Address standardization and space association method based on intelligent semantic analysis

Abstract

The invention discloses an address standardization and space association method based on intelligent semantic analysis, and relates to the technical field of address information processing, and the method comprises the steps: constructing a standard address library comprising space level information and geographic space geometric information according to an administrative space level data set and a house landmark space data set; performing hierarchical analysis on the to-be-processed text address based on the standard address library to obtain target space hierarchical data; and matching a spatial geometric position corresponding to the spatial hierarchy information in the standard address base, and mapping the spatial geometric position to a visual spatial geometric map. According to the method, automatic and batch efficient processing and accurate space matching of the text addresses are realized through standard address library construction, address hierarchical analysis and space geometry matching.

Description

Technical Field

[0001] This application relates to the field of address information processing technology, and in particular to address standardization and spatial association methods based on intelligent semantic parsing. Background Technology

[0002] With the rapid development of emerging applications such as digital twin cities, smart city management, and intelligent logistics, converting text addresses described in natural language into standardized data and accurately associating them with standardized spatial data (such as building boundaries, coordinate points, and administrative divisions in geographic information systems) has become a key technological requirement supporting upper-level applications. This technology aims to achieve standardized storage and structured management of address information, and to complete the precise mapping from text descriptions to specific spatial locations (such as buildings, units, and house numbers). It has broad application value and urgent demand in areas such as logistics distribution, smart city governance, optimization of public service facility layout, and intelligent navigation.

[0003] However, with the continuous advancement of urbanization, urban building density is constantly increasing, duplicate place names are common, and address expression formats are becoming increasingly diversified. Traditional address matching technology faces severe challenges and is unable to meet the needs of modern application scenarios for high-precision and high-efficiency spatial positioning at the building-unit-door number level. Specifically, existing technologies face three main technical bottlenecks: 1) Strong address ambiguity: the same text address may correspond to multiple physical locations (e.g., "No. 88, xx Road, xx District" may point to buildings in different business districts or functional areas), resulting in non-unique positioning results; 2) Prominent data heterogeneity: user-inputted text addresses are often presented in free text format (containing abbreviations, typos, colloquial expressions, and other non-standard information), which differs significantly from the standardized field structure (such as hierarchical geocoding) in standardized spatial data, hindering accurate information mapping; 3) High computational complexity: achieving accurate building-level matching requires comprehensive analysis of multi-source heterogeneous data (including administrative boundaries, road networks, point of interest (POI) distribution, and building physical boundaries). Traditional methods are computationally inefficient during data fusion and indexing, making it difficult to meet the real-time processing needs of large-scale address data. These technical challenges severely restrict the in-depth application and effectiveness of address matching technology in key areas such as smart cities. Summary of the Invention

[0004] In view of this, embodiments of this application provide a method and related equipment for address standardization and spatial association based on intelligent semantic parsing, so as to accurately parse and associate text addresses.

[0005] One aspect of this application provides an address standardization and spatial association method based on intelligent semantic parsing, the method comprising the following steps:

[0006] A standard address database, including spatial hierarchy information and geospatial geometric information, was constructed based on the administrative spatial hierarchy dataset and the building landmark spatial dataset.

[0007] Based on the standard address library, the text address to be processed is parsed hierarchically to obtain the target space hierarchy data;

[0008] The spatial geometric location corresponding to the spatial hierarchy information is matched in the standard address database, and the spatial geometric location is mapped to a visualized spatial geometric map.

[0009] In some embodiments, constructing a standard address database including spatial hierarchy information and geospatial geometric information based on an administrative spatial hierarchy dataset and a building landmark spatial dataset includes the following steps:

[0010] The process involves importing, spatially aligning, cleaning, and fusing raw address data from multiple sources to generate an intermediate dataset.

[0011] The expression for the intermediate dataset is:

[0012] ;

[0013] in, For the intermediate dataset, , These are datasets of administrative spatial levels. and housing landmark spatial dataset The cleaning function, This represents a data integration function. The integration strategy is to map administrative levels step by step. Indicates the associated field;

[0014] Each address record in the intermediate dataset is subjected to address standardization to obtain a standard address dataset; wherein, the address standardization process includes address text parsing and structuring, place name unification, administrative code mapping, spatial geographic coordinate standardization, and redundancy removal and merging;

[0015] The expression for address normalization is:

[0016] ;

[0017] in, For the standard address dataset, For the standardized function, A standardized set of rules;

[0018] The standard address dataset is logically split into layers according to spatial hierarchy and structured and stored in different data tables of a preset database, thereby constructing the standard address library that includes spatial hierarchy information and geospatial geometric information.

[0019] In some embodiments, the step of performing hierarchical parsing of the text address to be processed based on the standard address library to obtain target spatial hierarchy data includes the following steps:

[0020] The text address to be processed is preprocessed; the preprocessing methods include case conversion, Chinese character and number conversion, and place name alias conversion.

[0021] The administrative spatial hierarchy data in the text address to be processed was extracted by NLP word segmentation and recognition.

[0022] By utilizing the correlation between adjacent administrative space level data, semantic errors in the text address to be processed are corrected, and missing administrative space level data is supplemented, thereby restoring the text address to be processed into a complete address;

[0023] Using the TextRCNN deep neural network model and combining the spatial hierarchy information stored in the standard address library, the restored text address to be processed is classified by text hierarchy to obtain the target spatial hierarchy data.

