A method for recommending geographical entities

Abstract

The present invention discloses a method for recommending geographical entities, which relates to the technical field of urban transportation. First, the comment data sources and geographical entity lists required for the components of the geographical entity semantic network model in the research area are determined. Secondly, co-occurrence relationships, comment viewpoints, location semantics, comment topics, etc. of the comment data are analyzed, and different types of geographical entity semantic networks are calculated. Then, complex network technology is used to calculate the node index values of various geographical entity semantic networks to achieve clustering, evaluation and recommendation analysis of geographical entities. Finally, according to the usage of common recommended keywords, the corresponding geographical entity semantic networks are pre-generated, thereby accelerating the evaluation and recommendation speed of relevant geographical entity semantic networks, enhancing the scientificity and practicality of the construction of geographical entity semantic networks, and having a strong ability to recommend geographical entities based on text language analysis.

Description

Technical Field

[0001] The present invention relates to the technical fields of urban planning, tourism planning, urban traffic, and complex network modeling, and in particular to a method for recommending geographical entities. Background Art

[0002] Currently, comment data on various social media is closely related to geospatial entities and can reflect various characteristic information of geospatial entities, such as cultural attributes, tourism attributes, traffic attributes, and location attributes, etc. The comment data is basically mainly unstructured data, and the mining of these attribute information requires the assistance of natural language processing technology to complete the extraction of multiple attribute information.

[0003] There are still the following deficiencies in the prior art:

[0004] (1) Existing research often only calculates the "co-occurrence relationship" of certain keywords in comments to determine the semantic network relationship between different geographical entities, without effectively screening the keywords;

[0005] (2) The similarity relationship of comment viewpoints is not used to construct the semantic network relationship between geographical entities;

[0006] (3) Constructing the semantic network relationship of geographical entities based on the location semantic relationship is conducive to deeply revealing the spatial location connection relationship between geographical entities, which is rarely used in the prior art, especially in the modeling of the semantic network relationship of geographical entities. Summary of the Invention

[0007] The technical problem to be solved by the present invention is to overcome the deficiencies of the prior art and provide a method for recommending geographical entities. The present invention can construct semantic networks of different types of geographical entities from the mined multi-dimensional attribute information according to the comment data, thereby improving the evaluation ability of geographical entities based on text content and having a strong ability to recommend geographical entities based on text language analysis.

[0008] The present invention adopts the following technical solutions to solve the above technical problems:

[0009] A method for recommending geographical entities according to the present invention includes the following steps:

[0010] Step 1, determine the comment data and the list of geographical entities required for constructing the semantic network of geographical entities in the study area;

[0011] Step 2, extract the core focus words in each comment data, and obtain the first geographical entity semantic network in combination with the list of geographical entities;

[0012] Extract the comment viewpoints in each comment data, and obtain the second geographical entity semantic network in combination with the list of geographical entities;

[0013] Extract the keywords of the positional semantic relationship between geographical entities in each comment data, and combine with the geographical entity list to obtain the third geographical entity semantic network;

[0014] Extract the topic words reflecting the comment theme in each comment data, and combine with the geographical entity list to obtain the fourth geographical entity semantic network;

[0015] Step 3: Use the core focus words, comment viewpoints, keywords of positional semantic relationships, and topic words formed in Step 2 as the recommended keywords for geographical entities. According to the recommended keywords selected by the user, match the corresponding type of geographical entity semantic network, generate the geographical entity semantic network, and then push the geographical entities most relevant to the user's interests through the newly generated geographical entity semantic network.

[0016] As a further optimized solution of the geographical entity recommendation method described in the present invention, each geographical entity in the geographical entity list has longitude and latitude coordinate values, and the geographical entity list is obtained in the following manner:

[0017] Step 1.1: Each geographical entity in the study area has a unique number and name, and crawl the comment data for each geographical entity;

[0018] Step 1.2: Map each crawled comment data to the corresponding geographical entity and establish a one-to-one correspondence.

