Non-standard address standardization method and device

Abstract

The invention provides a non-standard address standardization method and device which are applied to a data processing server in a non-standard address standardization system, and the method comprises the steps: receiving an input non-standard address; the search engine server queries a recall result corresponding to the non-standard address; according to a set multi-dimensional address recall rule, the first N addresses in the front are taken as candidate addresses from recall results returned by the search engine server, and a first address list corresponding to rough sorting is obtained; the multi-dimensional address recall rule comprises query based on a text feature extraction algorithm, query based on Boolean and query based on a preset weight strategy; performing fine sorting on the N candidate addresses based on a BERT semantic vector model to obtain a second address list corresponding to the fine sorting; and performing fusion sorting on the first address list and the second address list by using a preset fusion algorithm to obtain standardized address data after fusion sorting, and storing the standardized address data to a database server. According to the scheme provided by the invention, the processing efficiency is improved, and the matching accuracy is effectively improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method and device for standardizing non-standard addresses. Background Art

[0002] With the rise of artificial intelligence technology, it has been widely used in urban planning and public services, such as the address standardization system of smart government affairs, which can process an average of tens of thousands of non-standard addresses every day. Non-standard addresses refer to address information that does not conform to the standard address format; the process of converting non-standard data into data that conforms to a specific standard format is called non-standard address standardization. At present, in the field of non-standard address standardization, it is mainly based on a single sorting and matching algorithm. For example, keyword-based matching, regular expression-based matching, and semantic vector-based matching can be used.

[0003] For the keyword-based matching method, first extract keywords such as street name, house number, city name, etc. from the non-standard address, then match the extracted keywords with the keywords in the standard address library, sort them according to the matching degree, and select the address corresponding to the result with the highest matching degree as the standardized address.

[0004] For the matching method based on regular expressions, first build a regular expression according to the common address format, then use the regular expression to match the non-standard address, extract the street, house number, city and other information, and then match the extracted information with the standard address library, and select the address corresponding to the result with the highest matching degree as the standardized address.

[0005] For the matching method based on semantic vectors, the non-standard address is first converted into a semantic vector, and then the semantic vector is matched with the vector in the standard address library, usually using a vector similarity calculation method such as cosine similarity; then the vectors are sorted according to their similarity, and the address corresponding to the result with the highest similarity is selected as the standardized address.

[0006] Although the above methods can achieve the standardization of non-standard addresses, since the existing technology is based on a single sorting and matching algorithm, no matter which method the existing technology adopts, as the non-standard address structure becomes more complicated, especially when processing large-scale non-standard addresses, it is easy to have problems of low processing efficiency and poor matching accuracy. Summary of the invention

[0007] The present application shows a non-standard address standardization method and device, aiming to improve the processing efficiency of the non-standard address standardization process and improve the matching accuracy.

[0008] The present application provides a non-standard address standardization method, which is applied to a data processing server in a non-standard address standardization system, wherein the non-standard address standardization system further includes a search engine server and a database server, wherein the data processing server is connected to the search engine server via an HTTP / HTTPS protocol, and the data processing server is connected to the database via a database connection protocol, wherein the method includes:

[0009] Receive non-standard addresses input by users;

[0010] Sending an address recall request carrying a non-standard address to the search engine server, so that after receiving the address recall request, the search engine server queries the database server for a recall result corresponding to the non-standard address and sends the result to the data processing server;

[0011] According to the set multi-dimensional address recall rules, the first N addresses with the highest ranking are taken as candidate addresses from the recall results returned by the search engine server, and a first address list corresponding to the rough ranking is obtained; wherein the multi-dimensional address recall rules include query based on text feature extraction algorithm, query based on Boolean and query based on preset weight strategy;

[0012] Based on the BERT semantic vector model, the N candidate addresses are precisely sorted to obtain a second address list corresponding to the precise sorting; the BERT semantic vector model is a semantic vector model that performs geographic location fine-tuning based on geographic location information;

[0013] The first address list and the second address list are merged and sorted using a preset fusion algorithm to obtain standardized address data after fusion and sorting, and the data is stored in the database server.

[0014] Optionally, the step of taking the top N addresses in the recall results returned by the search engine server as candidate addresses according to the set multi-dimensional address recall rule to obtain a first address list corresponding to the rough sorting includes:

[0015] Calculating the first weights corresponding to the recall results returned by the search engine server by using the TF·IDF algorithm, and sorting the results in descending order according to the first weights;

[0016] Perform Boolean filtering on the sorted recall results according to the set Boolean query conditions;

[0017] Adjust the weights of the recall results after Boolean filtering based on the preset weight strategy to obtain second weights corresponding to each recall result;

[0018] The second weights are sorted in descending order to obtain a first address list corresponding to the rough sorting.

[0019] Optionally, the step of adjusting the weights of the recall results after Boolean filtering based on the preset weight strategy to obtain a second weight corresponding to each recall result includes:

[0020] The product of the first weight and the weight empirical coefficient is used as the adjusted second weight.

[0021] Optionally, the step of finely sorting the N candidate addresses based on the BERT semantic vector model to obtain a second address list corresponding to the fine sorting includes:

[0022] Convert the N candidate addresses and the non-standard address into corresponding semantic vectors respectively;

[0023] Calculating the cosine similarity between the semantic vector of each address in the N candidate addresses and the semantic vector of the candidate address to obtain N similarity values;

[0024] The N similarity values ​​are sorted in descending order to obtain a second address list.

[0025] Optionally, the value of N is a natural number between 20 and 50.

[0026] Optionally, the step of fusing and sorting the first address list and the second address list using a preset fusion algorithm to obtain fused standardized address data includes:

[0027] Based on the inverse sorting fusion algorithm, the first address list and the second address list are fused and sorted to obtain standardized address data after fusion and sorting.

[0028] Optionally, before the step of storing in the database server, the method further comprises:

[0029] Based on the preset entity recognition model, the standardized address data after fusion sorting is structured sorted;

[0030] The standardized address data is updated using the structured candidate address list.

