is used to obtain semantic representation features corresponding to the inputted POI names, and input the semantic representation features into the fully connected classification layer for industry classification; The fully connected classification layer is used to output the industry classification result of the point of interest name.

12. The device according to claim 9, characterized in that The semantic representation layer is used to obtain semantic representation features corresponding to the inputted POI name, and input the semantic representation features into the conditional random field layer for word attribute recognition; The conditional random field layer is used to output the word attributes of each word segment in the name of the point of interest.

13. The device according to claim 9, characterized in that The training module includes: An acquisition unit, used to acquire a first training sample set matching the interest point name similarity recognition sub-model, and a second training sample set corresponding to each of the interest point name associated feature recognition sub-models; A selection unit, used to obtain a set number of training samples from the first training sample set and each of the second training sample sets, and sequentially add them to the newly created multi-task training sample set; A training unit, configured to train the multi-task learning model using the multi-task training sample set, and then determine whether the multi-task learning model meets a training end condition; A judgment unit is used to return to execute the operation of obtaining a set number of training samples from the first training sample set and each of the second training sample sets respectively if it is judged that the training end condition is not met, until the trained multi-task learning model is obtained when the training end condition is met.

14. The device according to claim 13, characterized in that Determining whether the multi-task learning model meets the training end condition includes: Using a test sample set to perform a recognition accuracy test on the interest point name similarity recognition sub-model; If the recognition accuracy of the interest point name similarity recognition sub-model exceeds a set threshold, it is determined that the training end condition is met.

15. An electronic device, characterized in that: include: at least one processor; as well as 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 so that the at least one processor can execute the POI name comparison method according to any one of claims 1 to 7.

16. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to enable the computer to execute the point of interest name comparison method according to any one of claims 1 to 7.