[0019] As a further optimized solution of the geographical entity recommendation method described in the present invention,

[0020] The method for obtaining the first geographical entity semantic network in Step 2 is specifically as follows:

[0021] Step 2.1: Based on the TF-TDF algorithm, extract the keywords in each comment data, and perform word frequency statistics on the keywords extracted from all comment data;

[0022] Step 2.2: Establish a co-occurrence relationship network of keywords with word frequency greater than the preset threshold in the comment data, and calculate the intensity value of the core keywords of the co-occurrence relationship network;

[0023] Step 2.3: Screen out the core focus words according to the word frequency quantity of keywords in all comment data and the intensity value of core keywords in the co-occurrence relationship network;

[0024] Step 2.4: Use geographical entities as the network nodes of the geographical entity semantic network. If a certain core focus word appears in the respective comment data of two different geographical entities, it is considered that there is a complex network connection edge between these two geographical entities;

[0025] Step 2.5: Based on complex network theory, construct the first geographical entity semantic network through the relationship between network nodes and complex network connection edges;

[0026] The method for obtaining the second geographical entity semantic network in Step 2 is specifically as follows:

[0027] Step 3.1: Based on the online API of Baidu natural language processing, extract the comment viewpoints in each comment data, and each comment viewpoint is composed of a combination of an attribute word and a descriptive word;

[0028] Step 3.2: According to the combination characteristics of the attribute words and descriptive words of the comment viewpoints, further refine and summarize the comment viewpoint list;

[0029] Step 3.3: Based on complex network theory, use geographical entities as the network nodes of the geographical entity semantic network. If a certain comment viewpoint in the comment viewpoint list appears in the respective comment data of two different geographical entities, it is considered that there is a complex network connection edge between these two geographical entities;

[0030] Step 3.4: Based on complex network theory, construct the second geographical entity semantic network through the relationship between network nodes and complex network connection edges;

[0031] The method for obtaining the third geographical entity semantic network in Step 2 is specifically as follows:

[0032] Step 4.1: Extract all geographical entities in each comment data;

[0033] Step 4.2: Construct keywords representing location semantic relationships and their corresponding weight values I; where the weight value I > 1;

[0034] Step 4.3: Based on natural language semantic analysis technology, extract the location semantic connection relationships between all pairs of geographical entities in each comment data, that is:

[0035] If there is no keyword representing the location semantic relationship described in Step 4.2 between two geographical entities in a comment data, it is considered that there is a complex network connection edge between these two geographical entities, and the weight of the edge is 1;

[0036] If there is a keyword representing the location semantic relationship described in Step 4.2 between two geographical entities in a comment data, it is considered that there is a complex network connection edge between these two geographical entities, and the weight of the edge is the weight value corresponding to the keyword representing the location semantic relationship described in Step 4.2; where, if there are multiple keywords representing the location semantic relationship described in Step 4.2, the keyword with the maximum weight value is used to determine the network edge connection relationship;

[0037] Step 4.4: Based on complex network theory, using geographical entities as network nodes of the geographical entity semantic network, and then constructing the third geographical entity semantic network through the relationships between the complex network connection edges determined in Step 4.3;

[0038] The method for obtaining the fourth geographical entity semantic network in Step 2 is specifically as follows:

[0039] Step 5.1: Based on the online API of Baidu natural language processing, extract the topic words of each comment data;

[0040] Step 5.2: Summarize the topic list of all comment data according to the topic words;

[0041] Step 5.3: Based on complex network theory, using geographical entities as network nodes of the geographical entity semantic network, if a certain topic in the topic list appears in the respective comment data of two different geographical entities, it is considered that there is a complex network connection edge between these two geographical entities;

[0042] Step 5.4: Based on complex network theory, construct the fourth geographical entity semantic network through the relationship between network nodes and complex network connection edges.

[0043] As a further optimization scheme of the geographical entity recommendation method described in the present invention, after generating the first to fourth geographical entity semantic networks, calculate various index values of network nodes for each geographical entity semantic network respectively; perform geographical entity clustering analysis based on the obtained various index values of network nodes, and perform spatial visualization display on geographical entities according to the clustering results.