[0031] Optionally, the step of performing structured sorting on the standardized address data after fusion sorting based on a preset entity recognition model includes:

[0032] Using an open source address entity recognition model, entity parsing and recognition are performed on the candidate address and the non-standard address to obtain entity recognition results; the entity recognition results at least include: province, city, district, township, village, street, building, house number;

[0033] Calculating the number of common entities between the entity recognition result of each address in the candidate address and the entity recognition result of the non-standard address respectively;

[0034] According to the number of common entities corresponding to each address in the candidate addresses and the proportion of the non-standard address entity recognition results, the standardized address data after the fusion sorting is structured sorted in descending order to obtain a structured candidate address list.

[0035] The present application also provides a non-standard address standardization method, which is applied to a search engine server in a non-standard address standardization system. The non-standard address standardization system also includes a data processing server and a database server. The data processing server is connected to the search engine server via an HTTP / HTTPS protocol, and the data processing server is connected to the database via a database connection protocol. The method includes:

[0036] receiving an address recall request sent by the data processing server, wherein the address recall request carries a non-standard address input by a user and received by the data processing server;

[0037] In response to the address recall request, the recall result corresponding to the non-standard address is queried from the database server and sent to the data processing server, so that the data processing server: according to the set multi-dimensional address recall rule, the first N addresses with the highest ranking are taken as candidate addresses from the recall results returned by the search engine server, and a first address list corresponding to the rough ranking is obtained; wherein the multi-dimensional address recall rule includes query based on text feature extraction algorithm, query based on Boolean query and query based on preset weight strategy; the N candidate addresses are fine-sorted based on the BERT semantic vector model to obtain a second address list corresponding to the fine-sort; the BERT semantic vector model is a semantic vector model that performs geographic location fine-tuning based on geographic location information; the first address list and the second address list are fused and sorted using a preset fusion algorithm to obtain standardized address data after fusion and sorting, and stored in the database server.

[0038] The present application provides a non-standard address standardization device, which is applied to a data processing server in a non-standard address standardization system. The non-standard address standardization system also includes a search engine server and a database server. The device includes:

[0039] An address receiving module, used to receive a non-standard address input by a user;

[0040] A recall request module, used to send an address recall request carrying a non-standard address to the search engine server, so that after receiving the address recall request, the search engine server queries the database server for a recall result corresponding to the non-standard address and sends it to the data processing server;

[0041] A rough sorting module is used to select the top N addresses from the recall results returned by the search engine server as candidate addresses according to the set multi-dimensional address recall rules, and obtain a first address list corresponding to the rough sorting; wherein the multi-dimensional address recall rules include query based on text feature extraction algorithm, query based on Boolean and query based on preset weight strategy;

[0042] A fine sorting module, used for fine sorting the N candidate addresses based on a BERT semantic vector model to obtain a second address list corresponding to the fine sorting; the BERT semantic vector model is a semantic vector model that performs geographic location fine-tuning based on geographic location information;

[0043] The fusion sorting module is used to use a preset fusion algorithm to perform fusion sorting on the first address list and the second address list, obtain standardized address data after fusion sorting, and store it in the database server.

[0044] Optionally, the rough sorting module is used to:

[0045] Calculating the first weights corresponding to the recall results returned by the search engine server by using the TF·IDF algorithm, and sorting the results in descending order according to the first weights;

[0046] Perform Boolean filtering on the sorted recall results according to the set Boolean query conditions;

[0047] Adjust the weights of the recall results after Boolean filtering based on the preset weight strategy to obtain second weights corresponding to each recall result;

[0048] The second weights are sorted in descending order to obtain a first address list corresponding to the rough sorting.

[0049] Optionally, the rough sorting module is specifically configured to: use the product of the first weight and the weight empirical coefficient as the adjusted second weight.

[0050] Optionally, the fine sorting module is used to:

[0051] Convert the N candidate addresses and the non-standard address into corresponding semantic vectors respectively;

[0052] Calculating the cosine similarity between the semantic vector of each address in the N candidate addresses and the semantic vector of the candidate address to obtain N similarity values;

[0053] The N similarity values ​​are sorted in descending order to obtain a second address list.

[0054] Optionally, the value of N is a natural number between 20 and 50.

[0055] Optionally, the fusion sorting module is used to:

[0056] Based on the inverse sorting fusion algorithm, the first address list and the second address list are fused and sorted to obtain standardized address data after fusion and sorting.

[0057] Optionally, the device further comprises:

[0058] The structured sorting module is used to perform structured sorting on the standardized address data after fusion sorting based on a preset entity recognition model; and update the standardized address data using the structured candidate address list.

[0059] The present application also provides a non-standard address standardization device, which is applied to a search engine server in a non-standard address standardization system. The non-standard address standardization system also includes a data processing server and a database server. The device includes:

[0060] A recall request receiving module, configured to receive an address recall request sent by the data processing server, wherein the address recall request carries a non-standard address input by a user and received by the data processing server;

[0061] The address recall module is used to respond to the address recall request, query the recall result corresponding to the non-standard address from the database server and send it to the data processing server, so that the data processing server: according to the set multi-dimensional address recall rule, take the top N addresses from the recall result returned by the search engine server as candidate addresses, and obtain a first address list corresponding to the rough sorting; wherein the multi-dimensional address recall rule includes query based on text feature extraction algorithm, query based on Boolean query and query based on preset weight strategy; based on the BERT semantic vector model, the N candidate addresses are fine-sorted to obtain a second address list corresponding to the fine sorting; the BERT semantic vector model is a semantic vector model that fine-tunes the geographic location based on geographic location information; the first address list and the second address list are fused and sorted using a preset fusion algorithm to obtain standardized address data after fusion sorting, and stored in the database server.

[0062] The present application provides an electronic device, comprising: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to execute the method as described in any one of the above aspects.

[0063] The present application provides a non-temporary computer-readable storage medium. When instructions in the storage medium are executed by a processor of an electronic device, the electronic device can execute the method described in any one of the above aspects.

