Characteristic and attribute structured compilation method and system based on multiple voucher data of real estate

Abstract

The invention discloses a feature and attribute structured compilation method and system based on multiple kinds of real estate voucher data, and belongs to the field of feature and attribute structured compilation, and the method comprises the steps: obtaining various kinds of voucher data in the real estate field, extracting the key information of a voucher for each kind of voucher, and constructing a voucher information feature vector; for a plurality of credible certificates, key information in the certificates is extracted, structured analysis is carried out on the certificate information through a natural language processing technology, and a unified real estate information model is constructed; in order to solve the problems of information conflict and inconsistency between different vouchers, a rule-based reasoning method is adopted, real estate domain knowledge is combined, conflict solving rules are set, conflict information is verified and corrected, and information consistency is guaranteed.

Description

Technical Field

[0001] The present invention belongs to the field of structured compilation of characteristics and attributes, and in particular, relates to a method and system for structured compilation of characteristics and attributes based on multiple voucher materials of real estate. Background Art

[0002] In the real estate field, there are many types of real estate voucher materials, such as property certificates, land certificates, real estate registration certificates, etc. The credibility and importance of these voucher materials vary. How to make full use of these heterogeneous voucher materials for feature and attribute structured compilation is a technical problem that needs to be solved urgently. Specifically, different vouchers have different issuing agencies, legal effects, and information integrity, and there may be conflicts and inconsistencies in the information between the vouchers.

[0003] How to reasonably evaluate the credibility of the voucher, assign appropriate weights to different vouchers, and design an effective hierarchical and progressive feature and attribute structured compilation method will be the focus of the research. In addition, real estate voucher information is complex, and how to effectively integrate multi-source voucher information during the refinement process to avoid logical contradictions and information loss in the refined data is also a major challenge. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a method for structuring the characteristics and attributes of various real estate voucher materials, including:

[0005] Acquire multiple credential data of real estate, extract key information of the credential data, and construct a credential information feature vector;

[0006] Constructing a credibility evaluation algorithm based on a support vector machine algorithm, evaluating the credential information feature vector based on the credibility evaluation algorithm, obtaining a credibility score for each type of credential, and determining the importance weight of each type of credential through a hierarchical analysis method;

[0007] The credibility score of each type of credential and the importance weight of each type of credential are weighted averaged to obtain a comprehensive score of each type of credential. If the comprehensive score of each type of credential exceeds a preset threshold, the credential is determined to be a credible credential.

[0008] Extracting key information of the trusted credential, performing structured analysis on the key information of the trusted credential, and constructing a real estate information model;

[0009] Based on the comprehensive score according to the voucher, the information of the real estate information model is layered to obtain layered information, and the layered information is heterogeneously processed and then fused to generate a real estate knowledge graph.

[0010] Preferably, the process of constructing the credential information feature vector includes:

[0011] Acquire various voucher data of real estate, perform text recognition on the voucher data, and obtain text information of the voucher data; wherein the voucher data includes real estate certification data, financial voucher data, tax and expense vouchers, compliance documents and auxiliary information;

[0012] The key information of the multiple credential data is represented by vectors to obtain the credential information feature vectors.

[0013] Preferably, the process of obtaining the credibility score of each type of credentials and the importance weight of each type of credentials includes:

[0014] A credential credibility evaluation model is constructed by using a support vector machine algorithm, and the credential information feature vector is calculated based on the credential credibility evaluation model to obtain a credibility score for each type of credential;

[0015] In view of the legal effect of each type of certificate and the authority of the issuing body, the hierarchical analysis method is used to establish an evaluation index system. The weight of each indicator is determined through expert scoring, and the comprehensive weighted score is calculated to obtain the weight coefficient of the certificate's importance.

[0016] Preferably, the process of constructing a credential credibility assessment model by using a support vector machine algorithm includes:

[0017] Collect relevant credential data with credibility labels, perform word segmentation and vectorization on the text data of the relevant credential data, process missing values, normalize numerical features, and finally encode categorical variables to obtain a preprocessed data set;

[0018] Constructing an SVR model, inputting the preprocessed data set into the SVR model for training to obtain a trained model; wherein the kernel function of the SVR model is an RBF kernel;

[0019] The trained model is verified based on the K-fold cross-validation method, and the trained model is fine-tuned and retrained based on the verification result to obtain the credential credibility assessment model.