[0044] As a further optimization scheme of the geographical entity recommendation method described in the present invention, the specific steps for calculating the intensity value of the core keyword in Step 2.2 are as follows:

[0045] Step 2.2.1: Based on complex network theory, for all comment data, using the core keyword as the network node, if two core keywords appear in one comment, it is considered that there is a complex network connection edge between these two core keywords, and a keyword complex network is established accordingly;

[0046] Step 2.2.2: For the keyword complex network established in Step 2.2.1, calculate the index values of degree centrality, weighted proximity centrality, weighted degree centrality, weighted betweenness centrality, and eigenvector centrality of the network node;

[0047] Step 2.2.3: Based on the entropy value method, calculate the comprehensive evaluation value of the network node for all the indexes obtained in Step 2.2.2; this comprehensive evaluation value is the intensity value of the core keyword;

[0048] The specific calculation steps for refining and summarizing the list of comment viewpoints in step 3.2 are as follows:

[0049] Step 3.2.1: Combine the attribute words with similar semantics in the comment viewpoints;

[0050] Step 3.2.2: Combine the descriptive words with similar semantics in the comment viewpoints;

[0051] Step 3.2.3: Based on the results of the combination processing in steps 3.2.1 and 3.2.2, remove the duplicate comment viewpoints with the same attribute words and descriptive words in the comment viewpoints, and then obtain the final list of comment viewpoints.

[0052] As a further optimization scheme of the geographical entity recommendation method described in the present invention, in step 4.1, all geographical entities in each comment data are extracted, including the geographical entity to which the comment data belongs, regardless of whether the comment data includes the geographical entity to which the comment data belongs.

[0053] As a further optimization scheme of the geographical entity recommendation method described in the present invention, the method for updating the semantic network of the mth geographical entity: respectively reconstruct the semantic networks of the first to fourth geographical entities according to the new comment data, or only construct the semantic network of the mth geographical entity for the new comment data; wherein, the network node index results of the newly constructed semantic network of the mth geographical entity are superimposed and calculated with the network node index results of the original semantic network of the mth geographical entity to update the network node index results of the semantic network of the mth geographical entity; m = 1, 2, 3, 4.

[0054] As a further optimization scheme of the geographical entity recommendation method described in the present invention, the core focus words, comment viewpoints, keywords of location semantic relationships, and topic words formed in step 2 are used as the recommended keywords for geographical entities;

[0055] Construct a geographical entity recommendation system, which provides an interactive interface and provides functions for clicking and querying the recommended keywords on the system interface. When the user clicks or queries a certain recommended keyword, the system automatically displays the list of geographical entities most relevant to the recommended keyword and visualizes it geospatially.

[0056] As a further optimization scheme of the geographical entity recommendation method described in the present invention, the implementation method of the geographical entity recommendation system is as follows:

[0057] Step A.1: When the user clicks or queries a certain recommended keyword, the geographical entity recommendation system determines which category among the semantic networks of the first to fourth geographical entities the recommended keyword belongs to;

[0058] Step A.2: According to the type of the geographical entity semantic network determined in Step A.1, only use the recommended keyword clicked or queried by the user as the judgment condition for determining the complex network connection edges in a certain type of geographical entity semantic network determined in Step A.1;

[0059] Step A.3: Generate the corresponding geographical entity semantic network according to the generation step requirements of a certain type of geographical entity semantic network determined in Step A.1;

[0060] Step A.4: Calculate various index values of the network nodes in the geographical entity semantic network generated in Step A.3, and screen the geographical entities according to the size order of any type of index value of the network nodes, and complete the spatial visualization display of the geographical entities.

[0061] Compared with the prior art, the present invention adopts the above technical solutions and has the following technical effects:

[0062] (1) The present invention provides a method for constructing a geographical entity semantic network and a geographical entity recommendation method. By analyzing co-occurrence relationships, comment views, location semantics, comment themes, etc. in comment data, different types of geographical entity semantic networks are calculated to more comprehensively reflect the multi-dimensional attribute characteristics of geographical entities; the present invention can also pre-generate the corresponding geographical entity semantic network according to the usage of common recommended keywords, thereby accelerating the recommendation speed of relevant geographical entity semantics;

[0063] (2) The present invention is based on the multi-dimensional attribute characteristics contained in the text of geographical entities, and performs semantic modeling on the multi-dimensional attribute characteristics, and proposes a method for constructing a geographical entity semantic network and a geographical entity recommendation method;

[0064] (3) The present invention proposes a method for constructing a geographical entity semantic network and a geographical entity recommendation method, and provides an offline processing technical method for updating result data, which can better realize that the existing operation system service is not affected when the data results are updated. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 is the overall flow schematic diagram of the present invention.

[0066] Figure 2 is the schematic diagram of the correspondence between comment data and geographical entities.

[0067] Figure 3 is the schematic diagram of the processing process of comment core focus words.

[0068] Figure 4 is the schematic diagram of the processing process of the first geographical entity semantic network A.