[0064] The present application provides a computer program product. When instructions in the computer program product are executed by a processor of an electronic device, the electronic device is enabled to execute the method as described in any one of the above aspects.

[0065] The technical solution provided by this application may have the following beneficial effects:

[0066] This application follows the multi-dimensional address recall rules and only takes the top N recall results returned by the search engine server as candidate data, rather than performing a full match of the standard address library for standard addresses, which greatly reduces the computational overhead and thus improves processing efficiency. Moreover, the multi-way merge fusion sorting algorithm that combines coarse sorting with fine sorting significantly improves the matching accuracy of non-standard address standardization. Experimental data show that the accuracy of non-standard address standardization based on the test set is as high as 86%. In addition, this application can adapt to different types of non-standard addresses, such as spelling errors, abbreviations, missing information and other complex situations, and has strong adaptability and enhanced robustness. In addition, through the process of automated data recall, coarse sorting, fine sorting and fusion sorting, the need for manual intervention is greatly reduced, reducing maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 A flow chart of the non-standard address standardization method provided by this application;

[0068] Figure 2 The architecture diagram of the non-standard address standardization system provided for this application;

[0069] Figure 3 Another flow chart of the non-standard address standardization method provided for this application;

[0070] Figure 4 Another flow chart of the non-standard address standardization method provided by this application;

[0071] Figure 5 A structural diagram of a non-standard address standardization device provided in this application;

[0072] Figure 6 Another structural diagram of the non-standard address standardization device provided by this application;

[0073] Figure 7 Another structural diagram of the non-standard address standardization device provided by this application;

[0074] Figure 8 A block diagram of an electronic device provided for this application;

[0075] Fig. 9 A block diagram of another electronic device provided by the present application. DETAILED DESCRIPTION

[0076] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0077] Since the prior art is implemented based on a single sorting and matching algorithm, and no matter what method the prior art adopts, as the non-standard address structure becomes more complicated, especially when processing large-scale non-standard addresses, it is easy to have problems of low processing efficiency and poor matching accuracy. In addition, the keyword-based matching method requires continuous updating and maintenance of the keyword dictionary to cope with new address formats and changes; building and maintaining complex regular expressions also requires a lot of time and manpower costs; based on a single sorting and matching algorithm, in order to cope with the diversity and complexity of non-standard addresses, it is necessary to regularly train and update the model to ensure matching accuracy. It can be seen that the maintenance cost of the prior art is very high. Therefore, in order to improve the processing efficiency of the non-standard address standardization process and improve the matching accuracy, the present application provides the following non-standard address standardization method and device.

[0078] The following is an explanation of the non-standard address standardization method provided by this application.

[0079] Embodiment 1

[0080] See also Figure 1 and Figure 2 , Figure 1 A flow chart of a non-standard address standardization method provided by the present application. The present application provides a non-standard address standardization method, such as Figure 2 The data processing server applied to the non-standard address standardization system further comprises a search engine server and a database server. The data processing server is connected to the search engine server via an HTTP / HTTPS protocol, and the data processing server is connected to the database via a database connection protocol (such as a JDBC protocol).

[0081] In a specific implementation, (1) the data processing server device information is as follows: Form, high-performance computing server, equipped with multi-core CPU and large-capacity memory. Type, general-purpose server. Function, responsible for the input of non-standard addresses, data recall, rough sorting, fine sorting and RRF fusion sorting calculation. (2) The search engine server device information is as follows: Form, distributed search engine cluster, with high availability and scalability. Type, dedicated search engine server. Function, store and index standard address data, and provide query services based on TF·IDF algorithm, bool query and weight strategy. (3) The database server device information is as follows: Form, relational database server, such as MySQL or PostgreSQL. Type, general-purpose database server. Function: store and manage standard address data, and provide data query and update services.

[0082] Specifically, the present application is applicable to non-standard address standardization scenarios, such as smart government address standardization systems, smart city management systems, etc. The method may include the following steps:

[0083] Step S101: receiving a non-standard address input by a user.

[0084] Step S102: sending an address recall request carrying a non-standard address to the search engine server, so that after receiving the address recall request, the search engine server queries the database server for a recall result corresponding to the non-standard address and sends it to the data processing server.

[0085] In one case, the first address list corresponding to the rough sorting can be obtained as follows:

[0086] (1) Calculate the first weights corresponding to the recall results returned by the search engine server using the TF·IDF algorithm, and sort the results in descending order according to the first weights.

[0087] TF·IDF (Term Frequency-Inverse Document Frequency, TF·IDF algorithm) is a commonly used text feature extraction method used to measure the importance of a word in a document.

[0088] (2) Perform Boolean filtering on the sorted recall results according to the set Boolean query conditions.

[0089] (3) The weights of the recall results after Boolean filtering are adjusted based on the preset weight strategy to obtain a second weight corresponding to each recall result.

[0090] For example, the product of the first weight and the weight empirical coefficient can be used as the adjusted second weight. Preferably, the value of the weight empirical coefficient is a positive number greater than 1, and the preferred value is 2. It should be noted that the method for determining the second weight is a preferred implementation method provided by this application and should not be understood as a limitation on this application; moreover, the weight empirical coefficient values ​​listed here are only preferred methods and are not intended to be a limitation on this application.

[0091] (4) Sorting the second weights in descending order to obtain a first address list corresponding to the rough sorting.

[0092] Step S103: According to the set multi-dimensional address recall rules, the top N addresses are taken from the recall results returned by the search engine server as candidate addresses to obtain a first address list corresponding to the rough sorting; wherein the multi-dimensional address recall rules include queries based on text feature extraction algorithms, queries based on Boolean queries, and queries based on preset weight strategies.

[0093] In one case, the second address list corresponding to the precise sorting can be obtained as follows:

[0094] (1) converting the N candidate addresses and the non-standard address into corresponding semantic vectors respectively;

[0095] (2) calculating the cosine similarity between the semantic vector of each of the N candidate addresses and the semantic vector of the candidate address to obtain N similarity values;

[0096] (3) Sort the N similarity values ​​in descending order to obtain a second address list.