[0024] In some embodiments, the TextRCNN deep neural network model performs the following steps on the restored text address to be processed:

[0025] The contextual semantic information of each word in the restored text address to be processed is obtained by a bidirectional recurrent neural network, and the output of the recurrent neural network is concatenated with the embedding vector of the word to form a target semantic representation.

[0026] word The left-hand context vector is calculated as follows:

[0027] ;

[0028] in, It has A dense vector of real values; It is a word Word embedding is a feature A dense vector of real values; It is the previous word The left-side context; the shared left-side context parameter of the first word in the restored text address to be processed. ; It involves transforming a hidden layer into the next hidden layer; This is a weight matrix used to combine the semantics of the current word with the upper left text of the next word; It is a non-linear activation function;

[0029] word The right-hand context vector is calculated as follows:

[0030] ;

[0031] Among them, the right-hand context shared parameter of the last word in the restored text address to be processed ;

[0032] left context vector Word embedding and the right-hand context vector By combining the words, we can obtain the word. The representation is as follows:

[0033] ;

[0034] After linear transformation and tanh activation function processing, the words are obtained. The final semantic representation is as follows:

[0035] ;

[0036] and words The final semantic representation The input is fed into subsequent layers of the bidirectional recurrent neural network for processing;

[0037] in, These are latent semantic vectors used to analyze each semantic factor to determine the most useful factors for representing the text. For the corresponding bias term;

[0038] The concatenated multi-dimensional word vectors are compressed into a low-dimensional semantic space through non-linear mapping, and key discriminative information is extracted from the restored text address using max pooling. Specifically, this includes performing max pooling on all word representations in the same dimension.

[0039] ;

[0040] in, This represents the j-th feature value of the i-th word;

[0041] The output layer is defined as follows, and is used to map the pooled features to the classification space:

[0042] ;

[0043] in, This is the output layer weight matrix. For bias terms, The feature vector after pooling;

[0044] The extracted high-dimensional features are classified using a fully connected softmax layer, and the probability prediction value for each spatial level category is calculated:

[0045] ;

[0046] in, This indicates that the restored address of the text to be processed belongs to the first... The probability of each spatial hierarchy category, This represents the total number of categories.

[0047] In some embodiments, matching the spatial geometric location corresponding to the spatial hierarchy information in the standard address database includes the following steps:

[0048] The spatial hierarchy information is matched one-to-one with the building landmark information in the standard address database, thereby realizing the conversion of text address to spatial data and obtaining the spatial geometric location of the spatial hierarchy information.

[0049] In some embodiments, mapping the spatial geometric location to a visualized spatial geometric map includes the following steps:

[0050] The spatial geometric location is mapped to a visualized spatial geometric map using a professional geographic information system engine; wherein the spatial geometric location includes coordinates, boundaries, and hierarchical metadata.

[0051] In some embodiments, the method further includes the following steps:

[0052] The spatial hierarchical text information and spatial geographic geometric data corresponding to the address of the spatial geometric location are stored in the standard address database.

[0053] Another aspect of this application embodiment provides an address standardization and spatial association device based on intelligent semantic parsing, the device comprising:

[0054] The standard address database construction unit is used to construct a standard address database that includes spatial hierarchy information and geospatial geometric information based on the administrative spatial hierarchy dataset and the building landmark spatial dataset.

[0055] The text address hierarchical unit is used to perform hierarchical parsing of the text address to be processed based on the standard address library to obtain target spatial hierarchy data.

[0056] The spatial matching and mapping unit is used to match the spatial geometric location corresponding to the spatial hierarchy information in the standard address database, and map the spatial geometric location to a visualized spatial geometric map.

[0057] Another aspect of this application embodiment provides an electronic device, including a processor and a memory;

[0058] The memory is used to store programs;

[0059] The processor executes the program to implement any of the methods described above.

[0060] Another aspect of this application provides a computer-readable storage medium storing a program that is executed by a processor to implement the method described in any of the above embodiments.

[0061] This application includes at least the following beneficial effects:

[0062] This application constructs a standard address database that includes spatial hierarchy information and geospatial geometric information based on administrative spatial hierarchy datasets and building landmark spatial datasets. It then performs hierarchical parsing of the text addresses to be processed based on the standard address database to obtain the target spatial hierarchy data. Finally, it matches the spatial geometric locations corresponding to the spatial hierarchy information in the standard address database and maps these locations onto a visualized spatial geometric map. This application achieves automated, batch, and efficient processing of text addresses, as well as accurate spatial matching, through the construction of a standard address database, hierarchical address parsing, and spatial geometric matching. Attached Figure Description

[0063] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0064] Figure 1 A flowchart illustrating the address standardization and spatial association method based on intelligent semantic parsing provided in this application embodiment;

[0065] Figure 2 Example flowchart of the address standardization and spatial association method based on intelligent semantic parsing provided in the embodiments of this application;

[0066] Figure 3 A structural diagram of the TextRCNN deep neural network model provided in the embodiments of this application;

[0067] Figure 4A schematic diagram illustrating the average accuracy of different data entities under different similarity thresholds provided in the embodiments of this application;

[0068] Figure 5 The structural block diagram of the address standardization and spatial association device based on intelligent semantic parsing provided in the embodiments of this application is shown. Detailed Implementation