[0069] Figure 5It is a schematic diagram of the processing process of the second geographical entity semantic network B.

[0070] Figure 6 It is a schematic diagram of the processing process of the third geographical entity semantic network C.

[0071] Figure 7 It is a schematic diagram of the situation where there are no location semantic keywords between geographical entities in the comment data.

[0072] Figure 8 It is a schematic diagram of the situation where there are location semantic keywords between geographical entities in the comment data.

[0073] Figure 9 It is a schematic diagram of the processing process of the fourth geographical entity semantic network D.

[0074] Figure 10 It is a schematic diagram of the geographical entity recommendation process.

[0075] Figure 11 It is a schematic diagram of the geographical entity semantic network update process. Detailed implementation manners

[0076] The technical solutions of the present invention will be further described in detail below with reference to the accompanying drawings:

[0077] The present invention proposes to establish the semantic network relationship between related geographical entities through the extracted multi-dimensional attribute information, so as to realize the associated processing of different geographical entities in the geographical space. Further, the geographical entities of the content concerned by the user can be recommended to the user and timely feedback can be provided by means of the completed semantic network relationship.

[0078] Specifically, the present invention solves the deficiencies of the existing problems and conducts technological innovation from the following aspects:

[0079] (1) Existing research often only calculates the "co-occurrence relationship" of certain keywords in the comments to determine the semantic network relationship between different geographical entities, without effectively screening the keywords. By calculating the complex network index value of the "co-occurrence relationship" of keywords in the comments, the core attention words are screened out. Then, according to the "co-occurrence relationship" of the core attention words in different geographical entities, the semantic network relationship between geographical entities is solved. The present invention first models the "co-occurrence relationship" between keywords to screen out the core attention words, and then determines the semantic connection relationship between geographical entities through the "co-occurrence relationship" between the core attention words. Generally speaking, the present invention calculates the "co-occurrence relationship" twice on the basis of the existing technology to construct the semantic network relationship between geographical entities.

[0080] (2) The present invention proposes to construct the semantic network relationship between geographical entities by using the similarity relationship of comment viewpoints, which is ignored in the existing technical methods. Comment viewpoints are the concentrated reflection of the core content of comments. A comment may contain multiple comment viewpoints. A comment viewpoint is composed of an attribute word and a descriptive word. Attribute words generally only represent the attributes or characteristics of people or things and have the function of distinction or classification. Descriptive words generally refer to adjectives, which are mainly used to describe or modify nouns or pronouns, indicating the nature, state, characteristics or attributes of people or things. They are commonly used as attributives and can also be used as predicatives or complements. The combination of an attribute word and a descriptive word can form a comment viewpoint. By means of the combination relationship between the attribute word and the descriptive word, the semantic similarity degree between comments can be obtained.

[0081] (3) The positional semantic relationship between different geographical entities in a comment is of great significance for the connection level between geographical entities. For example, in the comment sentence "Coming out of Guandong Street is the ancient canal. Now, no historical traces can be seen, only a small stele inscribed here as the ancient canal", there is a certain positional semantic relationship between the two geographical entities of Guandong Street and the ancient canal. That is to say, there is a strong geographical position connection relationship between them. Therefore, constructing the semantic network relationship for geographical entities based on the positional semantic relationship is conducive to deeply revealing the spatial position connection relationship between geographical entities. This is rarely used in the existing technology, especially in the modeling of the semantic network relationship of geographical entities.

[0082] At the same time, the present invention re - designs the weights of the keywords representing the positional semantic relationship. If a comment does not contain the pre - determined keywords representing the positional semantic relationship between two geographical entities, it is considered that there is only a positional semantic connection relationship with a weight of 1 between the two geographical entities. When a comment contains the pre - determined keywords representing the positional semantic relationship between two geographical entities, it is considered that there is only a positional semantic connection relationship with a weight greater than 1 between the two geographical entities. This technical processing method is conducive to deeper revealing the different - strength positional semantic connection relationships between geographical entities.

[0083] (4) The present invention also considers the similarity of themes between comments to construct the semantic network relationship between geographical entities and conducts integrated innovation research on technical methods by combining other types of semantic networks of geographical entities. Specifically, it is to conduct cluster analysis through the constructed semantic network relationship between different geographical entities to evaluate and analyze the connection relationship of geographical entities from multiple dimensions.