[0097] It should be noted that the TF·IDF algorithm is used to calculate the weight of each word in the non-standard address, so as to recall candidate addresses related to the non-standard address.

[0098] Bool query: is a Boolean query method that combines multiple query conditions through logical operators such as AND, OR, and NOT to accurately control the candidate addresses to be recalled. Boolean query supports the following four clauses: ① must: conditions that must be matched, similar to logical AND, and matching documents will participate in the scoring; ② should: conditions that can be matched, similar to logical OR, and matching documents will participate in the scoring; ③ must_not: conditions that must not be matched, similar to logical NOT, and matching documents will not participate in the scoring; ④ filter: conditions that must be matched, similar to must, but will not affect the scoring and are only used for filtering. The filtering mechanism of Boolean query is one of the core functions of Boolean query, which is mainly used to quickly filter documents without affecting the scoring of documents.

[0099] In addition, based on the bool query, the weights are adjusted through a preset weight strategy to further sort the words according to their weights, ensuring that the recalled candidate addresses are more relevant to the non-standard addresses.

[0100] The candidate addresses are sorted and scored through the above query method in the multi-dimensional recall rule, and the scope of standard address matching is narrowed based on the sorting score, which reduces the computing overhead and greatly shortens the overall system response time.

[0101] In one case, the value of N is a natural number between 20 and 50. Of course, the value range of N is only the preferred value range provided by this application, and more preferably, the value of N can be set to 30. It should be noted that the above is only the preferred value range and preferred value of the number of candidate addresses N provided by this application, and should not be understood as a limitation on this application.

[0102] In the prior art, whether it is keyword-based matching, regular expression-based matching, or semantic vectorization-based matching, they all belong to a single algorithm. Moreover, when processing large-scale non-standard addresses, they are all matched based on the full standard address library. It can be seen that its computational efficiency is too low. In addition, for algorithms with high computational complexity such as semantic vectorization, a large amount of computing resources are often required, resulting in excessive resource consumption. The present application solution selects some of the top-ranked addresses from the full standard address library as candidate addresses for subsequent processing, which greatly reduces the computing resource usage while ensuring the matching accuracy and improves the processing efficiency.

[0103] Step S104: fine-sorting the N candidate addresses based on the BERT semantic vector model to obtain a second address list corresponding to the fine-sorting, wherein the BERT semantic vector model is a semantic vector model that performs geographic location fine-tuning based on geographic location information.

[0104] BERT (Bidirectional Encoder Representations from Transformers, BERT semantic vector model:) is a pre-trained language model that can capture the deep semantic information of text. In step S104, the BERT model is used to generate semantic vectors of non-standard addresses and candidate addresses.

[0105] In a specific method, the BERT semantic vector model can be constructed using the multimodal geographic text pre-training model MGeo open source address analysis kit. Of course, this application does not limit the specific development kit used to construct the BERT semantic vector model, and technical personnel in this field can choose according to actual needs.

[0106] In order to improve the accuracy of address matching, the BERT model adds geographic location information during the training process to perform geographic location fine-tuning, so that the model can better understand the semantics of the address and the relationship between the geographic location.

[0107] In one implementation, during the semantic similarity calculation process, the cosine similarity between the semantic vectors of the non-standard address and the candidate address may be calculated, thereby achieving precise sorting of the candidate addresses.

[0108] It should be noted that the existing technology has a low matching accuracy when standardizing non-standard addresses. The reasons are as follows: (1) Keyword matching limitations: Inaccurate keyword extraction will lead to subsequent matching errors, especially when there are spelling errors, abbreviations, missing information, etc. in the non-standard address. (2) Regular expression matching limitations: Regular expressions have difficulty processing complex address structures, especially non-standard format addresses, which can easily lead to matching failures. (3) Semantic vectorization matching limitations: When natural language processing technology processes complex address structures, its semantic understanding ability is limited to the vectorization model, which can easily lead to matching errors. If you want to tune it, you need to retrain the semantic vectorization model, which is time-consuming, labor-intensive and costly.

[0109] It can be seen that since a single matching and sorting algorithm will always be subject to some limitations, the final standardized result matching accuracy is poor. Therefore, the present application provides a multi-way merge fusion sorting scheme based on a combination of coarse sorting and fine sorting to avoid the limitations of a single algorithm. Specifically, the coarse sorting of the present application is a multi-dimensional address recall rule, including queries based on text feature extraction algorithms, queries based on Boolean queries, and queries based on preset weight strategies, which belong to coarse sorting at the keyword level; then, a semantic-level fine sorting is performed based on a semantic vector model. Obviously, the two-level sorting of coarse sorting and fine sorting can greatly improve the matching accuracy.

[0110] The top 30 recalled data are sorted using the BERT semantic vector model fine-tuned based on geographic location. This step fully considers the semantic similarity of the address and makes up for the lack of semantic understanding in the keyword recall method in the rough sorting stage.

[0111] Step S105: using a preset fusion algorithm to merge and sort the first address list and the second address list to obtain standardized address data after fusion and sorting, and storing the data in the database server.

[0112] In one case, the fused standardized address data can be obtained in the following manner: based on the reciprocal rank fusion algorithm, the first address list and the second address list are fused and sorted to obtain the fused and sorted standardized address data. RRF (Reciprocal Rank Fusion) is a method of combining multiple result sets with different relevance indicators into a single result set. It is a ranking fusion technology used to improve the performance of information retrieval systems. It combines the results of multiple ranking lists to generate a comprehensive ranking list, thereby improving the accuracy and robustness of the overall ranking.

[0113] Furthermore, after the fusion sorting, the standardized address data is obtained, and the first or first M candidate addresses ranked high can be used as the final standardized address or standardized address set, and the determined standardized address or standardized address set is stored in the database server. Preferably, the value of M can be set to a natural number between 2 and 5. It should be noted that this value is not a limitation of this application.