[0069] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0070] Before providing a detailed description of the embodiments of this application, some related technologies involved in the embodiments of this application will be described first, as follows:

[0071] Spatial address matching techniques can be mainly divided into three categories: 1) Rule-based address parsing techniques. This method involves segmenting text addresses using a predefined rule base, extracting fields such as province, city, district, street, and neighborhood, and then matching them step-by-step with a standardized spatial database. 2) Machine learning-based address classification techniques. This transforms the address matching problem into a classification task, using machine learning models (such as decision trees, random forests, and deep learning) to learn the mapping relationship between historical address data and spatial location, predicting the building category to which the text address belongs after input. 3) Geocoding-based approximate matching techniques. This converts text addresses into geographic coordinates, and then uses spatial indexes (such as R-trees and quadtrees) to retrieve the nearest building boundary or center point in a GIS database, achieving a "coordinate-building" association. However, all three techniques have significant limitations: rule-based matching methods have poor ambiguity handling capabilities and low accuracy; machine learning-based address classification methods require large-scale labeled data, have poor model interpretability, and are difficult to debug and optimize; and geocoding-based matching methods have high computational complexity.

[0072] Regarding the address resolution and spatial association of text addresses, the aforementioned technologies still suffer from the following problems: 1. Spatial positioning accuracy is generally insufficient, making it difficult to effectively achieve refined spatial positioning at the building level (or even unit and door number), failing to meet the needs of high-precision application scenarios. 2. Strict requirements are placed on the quality of the original text data. When faced with free text formats such as abbreviations, typos, and colloquial expressions commonly found in real-world scenarios, traditional methods have limited processing capabilities, leading to a significant decrease in the accuracy of spatial association matching, thus restricting the practicality and reliability of the technology.

[0073] This application aims to address the core problems of weak text address data governance capabilities and insufficient spatial positioning accuracy in existing technologies. It proposes an innovative, lightweight address parsing and spatial association matching framework based on natural language processing. By constructing a standard address database, hierarchical address parsing, and spatial geometric matching, it achieves automated, batch, and efficient processing of text address data and accurate spatial matching, ultimately outputting standardized results and validating their effectiveness.

[0074] Specifically, this application effectively solves the governance problem of non-standard address data in real-world scenarios by mining the spatial hierarchical relationships in the standard address database, intelligently identifying and correcting missing information and format errors in the original text. At the same time, it integrates an address parsing engine and a spatial geometry matching engine to accurately extract spatial hierarchical information (such as province / city / county / township) from the address text and map it to the corresponding spatial geometric entities, ultimately forming a processing result that balances standardization and spatial accuracy. This achieves a technological breakthrough in building-level high-precision spatial positioning and significantly improves the accuracy, robustness, and practicality of address matching.

[0075] Reference Figure 1 This application provides an address standardization and spatial association method based on intelligent semantic parsing, specifically including the following steps S100~S120:

[0076] S100: Construct a standard address database that includes spatial hierarchy information and geospatial geometric information based on the administrative spatial hierarchy dataset and the building landmark spatial dataset;

[0077] S110: Based on the standard address library, perform hierarchical parsing of the text address to be processed to obtain target spatial hierarchy data;

[0078] S120: Match the spatial geometric location corresponding to the spatial hierarchy information in the standard address database, and map the spatial geometric location to a visualized spatial geometric map.

[0079] The following section will provide a detailed introduction and explanation of the solutions in the embodiments of this application, using specific application examples.

[0080] Given the challenges of diverse text addresses and low building-level spatial positioning accuracy in existing technologies, the purpose of this application is to provide a building-unit-address number-level spatial parsing and matching algorithm based on natural language processing. The system first constructs a standardized and structured standard address database. This is achieved by associating spatial geographic data with administrative spatial hierarchy data and dividing the data according to spatial hierarchy. Each level of data table is stored in a PostgreSQL database to obtain the standard address database. Based on this standard address database, an automated batch processing method for address standardization hierarchical and spatial matching based on natural language processing technology is proposed. Finally, the effectiveness of the algorithm is verified by calculating its text similarity and accuracy, ensuring reliable results.

[0081] Given the challenges of diverse text address representations and suboptimal building-level spatial positioning accuracy in existing technologies, the purpose of this application is to provide a building-unit-address number-level spatial parsing and matching algorithm based on natural language processing, achieving accurate parsing and matching of text addresses at a fine spatial level. This technical solution first constructs a standard address database containing spatial hierarchy information and geospatial geometric information. The data structure is divided according to spatial hierarchy, and each level of data table is stored in a PostgreSQL database, forming a structured address baseline database. Based on this database, natural language processing techniques are used to sequentially perform word segmentation, correction and completion, matching, and hierarchical parsing operations on the text address, transforming unstructured text into structured features that match the standard database. Subsequently, the parsed spatial data at each level is associated with the standard address database, and a GIS engine is used to achieve visual mapping of the spatial data, accurately anchoring the physical spatial location corresponding to the text address. The standardized results after parsing and matching are then centrally stored. Finally, by calculating text similarity and accuracy indicators, the validity of the stored results is verified from both text consistency and spatial accuracy dimensions, ensuring the reliability of the output results. The entire process, from address database construction to result verification, forms a complete technical chain. It not only solves the problem of parsing diverse text addresses, but also improves the accuracy of building-level spatial positioning by leveraging spatial geometric matching and visualization, ultimately achieving the technical goal of building-unit-door number level spatial parsing and matching.