[0020] Preferably, the process of determining whether the credential is a credible credential includes:

[0021] Obtain the credibility score and importance weight data of each type of credentials and use them as input parameters of the weighted average algorithm;

[0022] A weighted average algorithm is used to calculate the comprehensive score of each type of credential by combining the credibility score and importance weight of the credential;

[0023] According to the calculated comprehensive scores, various vouchers are prioritized to obtain the voucher priority ranking results;

[0024] Construct a trusted credential determination threshold to determine whether the comprehensive score of each type of credential exceeds the preset trusted credential determination threshold. If it exceeds the threshold, the credential is marked as a trusted credential.

[0025] Preferably, the process of constructing the real estate information model includes:

[0026] Structuring the key information of the trusted credential, extracting corresponding attribute values ​​through regular expression matching, and obtaining parsing results;

[0027] Based on the analysis results, the field names and data types of the model are determined, and the key attributes obtained by the analysis are mapped to the corresponding fields to construct the real estate information data model.

[0028] Preferably, the process of generating the real estate knowledge graph includes:

[0029] Determine the schema of the knowledge graph based on the real estate information data model, define entity types, attributes and relationships, and construct a real estate domain ontology;

[0030] Extracting entities and relationships of the credential information feature vector to construct an initial knowledge graph;

[0031] Through field matching and entity alignment algorithms, determine whether entities in different data sources point to the same object. If so, merge them to obtain merged data.

[0032] Calculating the relevance of the credential information feature vector based on a long short-term memory network to obtain associated information;

[0033] Inputting the merged data and the associated information into the initial knowledge graph to generate an information knowledge graph;

[0034] The information knowledge graph is dimensionally integrated with user behavior data and market data to obtain the real estate knowledge graph.

[0035] On the other hand, the present invention also provides a system for structuring the characteristics and attributes of various real estate voucher materials, including:

[0036] A vector construction module, used to obtain various credential data of real estate, extract key information of the credential data, and construct a credential information feature vector;

[0037] A calculation module, used to construct a credibility evaluation algorithm based on a support vector machine algorithm, evaluate the credential information feature vector based on the credibility evaluation algorithm, obtain the credibility score of each type of credential, and determine the importance weight of each type of credential through a hierarchical analysis method;

[0038] A judgment module, used for performing weighted average of the credibility score of each type of credentials and the importance weight of each type of credentials to obtain a comprehensive score of each type of credentials, and if the comprehensive score of each type of credentials exceeds a preset threshold, the credentials are judged to be credible;

[0039] A model building module, used to extract key information of the trusted credential, perform structured analysis on the key information of the trusted credential, and build a real estate information model;

[0040] The refinement module is used to perform hierarchical processing on the information of the real estate information model based on the comprehensive score according to the voucher, obtain hierarchical information, perform heterogeneous processing on the hierarchical information, and then fuse it to generate a real estate knowledge graph.

[0041] Compared with the prior art, the present invention has the following advantages and technical effects:

[0042] The invention discloses a method for processing real estate voucher information. The method first obtains various types of real estate vouchers, extracts key information to construct feature vectors, evaluates the credibility of the vouchers through a support vector machine algorithm, and uses a hierarchical analysis method to determine the weights. Combined with the credibility score and weight, the comprehensive score is calculated and sorted to determine the credible voucher. The credible voucher is structured and parsed to build a unified real estate information model. For information conflicts, rules are set to verify and correct. According to the comprehensive score of the voucher, the information is processed in layers to gradually improve the accuracy. Finally, data fusion technology is used to integrate multi-source heterogeneous information and build a complete real estate knowledge graph. The present invention effectively solves the problems of credibility judgment, conflict processing and knowledge integration of real estate voucher information through multi-dimensional evaluation, layered processing and data fusion, and provides high-quality data support for real estate business. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0044] Figure 1 A schematic diagram of a method flow chart of an embodiment of the present invention;

[0045] Figure 2 Schematic diagram of the data structure of an embodiment of the present invention. DETAILED DESCRIPTION

[0046] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application are combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0047] It should be noted that the steps shown in the flowcharts of the accompanying drawings are executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described are executed in an order different from that shown here.

[0048] Embodiment 1

[0049] like Figure 1-2 As shown, this embodiment provides a method for structuring the characteristics and attributes of various real estate voucher materials, including:

[0050] Obtain various types of certificate information in the real estate field, including property certificates, land certificates, real estate registration certificates, etc. For each type of certificate, extract key information of the certificate, such as issuing agency, issuing time, legal effect, etc., and construct a certificate information feature vector.