[0114] It is understandable that the address structure of non-standard addresses is often diverse, including spelling errors, abbreviations, mixed languages, etc., and the degree of diversity will become more and more complex. Obviously, a single algorithm is difficult to handle complex situations. The reasons are: (1) Missing address information: There is often missing information in non-standard addresses, such as missing street names, house numbers, etc., which makes matching difficult; (2) Inconsistent address formats: The address formats of different regions and languages ​​vary greatly, and a single algorithm is difficult to adapt to all formats. The BERT semantic vector model provided in this application can accurately sort candidate addresses from a semantic level, and the fusion sorting through the fusion algorithm can effectively deal with the diversity of non-standard addresses, which is conducive to improving matching accuracy.

[0115] This application follows the multi-dimensional address recall rules and only takes the top N recall results returned by the search engine server as candidate data, rather than performing a full match of the standard address library for standard addresses, which greatly reduces the computational overhead and thus improves processing efficiency. Moreover, the multi-way merge fusion sorting algorithm that combines coarse sorting with fine sorting significantly improves the matching accuracy of non-standard address standardization. Experimental data show that the accuracy of non-standard address standardization based on the test set is as high as 86%. In addition, this application can adapt to different types of non-standard addresses, such as spelling errors, abbreviations, missing information and other complex situations, and has strong adaptability and enhanced robustness. In addition, through the process of automated data recall, coarse sorting, fine sorting and fusion sorting, the need for manual intervention is greatly reduced, reducing maintenance costs.

[0116] To further improve the accuracy of address matching, this application provides another non-standard address standardization method, such as Figure 3 As shown, it is another flow chart of the non-standard address standardization method provided by the present application, and the method may include the following steps:

[0117] Step S201: receiving a non-standard address input by a user.

[0118] Step S202: Send an address recall request carrying a non-standard address to the search engine server, so that after receiving the address recall request, the search engine server queries the recall result corresponding to the non-standard address from the database server and sends it to the data processing server.

[0119] Step S203: According to the set multi-dimensional address recall rules, the top N addresses are taken from the recall results returned by the search engine server as candidate addresses to obtain a first address list corresponding to the rough sorting; wherein the multi-dimensional address recall rules include queries based on text feature extraction algorithms, queries based on Boolean queries, and queries based on preset weight strategies.

[0120] Step S204: fine-sorting the N candidate addresses based on the BERT semantic vector model to obtain a second address list corresponding to the fine-sorting; the BERT semantic vector model is a semantic vector model that performs geographic location fine-tuning based on geographic location information.

[0121] Step S205: using a preset fusion algorithm to merge and sort the first address list and the second address list to obtain standardized address data after fusion and sorting.

[0122] It should be noted that Figure 3 Steps S201 to S205 in the method embodiment shown in FIG. Figure 1 Steps S101 to S105 in the method embodiment shown are similar, and the relevant parts can be found in Figure 1 The method implementation example will not be described in detail here.

[0123] Step S206: Based on the preset entity recognition model, the standardized address data after fusion sorting is structured sorted.

[0124] In one case, the standardized address data after fusion sorting can be structured sorted as follows:

[0125] Using an open source address entity recognition model, entity parsing and recognition are performed on the candidate address and the non-standard address to obtain entity recognition results; the entity recognition results at least include: province, city, district, township, village, street, building, house number;

[0126] Calculating the number of common entities between the entity recognition result of each address in the candidate address and the entity recognition result of the non-standard address respectively;

[0127] According to the number of common entities corresponding to each address in the candidate addresses and the proportion of the non-standard address entity recognition results, the standardized address data after the fusion sorting is structured sorted in descending order to obtain a structured candidate address list.

[0128] Step S207: Update the standardized address data using the structured candidate address list.

[0129] It should be noted that Figure 3 The method embodiment shown has Figure 1 All the intended effects of the method embodiments shown, in addition to Figure 3 The method embodiment shown in Figure 1 On the basis of the illustrated method embodiment, the standardized address data after fusion sorting is sorted, and the structural proportion is determined by means of the number of common entities between the entity recognition results of the candidate address and the entity recognition results of the non-standard address, and further matching is performed at the address structure level, thereby improving the matching accuracy.

[0130] Embodiment 3

[0131] See also Figure 4 The present application also provides a non-standard address standardization method, which is applied to a search engine server in a non-standard address standardization system. The non-standard address standardization system also includes a data processing server and a database server. The data processing server is connected to the search engine server via an HTTP / HTTPS protocol, and the data processing server is connected to the database via a database connection protocol.

[0132] Specifically, the method may include the following steps:

[0133] Step S301: receiving an address recall request sent by the data processing server, wherein the address recall request carries a non-standard address input by a user and received by the data processing server.

[0134] Step S302: In response to the address recall request, query the recall result corresponding to the non-standard address from the database server and send it to the data processing server, so that the data processing server: according to the set multi-dimensional address recall rule, take the top N addresses from the recall results returned by the search engine server as candidate addresses, and obtain a first address list corresponding to the rough sorting; wherein the multi-dimensional address recall rule includes query based on text feature extraction algorithm, query based on Boolean query and query based on preset weight strategy; based on the BERT semantic vector model, the N candidate addresses are fine-sorted to obtain a second address list corresponding to the fine sorting; the BERT semantic vector model is a semantic vector model that performs geographic location fine-tuning based on geographic location information; the first address list and the second address list are fused and sorted using a preset fusion algorithm to obtain standardized address data after fusion sorting, and stored in the database server.

[0135] This application follows the multi-dimensional address recall rules and only takes the top N recall results returned by the search engine server as candidate data, rather than performing a full match of the standard address library for standard addresses, which greatly reduces the computational overhead and thus improves processing efficiency. Moreover, the multi-way merge fusion sorting algorithm that combines coarse sorting with fine sorting significantly improves the matching accuracy of non-standard address standardization. Experimental data show that the accuracy of non-standard address standardization based on the test set is as high as 86%. In addition, this application can adapt to different types of non-standard addresses, such as spelling errors, abbreviations, missing information and other complex situations, and has strong adaptability and enhanced robustness. In addition, through the process of automated data recall, coarse sorting, fine sorting and fusion sorting, the need for manual intervention is greatly reduced, reducing maintenance costs.