[0082] Reference Figure 2 This application provides a building-level spatial parsing and matching method based on natural language processing. The method includes standard address database construction, address hierarchical parsing, spatial geometric matching, standardized matching result storage, and validity verification.

[0083] Specifically, the embodiments of this application include the following steps:

[0084] Step 1: Construction of the standard address database. First, data preprocessing is performed on administrative spatial hierarchy data and building landmark spatial data, including data cleaning, data format unification, and data element encoding. These are then integrated to construct a standard address database, which includes spatial hierarchy information and geospatial geometric information.

[0085] Step 2: Address hierarchical parsing. Based on the standard address library built in Step 1, address hierarchical parsing is performed by using natural language processing methods to achieve address segmentation, address correction and completion, and address matching of text address information.

[0086] Step 3: Spatial geometric matching. By accurately associating the spatial hierarchical data obtained in Step 2 with the standard address database, the text address is converted into standard geometric data. The geometric data is then visualized and mapped using a GIS engine to complete spatial positioning and distribution analysis.

[0087] Step 4: Standardize the matching results storage, storing the text information of each spatial level in Step 2 and the spatial geographic geometry data in Step 3 in the database.

[0088] Step 5: Validation. The validity of the algorithm is verified by calculating text similarity and accuracy.

[0089] Furthermore, step 1 specifically includes the following steps:

[0090] Based on existing business addresses, operations such as cleaning, standardization, and database construction are performed to generate a high-quality standard address database.

[0091] 1a. Import, spatially align, clean, and fuse multi-source heterogeneous raw address data (such as house addresses, landmarks, and administrative division information from business systems) to generate an intermediate dataset. This serves as the input for subsequent standardization. The administrative spatial hierarchy dataset ( ) and housing landmark spatial dataset ( Data cleaning and integration are performed, and the formula is expressed as follows:

[0092] ;

[0093] in and These are datasets of administrative spatial levels. and housing landmark spatial dataset The cleaning function, Represented as a data integration function, The integration strategy is to map administrative levels step by step. This is represented as a related field.

[0094] 1b. For the intermediate dataset obtained in step 1a Address standardization processes are performed, including address structure decomposition, hierarchical normalization, place name unification, and coding standardization, ultimately generating a standard address dataset with a clear structure, standardized fields, and clear semantics. For each address record di∈ The following standardization operations are performed: (1) Address text parsing and structuring: Addresses are parsed into standard fields according to spatial hierarchy of province, city, county, township, street, community / administrative village, residential area / natural village, building, unit, and house number; (2) Place name unification: Place names with different spellings, aliases, and abbreviations are mapped to official standard names; (3) Administrative code mapping: Standard administrative codes are matched for each spatial level address information; (4) Spatial geographic coordinate standardization: Coordinate system, precision, and format are unified; (5) Redundancy removal and merging: Duplicate or highly similar addresses are merged or deduplicated. The formula is expressed as follows:

[0095] ;

[0096] in, For the standardized function, It is a set of standardized rules.

[0097] 1c. The standardized address dataset (Structured and standardized, with unified fields, including spatial hierarchy information such as province, city, district, street, etc.) Logically divided into layers according to spatial hierarchy dimensions (such as province, city, county, township, street, alley, community / administrative village, residential area / natural village group, building, unit, house number), and structured and stored in different data tables of PostgreSQL database. Each table stores address data at a specific level, and the following requirements are met: (1) Each table has a unified field structure; (2) The tables are linked through key fields to support inter-table association queries (such as querying the district / county from the street, querying from the residential area, etc.). Finally, a standardized address database that is standardized, hierarchical, scalable, and efficient in querying is formed.

[0098] Step 2 specifically includes the following steps:

[0099] 2a. Address preprocessing: Preprocess the input address information, including operations such as case conversion, Chinese character and number conversion, and place name alias conversion.

[0100] 2b. Address segmentation: Extract administrative spatial level data from the input address information through NLP segmentation and recognition, including province, city, county, township, street / lane, community / administrative village, and residential area / natural village.

[0101] 2c. Address Correction and Completion: By leveraging the relationships between adjacent spatial hierarchy information, the algorithm can adaptively correct semantic errors in addresses and complete missing spatial hierarchy information, thereby restoring the address to a complete one as much as possible while meeting the standards for standard address information. For example, when a text address lacks the "township" hierarchy information, the algorithm can infer and complete the missing hierarchy information through correlation analysis of adjacent hierarchy information and geographic location data, thus ensuring the completeness and consistency of the address resolution results.

[0102] 2d. Address Matching: This application embodiment utilizes contextual language logic inference technology, combined with spatial hierarchy data relationships stored in the standard address database, to achieve full and accurate matching of information at each spatial hierarchy level. Addressing the diversity of expression forms and structural complexity in the underlying spatial hierarchy texts such as building, unit, and address numbers, this application embodiment innovatively introduces the TextRCNN deep neural network model for text hierarchy classification, effectively improving the semantic understanding and hierarchy division capabilities of complex address texts.