[0051] Obtain various types of certificate information such as real estate certificates, land certificates, and real estate registration certificates in the real estate field, use optical character recognition technology to perform text recognition on the certificate information, and obtain the text information of the certificate information. According to the pre-constructed real estate field knowledge graph, extract key information such as the issuing agency, issuing time, and legal effect from the text information of the certificate information. For the extracted issuing agency information, obtain the authority, credibility and other attribute information of the agency through knowledge graph query. For the extracted issuing time information, determine whether the certificate is within the validity period. If it exceeds the validity period, mark it as an invalid certificate. For the extracted legal effect information, combine the knowledge graph and case library to determine the legal effect level of the certificate. Vectorize the key information such as the issuing agency, issuing time, and legal effect to construct the certificate information feature vector.

[0052] For the feature vector of each type of credential, the credibility of the credential is evaluated by the support vector machine algorithm to obtain the credibility score of each type of credential. At the same time, according to factors such as the legal effect of the credential and the authority of the issuing agency, the hierarchical analysis method is used to determine the importance weight of each type of credential.

[0053] According to the credential type, the corresponding feature vector is obtained, and the credential credibility evaluation model is constructed through the support vector machine algorithm. The model parameters are trained and optimized to obtain the credibility score of each type of credential. According to the legal effect of each type of credential and the authority of the issuing agency, the hierarchical analysis method is used to establish an evaluation index system. The weight of each index is determined by expert scoring, and the comprehensive weighted score is calculated to obtain the credential importance weight coefficient. The credential credibility score and the importance weight coefficient are weighted and integrated to obtain the credential comprehensive evaluation score;

[0054] According to the pre-built knowledge graph in the real estate field, key information such as issuing authority, issuing time and legal effect is extracted from it and converted into feature vectors. For example, the feature vector of a real estate certificate may contain information such as "issuing authority: Real Estate Registration Center of City A", "issuing time: January 1, 2023", and "legal effect: valid". Using these feature vectors, a credential credibility evaluation model based on support vector machine (SVR) is constructed. For example, a large number of real and forged real estate certificates are collected, their feature vectors are extracted, and they are marked as "credible" and "uncredible" respectively. Then, the SVR model is trained with these sample data to enable it to predict the credibility score of the certificate based on the input feature vector. The higher the score, the higher the credibility of the certificate. In addition to the credibility score, the importance of the certificate needs to be considered. For example, the importance of the real estate certificate is usually higher than that of the land certificate. The analytic hierarchy process (AHP) is used to establish an evaluation index system, such as the authority of the issuing authority and the legal effect of the certificate. The weight of each indicator is determined by expert scoring, and then the comprehensive weighted score of each certificate is calculated to obtain the weight coefficient of the importance of the certificate.

[0055] According to the credibility score and importance weight of the credential, the comprehensive score of each type of credential is calculated by weighted average to obtain the priority ranking result of the credential. If the comprehensive score of a credential exceeds the preset threshold, the credential is determined to be a credible credential.

[0056] For example, if the legal effect weight of a real estate certificate is 0.8 and the authority weight of the issuing agency is 0.6, then its importance weight coefficient may be 0.7. The credibility score of the certificate is weighted and integrated with its importance weight coefficient to obtain the comprehensive evaluation score of the certificate. For example, if the credibility score of a real estate certificate is 0.9 and the importance weight coefficient is 0.7, then its comprehensive evaluation score is 0.63. Certificates can be classified according to preset thresholds. For example, assume that the credibility threshold is 0.6 and the pending threshold is 0.3. If the comprehensive evaluation score of a real estate certificate is 0.7, it is judged as a credible certificate; if the score is 0.4, it is judged as a questionable certificate; if the score is 0.2, it is judged as an untrustworthy certificate.

[0057] For multiple trusted certificates, extract key information from the certificates, such as house area, owner, location, etc., use natural language processing technology to perform structured analysis on the certificate information and build a unified real estate information model.

[0058] Obtain multiple real estate-related trusted credentials, use natural language processing technology to analyze the credentials, and extract key information such as house area, owner, and location. Structural processing is performed on the extracted key information, and a named entity recognition algorithm is used to identify key attributes such as house area and owner, and the corresponding attribute values ​​are extracted through regular expression matching. According to the results of structured parsing, a unified real estate information data model is constructed, the field name and data type of the model are determined, and the key attributes obtained from the parsing are mapped to the corresponding fields. Standardization is performed on the field values ​​in the real estate information model, such as unifying the house area into square meters and formatting the owner's name, to ensure the consistency and standardization of the data in the model. The real estate information model is deduplicated using a data cleaning algorithm. For multiple records of the same property, whether it is duplicate data is determined based on key attributes such as house area and location, and the duplicate data is merged or deleted.