[0136] It should be noted that, for the method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the order of the actions described, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all optional embodiments, and the actions involved are not necessarily required by the present application.

[0137] The non-standard address standardization device provided by this application is described below.

[0138] Embodiment 4

[0139] See also Figure 5 , which is a structural diagram of the non-standard address standardization device provided in the present application, and is applied to a data processing server in a non-standard address standardization system, wherein the non-standard address standardization system also includes a search engine server and a database server.

[0140] Specifically, the device includes: an address receiving module 410 , a recall request module 420 , a rough sorting module 430 , a fine sorting module 440 and a fusion sorting module 450 .

[0141] The address receiving module 410 is used to receive a non-standard address input by a user.

[0142] The recall request module 420 is used to send an address recall request carrying a non-standard address to the search engine server, so that after receiving the address recall request, the search engine server queries the recall result corresponding to the non-standard address from the database server and sends it to the data processing server.

[0143] The rough sorting module 430 is used to take the top N addresses from the recall results returned by the search engine server as candidate addresses according to the set multi-dimensional address recall rules, and obtain the first address list corresponding to the rough sorting; wherein the multi-dimensional address recall rules include queries based on text feature extraction algorithms, queries based on Boolean queries, and queries based on preset weight strategies.

[0144] The fine sorting module 440 is used to finely sort the N candidate addresses based on the BERT semantic vector model to obtain a second address list corresponding to the fine sorting; the BERT semantic vector model is a semantic vector model that performs geographic location fine-tuning based on geographic location information.

[0145] The fusion sorting module 450 is used to use a preset fusion algorithm to perform fusion sorting on the first address list and the second address list, obtain standardized address data after fusion sorting, and store it in the database server.

[0146] In one case, the rough sorting module 430 is used to calculate the first weight corresponding to the recall results returned by the search engine server using the TF·IDF algorithm, and sort the results in descending order according to the first weight; perform Boolean filtering on the sorted recall results according to the set Boolean query condition; adjust the weight of the recall results after Boolean filtering based on the preset weight strategy to obtain the second weight corresponding to each recall result; sort the results in descending order according to the second weight to obtain the first address list corresponding to the rough sorting.

[0147] In one implementation, the rough sorting module 430 is specifically configured to use the product of the first weight and the weight empirical coefficient as the adjusted second weight.

[0148] In one case, the fine sorting module 440 is used to convert the N candidate addresses and the non-standard address into corresponding semantic vectors respectively; calculate the cosine similarity between the semantic vector of each address in the N candidate addresses and the semantic vector of the candidate address to obtain N similarity values; and sort the N similarity values ​​in descending order to obtain a second address list. Preferably, the value of N is a natural number between 20 and 50.

[0149] In one case, the fusion sorting module 450 is used to perform fusion sorting on the first address list and the second address list based on a reciprocal sorting fusion algorithm to obtain standardized address data after fusion sorting.

[0150] To further improve the matching accuracy, the present application provides another non-standard address standardization device, such as Figure 6 As shown, the device may further include: a structured sorting module 460, which is used to perform structured sorting on the standardized address data after fusion sorting based on a preset entity recognition model; and update the standardized address data using the structured candidate address list.

[0151] This application follows the multi-dimensional address recall rules and only takes the top N recall results returned by the search engine server as candidate data, rather than performing a full match of the standard address library for standard addresses, which greatly reduces the computational overhead and thus improves processing efficiency. Moreover, the multi-way merge fusion sorting algorithm that combines coarse sorting with fine sorting significantly improves the matching accuracy of non-standard address standardization. Experimental data show that the accuracy of non-standard address standardization based on the test set is as high as 86%. In addition, this application can adapt to different types of non-standard addresses, such as spelling errors, abbreviations, missing information and other complex situations, and has strong adaptability and enhanced robustness. In addition, through the process of automated data recall, coarse sorting, fine sorting and fusion sorting, the need for manual intervention is greatly reduced, reducing maintenance costs.

[0152] Embodiment 5

[0153] like Figure 7 As shown, the present application also provides a non-standard address standardization device, which is applied to a search engine server in a non-standard address standardization system. The non-standard address standardization system also includes a data processing server and a database server. The device includes: a recall request receiving module 510 and an address recall module 520.

[0154] The recall request receiving module 510 is used to receive an address recall request sent by the data processing server, wherein the address recall request carries a non-standard address input by a user and received by the data processing server.

[0155] The address recall module 520 is used to respond to the address recall request, query the recall result corresponding to the non-standard address from the database server and send it to the data processing server, so that the data processing server: according to the set multi-dimensional address recall rule, take the top N addresses from the recall results returned by the search engine server as candidate addresses, and obtain a first address list corresponding to the rough sorting; wherein the multi-dimensional address recall rule includes query based on text feature extraction algorithm, query based on Boolean query and query based on preset weight strategy; based on the BERT semantic vector model, the N candidate addresses are fine-sorted to obtain a second address list corresponding to the fine sorting; the BERT semantic vector model is a semantic vector model that performs geographic location fine-tuning based on geographic location information; the first address list and the second address list are fused and sorted using a preset fusion algorithm to obtain standardized address data after fusion sorting, and stored in the database server.

[0156] This application follows the multi-dimensional address recall rules and only takes the top N recall results returned by the search engine server as candidate data, rather than performing a full match of the standard address library for standard addresses, which greatly reduces the computational overhead and thus improves processing efficiency. Moreover, the multi-way merge fusion sorting algorithm that combines coarse sorting with fine sorting significantly improves the matching accuracy of non-standard address standardization. Experimental data show that the accuracy of non-standard address standardization based on the test set is as high as 86%. In addition, this application can adapt to different types of non-standard addresses, such as spelling errors, abbreviations, missing information and other complex situations, and has strong adaptability and enhanced robustness. In addition, through the process of automated data recall, coarse sorting, fine sorting and fusion sorting, the need for manual intervention is greatly reduced, reducing maintenance costs.