[0103] like Figure 3 As shown, the TextRCNN deep neural network model takes a text sequence D as input, which is represented as words. The model outputs ordered sets of elements, containing several categories to represent classification results at different spatial levels. The core of this network architecture lies in fusing the long-range dependency capture capability of recurrent neural networks (RNNs) with the key feature filtering mechanism of convolutional neural networks (CNNs), enabling it to comprehensively understand textual context information without being limited by a fixed window size, thus achieving accurate identification and classification of information at various spatial levels within address text. The specific processing flow of the TextRCNN model is as follows:

[0104] Word representation learning layer: First, the contextual semantic information of each word is obtained through a bidirectional recurrent neural network (Bi-RNN). Then, the RNN output is concatenated with the original word embedding vector to form a richer semantic representation.

[0105] word The left-side context vector is calculated as shown in the following formula. It has A dense vector of real values.

[0106] ;

[0107] in, It is a word Word embedding is a feature A dense vector of real values. It is the previous word The left-hand context. The left-hand context of the first word in the text shares parameters. . It transforms the hidden layer to the next hidden layer. This is a weight matrix used to combine the semantics of the current word with the upper left text of the next word. It is a non-linear activation function.

[0108] The right-side context vector is calculated as shown in the formula below, capturing contextual information to the right of a word through a similar mechanism. The right-hand context shared parameter of the last word in the text sequence .

[0109] ;

[0110] Then, using the formula, the left-side context vector... Word embedding and the right-hand context vector By combining the words, we can obtain the word. The representation of:

[0111] ;

[0112] As shown in the following formula, this representation is further processed by a linear transformation and the tanh activation function to obtain the final semantic representation of the word, which is then input into subsequent layers of the network for further processing:

[0113] ;

[0114] in, It is a potential semantic vector that analyzes each semantic factor to determine the most useful factor for representing the text. For the corresponding bias term.

[0115] Text representation learning layer: This layer compresses the concatenated multi-dimensional word vectors into a low-dimensional semantic space using a non-linear mapping, and then employs max pooling to extract key discriminative information from the text. As shown in the following formula, max pooling is performed on all word representations along the same dimension:

[0116] ;

[0117] in, Let represent the j-th feature value of the i-th word.

[0118] The output layer is defined as shown in the following formula, which maps the pooled features to the classification space:

[0119] ;

[0120] in, This is the output layer weight matrix. For bias terms, This is the feature vector after pooling.

[0121] Classification Decision Layer: As shown in the final formula below, the extracted high-dimensional features are classified through a fully connected softmax layer, and the probability prediction value of each spatial level category is calculated:

[0122] ;

[0123] in, Indicates that the text belongs to the first The probability of each spatial hierarchy category, This represents the total number of categories.

[0124] Step 3 specifically includes the following steps:

[0125] Based on key information in the text address, such as house number, community name, or keywords, locate the corresponding standard address in the standard address database, and attach the corresponding attribute information of the standard address, such as coordinates and tags.

[0126] 3a. Spatial geometric matching: Match the spatial level matching data obtained in step 2 with the building landmark information in the standard address database to realize the conversion of text address to spatial data and obtain its specific standard geospatial geometric data.

[0127] 3b. Spatial geometric map visualization mapping: Relying on a professional geographic information system engine, the standardized geospatial geometric data (including coordinates, boundaries, and hierarchical metadata) output in step 3a is transformed into an intuitive visualization, enabling spatial positioning and distribution pattern analysis of the matching results.

[0128] Step 4 specifically includes the following steps:

[0129] 4a. Obtain results that meet the standardized address requirements: address layering results according to the spatial hierarchy of province, city, county, township, street / lane, community / administrative village, neighborhood / natural village group, building, unit, and house number, as well as the corresponding address coding data and spatial geographic geometric data.

[0130] 4b. Ultimately, two standardized data tables will be obtained in the database, storing structured spatial layering information and spatial geographic geometric data respectively.

[0131] For example, Table 1 shows the core table structure of the address standardization processing result library.

[0132] Table 1

[0133] Table name Field Description standardized_data_YYYYMMDD lng(float8),lat(float8),energyAddress(vachar),province_id(vachar),provance_name(vachar),city_id(vachar),city_name(vachar)... standardized_gemo_YYYYMMDD eneryid(vachar),town_gemo(vachar),road_gemo(vachar),village_gemo(vachar),community_gemo(vacher)

[0134] The embodiments of this application can automatically and batch - analyze and process text address information, achieve efficient conversion from text to structured data, and provide accurate basic data for spatial geometric matching to achieve building - level spatial positioning.

[0135] Step 5 specifically includes the following steps:

[0136] To objectively evaluate the processing effect and classification accuracy of the address standardization classification algorithm in the embodiments of this application, this step proposes a classification effect verification method based on quantitative analysis of text similarity. By calculating the semantic similarity index between the original address text and the reconstructed address text after standardization classification, quantitative evaluation of the algorithm's classification effect and accuracy analysis are achieved, providing a scientific basis for algorithm performance optimization and threshold parameter determination.