[0059] For example, a house purchase contract contains a large amount of text information. Natural language processing technology can be used to extract key information such as house area, house location, and owner. Specifically, a named entity recognition algorithm can be used to identify the entity "house area" and extract the corresponding attribute value, such as "100 square meters". The extracted key information needs to be structured for subsequent storage and analysis. For example, key attributes such as house area, owner, location, and their corresponding attribute values ​​can be stored in a structured data model. House area can be stored as a numeric type in square meters; owner can be stored as a string type; location can be stored as a string type, such as "Chaoyang District, Beijing". A unified real estate information data model needs to be built. This model needs to define the names and data types of various fields, such as house area (numeric type), owner (string type), location (string type), construction year (date type), etc. The key information extracted from different vouchers is mapped to the corresponding fields of the data model according to unified specifications. Standardize the field values ​​in the real estate information model, for example, standardize the house area into square meters and the owner's name into the format of "surname + first name". This can ensure the consistency and standardization of the data and avoid query and analysis errors caused by inconsistent data formats.

[0060] In order to solve the information conflicts and inconsistencies between different credentials, a rule-based reasoning method is adopted, combined with real estate field knowledge, conflict resolution rules are set, and conflicting information is verified and corrected to ensure the consistency of information.

[0061] Compare the voucher information to determine whether there are information conflicts and inconsistencies; if there are information conflicts and inconsistencies, obtain the domain knowledge related to the conflicting information based on the pre-established real estate knowledge graph; based on the acquired domain knowledge, use rule-based reasoning methods to set resolution rules for different types of conflicts; based on the set resolution rules, verify the conflicting and inconsistent voucher information to determine which information is correct and which is wrong; for the wrong conflicting information, use the information correction algorithm to correct the wrong information based on the correct information and resolution rules to obtain the corrected consistency information; save the verified and corrected consistency voucher information to form a unified and standardized real estate voucher information database;

[0062] There are various sources of information in the real estate field, and the information of different vouchers may conflict and be inconsistent, which affects the reliability of the data. In order to solve this problem, it is necessary to compare, verify and correct the voucher information. For example, multiple vouchers such as real estate certificates, purchase contracts, and house appraisal reports are obtained. The house area registered in the real estate certificate is 100 square meters, and the owner is Zhang San; the house area agreed in the purchase contract is 98 square meters, and the owner is Zhang San; the house area assessed in the house appraisal report is 102 square meters, and the owner information is missing. There are differences in the house area and owner information in these three vouchers. In view of these differences, it is necessary to use the real estate field knowledge map to determine which information is more reliable. The knowledge map contains professional knowledge in the real estate field. For example, the legal effect of the real estate certificate is higher than other vouchers, and the area value of the house appraisal report is more credible. Based on the knowledge map, some rules can be set to resolve conflicts. For example, when there is a difference in the house area between the real estate certificate and the purchase contract, the value of the real estate certificate is preferred; when the owner information is missing, the information of the real estate certificate is preferred. According to these rules, the above examples are verified. The house area in the real estate certificate is 100 square meters, and the area in the purchase contract is 98 square meters. Since the real estate certificate is more effective, 100 square meters is judged to be correct information and 98 square meters is potential error information. The area of ​​the house appraisal report is 102 square meters, which is close to the value in the real estate certificate and can be considered as an error within a reasonable range. In terms of owner information, both the real estate certificate and the purchase contract record it as Zhang San, but the appraisal report lacks this information, so Zhang San is used as the correct information. For potentially erroneous information, such as the house area of ​​98 square meters in the purchase contract, an information correction algorithm can be used to correct it.

[0063] On the basis of the real estate information model, the real estate information is processed in layers according to the comprehensive score and priority of the voucher. On the basis of layered processing, data fusion technology is used to integrate real estate information from different sources and types. Through methods such as field matching and entity alignment, the association and fusion of multi-source heterogeneous information is achieved, and a complete and accurate real estate knowledge graph is constructed to provide high-quality data support for subsequent business applications.