[0084] (5) Step 9 of the present invention provides a new method and idea to solve the problem of how to quickly implement geographical entity recommendation. First, a recommendation system is provided. The recommendation system is designed to match the corresponding type of geographical entity semantic network according to the recommended keywords selected by the user, generate the relevant geographical entity semantic network, and then push the geographical entities most relevant to the user's interests through the newly generated geographical entity semantic network. Secondly, according to the recommended keywords frequently selected by the user, different geographical entity semantic networks are pre-generated, thereby improving the response speed of the geographical entity recommendation system.

[0085] (6) Step 10 of the present invention further proposes an update strategy for the geographical entity semantic network. It can either adopt the strategy of merging the updated new data sources or only construct the geographical entity semantic network for the new data, and then update the final result by means of superposition calculation of the node index values of the geographical entity semantic network. This is of great significance for offline updating the core result data of the system without affecting the services of the existing operating system.

[0086] Step 1) Refer to Appendix Figure 1 , and determine the comment data and geographical entity list required for the components of the geographical entity semantic network model in the study area. Among them, each geographical entity in the geographical entity list has longitude and latitude coordinate values;

[0087] Step 1.1) Each geographical entity in the study area has a unique number and name, and the comment data of each geographical entity is captured;

[0088] Step 1.2) Refer to Appendix Figure 2 , and each captured comment data is corresponding to the geographical entity it belongs to, and a one-to-one correspondence is established.

[0089] Step 2) Refer to Appendix Figure 3 , extract the core focus words in each comment data, and calculate to obtain the first geographical entity semantic network A;

[0090] Step 2.1) Refer to Appendix Figure 4 , based on the TF-TDF algorithm, extract the keywords in each comment data, and conduct word frequency statistics on the keywords extracted from all comment data;

[0091] Step 2.2) Establish the co-occurrence relationship network of the keywords with word frequency greater than the preset threshold in the comment data, and calculate the intensity value of the core keywords of the co-occurrence relationship network.

[0092] Among them, the calculation steps of the intensity value of the core keywords specifically include:

[0093] Step 2.2.1: Based on complex network theory, for all comment data, using the core keywords as network nodes, if two core keywords appear in the same comment, it is considered that there is a complex network connection edge between these two core keywords, and a keyword complex network is established accordingly;

[0094] Step 2.2.2: For the keyword complex network established in Step 2.2.1, calculate the index values of degree centrality, weighted proximity centrality, weighted degree centrality, weighted betweenness centrality, and eigenvector centrality of the network nodes;

[0095] Step 2.2.3: Based on the entropy value method, calculate the comprehensive evaluation value of the network nodes for all the indicators obtained in Step 2.2.2; this comprehensive evaluation value is the intensity value of the core keyword;

[0096] Step 2.3: According to the word frequency of keywords in all comment data and the intensity value of core keywords in the co-occurrence relationship network, screen out the core attention words;

[0097] Step 2.4: Using geographical entities as the network nodes of the geographical entity semantic network, if a core attention word appears in the comment data of two different geographical entities respectively, it is considered that there is a complex network connection edge between these two geographical entities;

[0098] Step 2.5: Based on complex network theory, construct the first geographical entity semantic network through the relationship between network nodes and complex network connection edges;

[0099] Step 3: Refer to Appendix Figure 5 , extract the comment viewpoints in each comment data, and calculate to obtain the second geographical entity semantic network B;

[0100] Step 3.1: Based on the online API of Baidu natural language processing, extract the comment viewpoints in each comment data, and each comment viewpoint is composed of an attribute word and a descriptive word;

[0101] Step 3.2: According to the combination characteristics of the attribute words and descriptive words of the comment viewpoints, further refine and summarize the comment viewpoint list;

[0102] Among them, the calculation steps for refining and summarizing the comment viewpoint list specifically include:

[0103] Step 3.2.1: Merge the attribute words with similar semantics in the comment viewpoints;

[0104] Step 3.2.2: Merge the descriptive words with similar semantics in the comment viewpoints;

[0105] Step 3.2.3) Based on the results after the merging process in Steps 3.2.1 and 3.2.2, duplicate comments with the same attribute words and descriptive words in the comment viewpoints are removed, and then the final list of comment viewpoints is obtained.

[0106] Step 3.3) Based on the complex network theory, with geographical entities as the network nodes of the geographical entity semantic network, if a certain comment viewpoint in the list of comment viewpoints appears in the respective comment data of two different geographical entities, it is considered that there is a complex network connection edge between these two geographical entities.