[0137] 5a. Design of verification method and evaluation index:

[0138] Based on text address information as the verification basis, a typical original address sample X in a certain place is selected: 'Lai* Community and x* Court, Building b, Unit 2, No. 311, D Town, C County, B City, A Province', as the original address benchmark data without algorithm processing. Through the algorithm in the embodiments of this application, standardization classification processing is performed on the original address X, and the following structural decomposition results can be obtained:

[0139] province_name = ’A Province’ / / Province name

[0140] city_name = ’B City’ / / City name

[0141] county_name = ’C County’ / / County name

[0142] town_name = 'D Town' / / Township - level administrative division name

[0143] village_name = 'Lai* Community' / / Community / administrative village name

[0144] road_name = 'Lai* Road' / / Street, road or lane name

[0145] communal_name = 'and x* Court' / / Community / natural village / group name

[0146] building_name = 'Building b' / / Building name

[0147] unit_name = 'Unit 2' / / Unit name

[0148] house_number = '311' / / House number

[0149] other_address = '' / / Unmatched address string fragment

[0150] Based on the administrative hierarchy standard (province-city-county-township-street-community / administrative village-residential area / natural village group-building-unit-door number-(unmatched string), the above structured parsing results are reassembled to generate a standardized reconstructed address Y: 'A province B city C county D town Lai* community Lai* road and ** courtyard b building 2 unit 311' (* represents the corresponding Chinese characters).

[0151] 5b. Text similarity calculation model:

[0152] To quantify the semantic similarity between the original address X and the reconstructed address Y, embodiments of this application employ a text similarity mapping function for calculation:

[0153] ;

[0154] in, This is a mapping function for calculating string similarity, where N is the total number of samples. Average text similarity ( , indicating complete dissimilarity , indicating complete similarity.

[0155] 5c. Classification accuracy calculation model:

[0156] Based on the text similarity calculation results, this application embodiment sets a similarity threshold determination mechanism to determine the classification correctness, and then calculates the overall classification accuracy of the algorithm. The formula is as follows:

[0157] ;

[0158] in, The average accuracy rate. A mapping function for determining the correctness of a string, if , The average similarity is greater than or equal to the threshold similarity. The value is 1 if it is not 1, otherwise it is 0.

[0159] 5d. Empirical verification and parameter optimization:

[0160] To comprehensively verify the effectiveness of the algorithm, this application's embodiments selected three typical address datasets—actual population, actual units, and actual buildings—for comprehensive evaluation. The actual population dataset contained 191,115 data entries, the actual unit dataset contained 5,862 data entries, and the actual building dataset contained 85,006 data entries. The average similarity index was calculated for each dataset, and the evaluation results are shown in Table 2.

[0161] Table 2. Average similarity of different data entities

[0162] Dataset Average similarity Actual population 0.9499 Actual units 0.8976 Existing houses 0.9359

[0163] Analysis of the data in Table 2 shows that, assuming the standard address database information is complete, the algorithm in this application maintains an average similarity of around 0.9 for the three main address data types, demonstrating its excellent performance in address text segmentation and classification. Furthermore, this application set multiple similarity thresholds (0.75, 0.80, 0.85, 0.89, 0.93, 0.94) to evaluate the classification accuracy under different threshold conditions. The results are shown in Table 3. Figure 4 It visually demonstrates the dynamic trend of accuracy as a function of the similarity threshold.

[0164] Table 3. Average accuracy of different data entities under different similarity thresholds.

[0165]

[0166] Combination Figure 4 The accuracy trend shown is clearly observable: when the similarity threshold is set to 0.80, the classification accuracy of the three types of data entities reaches a relatively balanced and optimal state—the accuracy for actual population is 0.9647%, for actual units it is 0.9051%, and for actual houses it is 0.9535%. Under this threshold condition, the algorithm maintains good recall and practicality while ensuring high classification accuracy. Therefore, this application embodiment determines 0.80 as the optimal similarity threshold parameter for address matching judgment. In practical applications, when the calculated text similarity is greater than or equal to 0.80, the algorithm classification result is judged to be correct; otherwise, it is judged as a classification anomaly, requiring manual review or algorithm optimization. This threshold parameter provides a clear decision-making basis and quality control standard for the engineering application of the algorithm in this application embodiment.

[0167] In summary, the embodiments of this application include the following innovative points:

[0168] (1) The embodiments of this application propose a refined matching method for spatial levels such as building-unit-door number, which breaks through the traditional macro-level limitation, realizes accurate analysis and association of the bottom spatial units, and significantly improves the address positioning accuracy.

[0169] (2) The NLP-based address full-level standardization parsing and spatial association method innovatively uses natural language processing technology to deeply analyze the semantic structure of text address, realize the standardization decomposition and mapping of full-level information from country / province / city / county / township to building / unit / door number, and accurately associate with standard address database building landmark data to complete the automatic conversion of unstructured text address to structured geospatial entity.

[0170] (3) Lightweight and efficient automated batch processing architecture supports unattended automated processing of large-scale address data in a limited environment, taking into account processing efficiency, scalability and practicality, solving the problem of large-scale application of massive data processing, and forming a full-process automated technology system.

[0171] This application addresses the technical bottlenecks in address text processing and spatial matching, focusing on the fully automated processing from unstructured text addresses to standardized geospatial entities, thus forming a core solution with significant technological innovation and practical value. Specific key technologies are as follows:

[0172] 1. A high-precision spatial matching method at the building-unit-door number level.

[0173] This application proposes a refined spatial matching method for micro-level spatial hierarchy, which breaks through the limitations of traditional address matching technology that is limited to high-level administrative divisions (such as provinces / cities / counties). It realizes high-precision spatial association and matching of low-level spatial units such as buildings, units, and house numbers, significantly improving the accuracy and practicality of address spatial positioning, and meeting the needs of refined management scenarios (such as community governance, facility operation and maintenance, and precision delivery) for fine-grained spatial information.