[0064] For example, convert housing transaction data from different regions into a unified format, and handle missing values ​​and outliers. For data from real estate agency websites, it is necessary to pay attention to the real-time and accuracy of the data, and filter and identify false housing information. For example, by analyzing housing descriptions, pictures, prices and other information, identify and remove false housing. For data from map service providers, it is necessary to pay attention to the accuracy and coverage of geographic location information, and calibrate and convert geographic coordinates. For example, convert geographic location information under different coordinate systems into a unified coordinate system. The schema of the real estate knowledge graph needs to be designed according to specific business needs. For example, define the "community" entity type, whose attributes include community name, construction year, volume ratio, greening rate, etc.; define the "house" entity type, whose attributes include house area, orientation, floor, price, etc.; define the "transaction" relationship, connect the "house" and "buyer" and "seller" entities, and represent housing transaction behavior. Building a real estate domain ontology can help better understand and organize concepts and relationships in the real estate field. For example, define the concept of "school district housing", which has an association relationship with entities such as "school" and "community". Using natural language processing technology, entities and relationships can be extracted from unstructured data. For example, entities such as "community name", "house area", and "orientation" can be identified from the house description text, as well as the relationship between them. For example, from the sentence "a three-bedroom apartment with north-south ventilation, located in a high-end community in the core area of ​​Jing'an District", information such as "house area" is "three-bedroom", "orientation" is "north-south ventilation", and "community location" is "core area of ​​Jing'an District" can be extracted, and these information can be associated with the corresponding entities. Through field matching and entity alignment algorithms, the same entity in different data sources can be merged. For example, house transaction information in government public data and house information on real estate agency websites may point to the same house. By comparing house address, area, floor and other information, it can be determined whether they point to the same entity and merge them to avoid information duplication and redundancy. For example, the house address recorded in the government data is "Room 301, Unit 2, Building 1, Community A", and the house address recorded on the agency website is "1-2-301, Community A". Through the address matching algorithm, it can be determined that they point to the same house. Using machine learning algorithms, implicit relationships between entities can be mined. For example, by analyzing the user's browsing history and purchase history, we can find out the user's preference for certain types of houses and add this information to the knowledge graph. For example, if the user often browses the information of school district houses, we can infer that the user has a high interest in school district houses and establish an "interested" relationship between "user" and "school district house". The quality assessment of the constructed real estate knowledge graph can be evaluated from the dimensions of completeness, accuracy, consistency, etc. For example, check whether there are missing entities or relationships in the knowledge graph, whether there are wrong attribute values, whether there are logical contradictions, etc.Through quality assessment, we can identify problems and defects in the knowledge graph and make targeted optimization and improvements. For example, if we find that the knowledge graph lacks information about certain communities, we can supplement the data source or collect information manually. By associating the real estate knowledge graph with other data sources, we can further enrich the information dimension of the knowledge graph. For example, by associating user behavior data with the knowledge graph, we can analyze user preferences for different types of houses; by associating market data with the knowledge graph, we can predict housing price trends.

[0065] On the other hand, this embodiment also provides a structured compilation system based on the characteristics and attributes of various real estate voucher materials, including:

[0066] A vector construction module, used to obtain various credential data of real estate, extract key information of the credential data, and construct a credential information feature vector;

[0067] A calculation module, used to construct a credibility evaluation algorithm based on a support vector machine algorithm, evaluate the credential information feature vector based on the credibility evaluation algorithm, obtain the credibility score of each type of credential, and determine the importance weight of each type of credential through a hierarchical analysis method;

[0068] A judgment module, used for performing weighted average of the credibility score of each type of credentials and the importance weight of each type of credentials to obtain a comprehensive score of each type of credentials, and if the comprehensive score of each type of credentials exceeds a preset threshold, the credentials are judged to be credible;

[0069] A model building module, used to extract key information of the trusted credential, perform structured analysis on the key information of the trusted credential, and build a real estate information model;

[0070] The refinement module is used to perform hierarchical processing on the information of the real estate information model based on the comprehensive score according to the voucher, obtain hierarchical information, perform heterogeneous processing on the hierarchical information, and then fuse it to generate a real estate knowledge graph.

[0071] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application 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 structuring the characteristics and attributes of various real estate voucher materials, characterized in that: include: Acquire multiple credential data of real estate, extract key information of the credential data, and construct a credential information feature vector; Constructing a credibility evaluation algorithm based on a support vector machine algorithm, evaluating the credential information feature vector based on the credibility evaluation algorithm, obtaining a credibility score for each type of credential, and determining the importance weight of each type of credential through a hierarchical analysis method; The credibility score of each type of credential and the importance weight of each type of credential are weighted averaged to obtain a comprehensive score of each type of credential. If the comprehensive score of each type of credential exceeds a preset threshold, the credential is determined to be a credible credential. Extracting key information of the trusted credential, performing structured analysis on the key information of the trusted credential, and constructing a real estate information model; Based on the comprehensive score according to the voucher, the information of the real estate information model is layered to obtain layered information, and the layered information is heterogeneously processed and then fused to generate a real estate knowledge graph.