[0107] This algorithm step also includes the following considerations: After the comment viewpoints of each geographical entity are replaced with the combined attribute words and descriptive words after merging in Steps 3.2.1 and 3.2.2, the condition is then judged that if a certain comment viewpoint in the list of comment viewpoints appears in the respective comments of two different geographical entities.

[0108] Step 3.4) Based on the complex network theory, through the relationship between the network nodes and the complex network connection edges, the second geographical entity semantic network B is constructed.

[0109] Step 4) Refer to the appendix Figure 6 , extract the keywords of the positional semantic relationship between geographical entities in each comment data, and calculate to obtain the third geographical entity semantic network C;

[0110] Step 4.1) Extract all geographical entities in each comment data; all geographical entities extracted in each comment statement in this step include the geographical entity to which this comment belongs, regardless of whether this comment statement includes the geographical entity to which this comment belongs.

[0111] Step 4.2) Construct the keywords representing the positional semantic relationship and the corresponding weight value I; where the weight value I > 1;

[0112] Step 4.3) Based on the natural language semantic analysis technology, extract the positional semantic connection relationship between all pairs of geographical entities in each comment data, that is:

[0113] Refer to the appendix Figure 7 , if there is no keyword representing the positional semantic relationship described in Step 4.2 between two geographical entities in a comment statement, it is considered that there is a complex network connection edge between these two geographical entities, and the weight of the edge is 1;

[0114] Refer to the appendix Figure 8, if there are keywords indicating the location semantic relationship described in Step 4.2 between two geographical entities in a comment statement, it is considered that there is a complex network connection edge between these two geographical entities, and the weight of the edge is the weight value determined by the keyword expressing the location semantic relationship described in Step 4.2. Among them, if there are multiple keywords indicating the location semantic relationship described in Step 4.2, the keyword with the maximum weight value is used to determine the calculation of the network edge connection relationship.

[0115] Step 4.4) Based on complex network theory, using geographical entities as the network nodes of the geographical entity semantic network, and then constructing the third geographical entity semantic network C through the relationship between the complex network connection edges determined in Step 4.3.

[0116] Step 5) Refer to Appendix Figure 9 , extract the topic words reflecting the comment topics in each comment data, and calculate to obtain the fourth geographical entity semantic network D;

[0117] Step 5.1) Based on the online API of Baidu natural language processing, extract the topic words of each comment data;

[0118] Step 5.2) Summarize the topic list of all comment data according to the topic words;

[0119] Step 5.3) Based on complex network theory, using geographical entities as the network nodes of the geographical entity semantic network, if a certain topic in the topic list appears in the respective comment data of two different geographical entities, it is considered that there is a complex network connection edge between these two geographical entities;

[0120] Step 5.4) Based on complex network theory, construct the fourth geographical entity semantic network through the relationship between the network nodes and the complex network connection edges.

[0121] Step 6) For all the first to fourth geographical entity semantic networks generated in Step 2, Step 3, Step 4, and Step 5, calculate various index values of the network nodes for each geographical entity semantic network respectively. Various index values of the network nodes include degree centrality, weighted closeness centrality, weighted degree centrality, weighted betweenness centrality, and eigenvector centrality index values of the network nodes.

[0122] Step 7) Conduct geographical entity clustering analysis based on the various index values of the network nodes obtained above, and perform spatial visualization display on the geographical entities according to the clustering results.

[0123] Step 8) Take the core focus words, comment viewpoints, keywords of the location semantic relationship, and topic words formed in Step 2, Step 3, Step 4, and Step 5 as the recommended keywords of the geographical entities.

[0124] Step 9) Refer to AppendixFigure 10 , construct a geographical entity recommendation system that provides an interactive interface and offers functions for clicking and querying recommended keywords on the system interface. When a user clicks or queries a certain recommended keyword, the system automatically displays a list of geographical entities most relevant to the recommended keyword and visualizes them geospatially.

[0125] Step 9.1) When a user clicks or queries a certain recommended keyword, the system determines which category among the first to fourth geographical entity semantic networks A, B, C, and D it belongs to based on the recommended keyword.

[0126] Step 9.2) According to the type of geographical entity semantic network determined in Step 9.1, only use the recommended keyword clicked or queried by the user as the judgment condition for determining the complex network connection edges in the geographical entity semantic network.