[0174] 2. Address standardization hierarchical resolution and spatial association method based on NLP technology.

[0175] This application innovatively proposes a hierarchical parsing and spatial matching method for address standardization based on Natural Language Processing (NLP) technology. By deeply parsing the semantic structure of unstructured text addresses, it achieves hierarchical decomposition of address information (covering all levels such as province / city / county / township) and standardized mapping. Based on building landmark data in a standard address database, it completes the automated and accurate conversion from text addresses described in natural language to structured, hierarchical, and standardized geospatial entities. This method solves the technical problems of traditional address processing, such as reliance on human experience, low standardization, and inaccurate spatial association.

[0176] 3. Lightweight and highly efficient automated batch processing architecture.

[0177] This application presents and implements a lightweight, high-efficiency automated batch processing algorithm architecture. By optimizing the allocation of computing resources and the processing flow, it maintains high efficiency even in resource-constrained environments and supports unattended automated batch processing and spatial matching of large-scale address datasets. This architecture significantly improves the processing efficiency, system scalability, and practical applicability of address standardization and spatial matching, solving the problems of high computational cost, low efficiency, and difficulty in scaling up traditional methods when processing massive amounts of data.

[0178] Compared with the three mainstream spatial address matching technologies currently available, the embodiments of this application have significant technical advantages and improved practicality. Traditional rule-based matching methods rely on predefined rule bases for address segmentation and parsing. Although the logic is clear, the ability to handle ambiguity is poor, making it difficult to adapt to the diversity and complexity of address representations, resulting in low spatial positioning accuracy, especially in finer-grained spatial hierarchical matching such as building-unit-door number. Although machine learning-based classification technology attempts to learn the mapping relationship between address and space through models, it relies on large-scale labeled data, has poor interpretability, and is difficult to optimize specifically, thus limiting its generalization ability. Geocoding-based approximate matching technology achieves association through coordinate transformation and spatial index retrieval. Although it can obtain coarse-grained location, the computational complexity is high, and the positioning accuracy is limited by the index granularity, making it impossible to accurately match to specific building units.

[0179] This application's embodiments utilize NLP-driven address standardization hierarchical parsing technology to deeply analyze the semantic structure of text addresses, achieving accurate decomposition and standardized mapping across all levels, from country / province / city / county to building / unit / door number. Combined with building landmark data from a standard address database, it completes precise association, significantly improving spatial positioning accuracy at the micro-level. Simultaneously, it integrates adaptive text correction and completion mechanisms to effectively handle non-standard text such as abbreviations, typos, and colloquial expressions in real-world scenarios, reducing reliance on the quality of the original data. Furthermore, it optimizes computational efficiency through a lightweight batch processing architecture, supporting large-scale automated data processing even in resource-constrained environments. This solves multiple bottlenecks in accuracy, adaptability, and scalability of existing technologies, resulting in a breakthrough improvement in accuracy, robustness, and practicality for the overall technical solution.

[0180] In summary, the embodiments of this application systematically solve the core problems of ambiguous input address text data and insufficient spatial positioning accuracy in existing technologies through targeted technological innovation, providing effective assistance for tasks such as logistics distribution, smart city management, and map navigation.

[0181] Reference Figure 5 This application provides an address standardization and spatial association device based on intelligent semantic parsing, including:

[0182] The standard address database construction unit is used to construct a standard address database that includes spatial hierarchy information and geospatial geometric information based on the administrative spatial hierarchy dataset and the building landmark spatial dataset.

[0183] The text address hierarchical unit is used to perform hierarchical parsing of the text address to be processed based on the standard address library to obtain target spatial hierarchy data.

[0184] The spatial matching and mapping unit is used to match the spatial geometric location corresponding to the spatial hierarchy information in the standard address database, and map the spatial geometric location to a visualized spatial geometric map.

[0185] It is understood that the content of the above method embodiments is applicable to the present device embodiments. The specific functions implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0186] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this application are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and sub-operations described as part of a larger operation are executed independently.

[0187] Furthermore, although this application is described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the described functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding this application. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional technology for an engineer. Therefore, those skilled in the art can implement the application set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of this application, which is determined by the full scope of the appended claims and their equivalents.

[0188] If the aforementioned functions are implemented as 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 this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0189] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0190] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0191] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0192] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0193] Although embodiments of this application have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the claims and their equivalents.

[0194] The above is a detailed description of the preferred embodiments of this application, but this application is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.

Claims

1. An address standardization and spatial association method based on intelligent semantic parsing, characterized in that, The method includes the following steps: A standard address database, including spatial hierarchy information and geospatial geometric information, is constructed based on the administrative spatial hierarchy dataset and the building landmark spatial dataset. Based on the standard address library, the text address to be processed is parsed hierarchically to obtain the target space hierarchy data; The spatial geometric location corresponding to the spatial hierarchy information is matched in the standard address database, and the spatial geometric location is mapped to a visualized spatial geometric map.