[0157] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0158] Embodiment 6

[0159] Optionally, an embodiment of the present application further provides an electronic device, comprising: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the various processes of the above-mentioned method embodiment are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be described here.

[0160] The embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, each process of the above method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it is not repeated here. The computer-readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0161] Figure 8 The present application provides a block diagram of an electronic device 800. For example, the electronic device 800 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.

[0162] Reference Figure 8 , the electronic device 800 may include one or more of the following components: a processing component 802 , a memory 804 , a power component 806 , a multimedia component 808 , an audio component 810 , an input / output (I / O) interface 812 , a sensor component 814 , and a communication component 816 .

[0163] The processing component 802 generally controls the overall operation of the electronic device 800, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps of the above method. In addition, the processing component 802 may include one or more modules to facilitate the interaction between the processing component 802 and other components. For example, the processing component 802 may include a multimedia module to facilitate the interaction between the multimedia component 808 and the processing component 802.

[0164] The memory 804 is configured to store various types of data to support operations on the device 800. Examples of such data include instructions for any application or method operating on the electronic device 800, contact data, phone book data, messages, images, videos, etc. The memory 804 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0165] The power supply component 806 provides power to the various components of the electronic device 800. The power supply component 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the electronic device 800.

[0166] The multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 808 includes a front camera and / or a rear camera. When the device 800 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera may receive external multimedia data. Each front camera and the rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.

[0167] The audio component 810 is configured to output and / or input audio signals. For example, the audio component 810 includes a microphone (MIC), and when the electronic device 800 is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in the memory 804 or sent via the communication component 816. In some embodiments, the audio component 810 also includes a speaker for outputting audio signals.

[0168] I / O interface 812 provides an interface between processing component 802 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include but are not limited to: home button, volume button, start button, and lock button.

[0169] The sensor assembly 814 includes one or more sensors for providing various aspects of status assessment for the electronic device 800. For example, the sensor assembly 814 can detect the open / closed state of the device 800, the relative positioning of components, such as the display and keypad of the electronic device 800, and the sensor assembly 814 can also detect the position change of the electronic device 800 or a component of the electronic device 800, the presence or absence of contact between the user and the electronic device 800, the orientation or acceleration / deceleration of the electronic device 800, and the temperature change of the electronic device 800. The sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 814 may also include an optical sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 may also include an accelerometer, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0170] The communication component 816 is configured to facilitate wired or wireless communication between the electronic device 800 and other devices. The electronic device 800 can access a wireless network based on a communication standard, such as WiFi, a carrier network (such as 2G, 3G, 4G or 5G), or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast operation information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (I rDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0171] In an exemplary embodiment, the electronic device 800 may be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above methods.

[0172] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 804 including instructions, and the instructions can be executed by a processor 820 of an electronic device 800 to perform the above method. For example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0173] Embodiment 7

[0174] Fig. 9 This is a block diagram of another electronic device 1900 provided in the present application. For example, the electronic device 1900 may be provided as a server.

[0175] Reference Fig. 9 , the electronic device 1900 includes a processing component 1922, which further includes one or more processors, and a memory resource represented by a memory 1932 for storing instructions executable by the processing component 1922, such as an application. The application stored in the memory 1932 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 1922 is configured to execute instructions to perform the above method.

[0176] The electronic device 1900 may also include a power supply component 1926 configured to perform power management of the electronic device 1900, a wired or wireless network interface 1950 configured to connect the electronic device 1900 to a network, and an input / output (I / O) interface 1958. The electronic device 1900 may operate based on an operating system stored in the memory 1932, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™ or the like.

[0177] Embodiment 8

[0178] The present application shows a computer program product. When instructions in the computer program product are executed by a processor of an electronic device, the electronic device is enabled to execute the method as described in any one of the above aspects.

[0179] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0180] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0181] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

[0182] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present application can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0183] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0184] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0185] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0186] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard drives, ROM, RAM, magnetic disks, or optical disks.

[0187] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for standardizing non-standard addresses, characterized in that: A data processing server applied to a non-standard address standardization system, wherein the non-standard address standardization system further comprises a search engine server and a database server, wherein the data processing server is connected to the search engine server via an HTTP / HTTPS protocol, and the data processing server is connected to the database via a database connection protocol, wherein the method comprises: Receive non-standard addresses input by users; Sending an address recall request carrying a non-standard address to the search engine server, so that after receiving the address recall request, the search engine server queries the database server for a recall result corresponding to the non-standard address and sends the result to the data processing server; According to the set multi-dimensional address recall rules, the first N addresses with the highest ranking are taken as candidate addresses from the recall results returned by the search engine server, and a first address list corresponding to the rough ranking is obtained; wherein the multi-dimensional address recall rules include query based on text feature extraction algorithm, query based on Boolean and query based on preset weight strategy; Based on the BERT semantic vector model, the N candidate addresses are precisely sorted to obtain a second address list corresponding to the precise sorting; the BERT semantic vector model is a semantic vector model that performs geographic location fine-tuning based on geographic location information; The first address list and the second address list are merged and sorted using a preset fusion algorithm to obtain standardized address data after fusion and sorting, and the data is stored in the database server.

2. The non-standard address standardization method according to claim 1, characterized in that: The step of taking the first N addresses with the highest ranking as candidate addresses from the recall results returned by the search engine server according to the set multi-dimensional address recall rule to obtain a first address list corresponding to the rough ranking includes: Calculating the first weights corresponding to the recall results returned by the search engine server by using the TF·IDF algorithm, and sorting the results in descending order according to the first weights; Perform Boolean filtering on the sorted recall results according to the set Boolean query conditions; Adjust the weights of the recall results after Boolean filtering based on the preset weight strategy to obtain second weights corresponding to each recall result; The second weights are sorted in descending order to obtain a first address list corresponding to the rough sorting.