2. The method according to claim 1, characterized in that The process of constructing the credential information feature vector includes: Acquire various voucher data of real estate, perform text recognition on the voucher data, and obtain text information of the voucher data; wherein the voucher data includes real estate certification data, financial voucher data, tax and expense vouchers, compliance documents and auxiliary information; The key information of the multiple credential data is represented by vectors to obtain the credential information feature vectors.

3. The method according to claim 1, characterized in that The process of obtaining the credibility score of each type of credentials and the importance weight of each type of credentials includes: A credential credibility evaluation model is constructed by using a support vector machine algorithm, and the credential information feature vector is calculated based on the credential credibility evaluation model to obtain a credibility score for each type of credential; In view of the legal effect of each type of certificate and the authority of the issuing body, the hierarchical analysis method is used to establish an evaluation index system. The weight of each indicator is determined through expert scoring, and the comprehensive weighted score is calculated to obtain the weight coefficient of the certificate's importance.

4. The method according to claim 3, characterized in that The process of constructing a credential credibility evaluation model by using a support vector machine algorithm includes: Collect relevant credential data with credibility labels, perform word segmentation and vectorization on the text data of the relevant credential data, process missing values, normalize numerical features, and finally encode categorical variables to obtain a preprocessed data set; Constructing an SVR model, inputting the preprocessed data set into the SVR model for training to obtain a trained model; wherein the kernel function of the SVR model is an RBF kernel; The trained model is verified based on the K-fold cross-validation method, and the trained model is fine-tuned and retrained based on the verification result to obtain the credential credibility assessment model.

5. The method according to claim 1, characterized in that: The process of determining whether the credential is a credible credential includes: Obtain the credibility score and importance weight data of each type of credentials and use them as input parameters of the weighted average algorithm; A weighted average algorithm is used to calculate the comprehensive score of each type of credential by combining the credibility score and importance weight of the credential; According to the calculated comprehensive scores, various vouchers are prioritized to obtain the voucher priority ranking results; Construct a trusted credential determination threshold to determine whether the comprehensive score of each type of credential exceeds the preset trusted credential determination threshold. If it exceeds the threshold, the credential is marked as a trusted credential.

6. The method according to claim 1, characterized in that The process of constructing the real estate information model includes: Structuring the key information of the trusted credential, extracting corresponding attribute values ​​through regular expression matching, and obtaining parsing results; Based on the analysis results, the field names and data types of the model are determined, and the key attributes obtained by the analysis are mapped to the corresponding fields to construct the real estate information data model.

7. The method according to claim 1, characterized in that The process of generating the real estate knowledge graph includes: Determine the schema of the knowledge graph based on the real estate information data model, define entity types, attributes and relationships, and construct a real estate domain ontology; Extracting entities and relationships of the credential information feature vector to construct an initial knowledge graph; Through field matching and entity alignment algorithms, determine whether entities in different data sources point to the same object. If so, merge them to obtain merged data. Calculating the relevance of the credential information feature vector based on a long short-term memory network to obtain associated information; Inputting the merged data and the associated information into the initial knowledge graph to generate an information knowledge graph; The information knowledge graph is dimensionally integrated with user behavior data and market data to obtain the real estate knowledge graph.

8. A structured compilation system based on the characteristics and attributes of various real estate voucher materials, characterized in that: include: A vector construction module, used to obtain various credential data of real estate, extract key information of the credential data, and construct a credential information feature vector; A calculation module, used to construct a credibility evaluation algorithm based on a support vector machine algorithm, evaluate the credential information feature vector based on the credibility evaluation algorithm, obtain the credibility score of each type of credential, and determine the importance weight of each type of credential through a hierarchical analysis method; A judgment module, used for performing weighted average of the credibility score of each type of credentials and the importance weight of each type of credentials to obtain a comprehensive score of each type of credentials, and if the comprehensive score of each type of credentials exceeds a preset threshold, the credentials are judged to be credible; A model building module, used to extract key information of the trusted credential, perform structured analysis on the key information of the trusted credential, and build a real estate information model; The refinement module is used to perform hierarchical processing on the information of the real estate information model based on the comprehensive score according to the voucher, obtain hierarchical information, perform heterogeneous processing on the hierarchical information, and then fuse it to generate a real estate knowledge graph.

Images