2. The address standardization and spatial association method based on intelligent semantic parsing according to claim 1, characterized in that, The construction of a standard address database, which includes spatial hierarchy information and geospatial geometric information, based on the administrative spatial hierarchy dataset and the building landmark spatial dataset includes the following steps: The process involves importing, spatially aligning, cleaning, and fusing raw address data from multiple sources to generate an intermediate dataset. The expression for the intermediate dataset is: ； in, For the intermediate dataset, , These are datasets of administrative spatial levels. and housing landmark spatial dataset The cleaning function, This represents a data integration function. The integration strategy is to map administrative levels step by step. Indicates the associated field; Each address record in the intermediate dataset is subjected to address standardization to obtain a standard address dataset; wherein, the address standardization process includes address text parsing and structuring, place name unification, administrative code mapping, spatial geographic coordinate standardization, and redundancy removal and merging; The expression for address normalization is: ； in, For the standard address dataset, For the standardized function, A standardized set of rules; The standard address dataset is logically split into layers according to spatial hierarchy and structured and stored in different data tables of a preset database, thereby constructing the standard address library that includes spatial hierarchy information and geospatial geometric information.

3. The address standardization and spatial association method based on intelligent semantic parsing according to claim 1, characterized in that, The process of performing hierarchical parsing of the text address to be processed based on the standard address library to obtain target spatial hierarchy data includes the following steps: The text address to be processed is preprocessed; the preprocessing methods include case conversion, Chinese character and number conversion, and place name alias conversion. The administrative spatial hierarchy data in the text address to be processed was extracted by NLP word segmentation and recognition. By utilizing the correlation between adjacent administrative space level data, semantic errors in the text address to be processed are corrected, and missing administrative space level data is supplemented, thereby restoring the text address to be processed into a complete address; Using the TextRCNN deep neural network model and combining the spatial hierarchy information stored in the standard address library, the restored text address to be processed is classified by text hierarchy to obtain the target spatial hierarchy data.

4. The address standardization and spatial association method based on intelligent semantic parsing according to claim 3, characterized in that, The TextRCNN deep neural network model performs the following steps on the restored text address to be processed: The contextual semantic information of each word in the restored text address to be processed is obtained by a bidirectional recurrent neural network, and the output of the recurrent neural network is concatenated with the embedding vector of the word to form a target semantic representation. word The left-hand context vector is calculated as follows: ； in, It has A dense vector of real values; It is a word Word embedding is a feature A dense vector of real values; It is the previous word The left-side context; the shared left-side context parameter of the first word in the restored text address to be processed. ; It transforms the hidden layer into the next hidden layer; This is a weight matrix used to combine the semantics of the current word with the upper left text of the next word; It is a non-linear activation function; word The right-hand context vector is calculated as follows: ； Among them, the right-hand context shared parameter of the last word in the restored text address to be processed ; left context vector Word embedding and the right-hand context vector By combining the words, we can obtain the word. The representation is as follows: ； After linear transformation and tanh activation function processing, the words are obtained. The final semantic representation is as follows: ； and words The final semantic representation The input is fed into subsequent layers of the bidirectional recurrent neural network for processing; in, These are latent semantic vectors used to analyze each semantic factor to determine the most useful factors for representing the text. For the corresponding bias term; The concatenated multi-dimensional word vectors are compressed into a low-dimensional semantic space through non-linear mapping, and key discriminative information is extracted from the restored text address using max pooling. Specifically, this includes performing max pooling on all word representations in the same dimension. ； in, This represents the j-th feature value of the i-th word; The output layer is defined as follows, and is used to map the pooled features to the classification space: ； in, This is the output layer weight matrix. For bias terms, The feature vector after pooling; The extracted high-dimensional features are classified using a fully connected softmax layer, and the probability prediction value for each spatial level category is calculated: ； in, This indicates that the restored address of the text to be processed belongs to the first... The probability of each spatial hierarchy category, This represents the total number of categories.

5. The address standardization and spatial association method based on intelligent semantic parsing according to claim 1, characterized in that, The process of matching the spatial geometric location corresponding to the spatial hierarchy information in the standard address database includes the following steps: The spatial hierarchy information is matched one-to-one with the building landmark information in the standard address database, thereby realizing the conversion of text address to spatial data and obtaining the spatial geometric location of the spatial hierarchy information.

6. The address standardization and spatial association method based on intelligent semantic parsing according to claim 1, characterized in that, The process of mapping the spatial geometric location to a visualized spatial geometric map includes the following steps: The spatial geometric location is mapped to a visualized spatial geometric map using a professional geographic information system engine; wherein the spatial geometric location includes coordinates, boundaries, and hierarchical metadata.

7. The address standardization and spatial association method based on intelligent semantic parsing according to any one of claims 1 to 6, characterized in that, The method further includes the following steps: The spatial hierarchical text information and spatial geographic geometric data corresponding to the address of the spatial geometric location are stored in the standard address database.

8. An address standardization and spatial association device based on intelligent semantic parsing, characterized in that, The device includes: The standard address database construction unit is used to construct a standard address database that includes spatial hierarchy information and geospatial geometric information based on the administrative spatial hierarchy dataset and the building landmark spatial dataset. The text address hierarchical unit is used to perform hierarchical parsing of the text address to be processed based on the standard address library to obtain target spatial hierarchy data. The spatial matching and mapping unit is used to match the spatial geometric location corresponding to the spatial hierarchy information in the standard address database, and map the spatial geometric location to a visualized spatial geometric map.

9. An electronic device, characterized in that, The electronic device includes a processor and a memory; The memory is used to store programs; The processor executes the program to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The storage medium stores a program that is executed by a processor to implement the method as described in any one of claims 1 to 7.

Images