3. The non-standard address standardization method according to claim 2, characterized in that: The step of adjusting the weights of the recall results after Boolean filtering based on the preset weight strategy to obtain the second weights corresponding to the recall results includes: The product of the first weight and the weight empirical coefficient is used as the adjusted second weight.

4. The non-standard address standardization method according to claim 1, characterized in that: The step of finely sorting the N candidate addresses based on the BERT semantic vector model to obtain a second address list corresponding to the fine sorting includes: Convert the N candidate addresses and the non-standard address into corresponding semantic vectors respectively; Calculating the cosine similarity between the semantic vector of each address in the N candidate addresses and the semantic vector of the candidate address to obtain N similarity values; The N similarity values ​​are sorted in descending order to obtain a second address list.

5. The non-standard address standardization method according to claim 4, characterized in that: The value of N is a natural number between 20 and 50.

6. The non-standard address standardization method according to claim 1, characterized in that: The step of fusing and sorting the first address list and the second address list using a preset fusion algorithm to obtain fused standardized address data includes: Based on the inverse sorting fusion algorithm, the first address list and the second address list are fused and sorted to obtain standardized address data after fusion and sorting.

7. The non-standard address standardization method according to claim 1, characterized in that: Before the step of storing in the database server, the method further comprises: Based on the preset entity recognition model, the standardized address data after fusion sorting is structured sorted; The standardized address data is updated using the structured candidate address list.

8. The non-standard address standardization method according to claim 7, characterized in that: The step of performing structured sorting on the standardized address data after fusion sorting based on the preset entity recognition model includes: Using an open source address entity recognition model, entity parsing and recognition are performed on the candidate address and the non-standard address to obtain entity recognition results; the entity recognition results at least include: province, city, district, township, village, street, building, house number; Calculating the number of common entities between the entity recognition result of each address in the candidate address and the entity recognition result of the non-standard address respectively; According to the number of common entities corresponding to each address in the candidate addresses and the proportion of the non-standard address entity recognition results, the standardized address data after the fusion sorting is structured sorted in descending order to obtain a structured candidate address list.

9. A method for standardizing non-standard addresses, characterized in that: A search engine server applied to a non-standard address standardization system, wherein the non-standard address standardization system further comprises a data processing server and a database server, wherein the data processing server is connected to the search engine server via an HTTP / HTTPS protocol, and the data processing server is connected to the database via a database connection protocol, wherein the method comprises: receiving an address recall request sent by the data processing server, wherein the address recall request carries a non-standard address input by a user and received by the data processing server; In response to the address recall request, the recall result corresponding to the non-standard address is queried from the database server and sent to the data processing server, so that the data processing server: according to the set multi-dimensional address recall rule, the first N addresses with the highest ranking are taken as candidate addresses from the recall results returned by the search engine server, and a first address list corresponding to the rough ranking is obtained; wherein the multi-dimensional address recall rule includes query based on text feature extraction algorithm, query based on Boolean query and query based on preset weight strategy; the N candidate addresses are fine-sorted based on the BERT semantic vector model to obtain a second address list corresponding to the fine-sort; the BERT semantic vector model is a semantic vector model that performs geographic location fine-tuning based on geographic location information; the first address list and the second address list are fused and sorted using a preset fusion algorithm to obtain standardized address data after fusion and sorting, and stored in the database server.

10. A non-standard address standardization device, characterized in that: A data processing server applied to a non-standard address standardization system, wherein the non-standard address standardization system further comprises a search engine server and a database server, and the device comprises: An address receiving module, used to receive a non-standard address input by a user; A recall request module, used to send an address recall request carrying a non-standard address to the search engine server, so that after receiving the address recall request, the search engine server queries the database server for a recall result corresponding to the non-standard address and sends it to the data processing server; A rough sorting module is used to select the top N addresses from the recall results returned by the search engine server as candidate addresses according to the set multi-dimensional address recall rules, and obtain a first address list corresponding to the rough sorting; wherein the multi-dimensional address recall rules include query based on text feature extraction algorithm, query based on Boolean and query based on preset weight strategy; A fine sorting module, used for fine sorting the N candidate addresses based on a BERT semantic vector model to obtain a second address list corresponding to the fine sorting; the BERT semantic vector model is a semantic vector model that performs geographic location fine-tuning based on geographic location information; The fusion sorting module is used to use a preset fusion algorithm to perform fusion sorting on the first address list and the second address list, obtain standardized address data after fusion sorting, and store it in the database server.

11. The non-standard address standardization device according to claim 10, characterized in that: The device also includes: The structured sorting module is used to perform structured sorting on the standardized address data after fusion sorting based on a preset entity recognition model; and update the standardized address data using the structured candidate address list.

12. A non-standard address standardization device, characterized in that: A search engine server applied to a non-standard address standardization system, wherein the non-standard address standardization system further comprises a data processing server and a database server, and the device comprises: A recall request receiving module, configured to receive an address recall request sent by the data processing server, wherein the address recall request carries a non-standard address input by a user and received by the data processing server; The address recall module is used to respond to the address recall request, query the recall result corresponding to the non-standard address from the database server and send it to the data processing server, so that the data processing server: according to the set multi-dimensional address recall rule, take the top N addresses from the recall result returned by the search engine server as candidate addresses, and obtain a first address list corresponding to the rough sorting; wherein the multi-dimensional address recall rule includes query based on text feature extraction algorithm, query based on Boolean query and query based on preset weight strategy; based on the BERT semantic vector model, the N candidate addresses are fine-sorted to obtain a second address list corresponding to the fine sorting; the BERT semantic vector model is a semantic vector model that fine-tunes the geographic location based on geographic location information; the first address list and the second address list are fused and sorted using a preset fusion algorithm to obtain standardized address data after fusion sorting, and stored in the database server.

13. An electronic device, characterized in that: include: A processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program implements the method according to any one of claims 1 to 9 when executed by the processor.

14. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 9 is implemented.

15. A computer program product, characterized in that When the instructions in the computer program product are executed by a processor of an electronic device, the electronic device implements the method according to any one of claims 1 to 9.

Images