[0127] Step 9.3) Generate the corresponding geographical entity semantic network according to the requirements of the geographical entity semantic network generation steps in Steps 2, 3, 4, and 5.

[0128] Step 9.4) Calculate various index values of the network nodes in the geographical entity semantic network generated in Step 9.3. According to the order of the size of a certain type of index value of the network nodes, screen the relevant geographical entities and complete the spatial visualization display of the geographical entities. Various index values of the network nodes include degree centrality, weighted proximity centrality, weighted degree centrality, weighted betweenness centrality, and eigenvector centrality index values.

[0129] The geographical entity recommendation system constructed in this step also includes the function of selecting different types of recommended keywords to achieve the recommendation of the most relevant geographical entities, that is, comprehensively recommend the most relevant geographical entities by matching multiple types of geographical entity semantic networks with multiple recommended keywords and calculating the cumulative sum of the index values of different types of geographical entity semantic networks.

[0130] In this step, different geographical entity semantic networks can be pre-generated according to the recommended keywords frequently selected by users, thereby improving the response speed of the geographical entity recommendation system.

[0131] Step 10) Refer to Appendix Figure 11 , the first to fourth geographical entity semantic networks (A - D) can be reconstructed separately according to the new comment data source or only construct the geographical entity semantic network for the new comment data source. Among them, the index results of the nodes of the newly constructed geographical entity semantic network can be superimposed and calculated with the index results of the nodes of the original same type of geographical entity semantic network to update the index results of the nodes of the geographical entity semantic network.

[0132] The algorithm steps in Step 3.3 include the following considerations: After replacing the comment views of each geographical entity with the combined attribute words and descriptive word combinations in Step 3.2.1 and Step 3.2.2, a judgment is made on the condition that a certain comment view in the comment view list appears in a certain comment of two different geographical entities respectively.

[0133] All geographical entities in each comment data extracted in Step 4.1 include the geographical entity to which this comment data belongs, regardless of whether this comment data includes the geographical entity to which this comment data belongs.

[0134] The various index values of the network nodes in Step 6 and Step 9.4 include the degree centrality, weighted proximity centrality, weighted degree centrality, weighted betweenness centrality, and eigenvector centrality index values of the network nodes.

[0135] The geographical entity recommendation system constructed in Step 9 also includes the function of realizing the recommendation of the most relevant geographical entity by selecting different types of recommendation keywords, that is, matching the semantic networks of multiple types of geographical entities through multiple recommendation keywords, and comprehensively recommending the most relevant geographical entity according to the cumulative calculation method of the index values of the semantic network nodes of different types of geographical entities. Among them, when calculating the cumulative sum of the index values of the semantic networks of different types of geographical entities, the weight values can be different. "Essentially, it is to let the user select multiple recommendation keywords, so that different types of geographical entity semantic networks can be associated according to the selected multiple recommendation keywords. Then, for the associated different types of geographical entity semantic networks, calculate the index values of the network nodes. The same type of network nodes in different types of geographical entity semantic networks can be set with different weights. The result after such cumulative calculation is used for comprehensive recommendation. And the weight values here can be set according to the frequencies used when different types of geographical entity semantic networks are selected by the user.

[0136] In Step 9, different geographical entity semantic networks are pre-generated according to the recommendation keywords frequently selected by the user, thereby improving the response speed of geographical entity recommendation.

[0137] The above content is a further detailed description of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is only limited to these descriptions. For those of ordinary skill in the technical field to which the present invention belongs, without departing from the concept of the present invention, several simple deductions or substitutions can still be made, and all should be regarded as belonging to the protection scope of the present invention.

Claims

1. A method for recommending geographical entities, characterized in that, it includes the following steps: Step 1, determine the comment data and geographical entity list required for the semantic network components of geographical entities in the study area; Step 2, extract the core focus words in each comment data, and combine with the geographical entity list to obtain the first geographical entity semantic network; Extract the comment viewpoints in each comment data, and combine with the geographical entity list to obtain the second geographical entity semantic network; extract the keywords of the location semantic relationship between geographical entities in each comment data, and combine with the geographical entity list to obtain the third geographical entity semantic network; Extract the topic words reflecting the comment theme in each comment data, and combine with the geographical entity list to obtain the fourth geographical entity semantic network; Step 3, use the core focus words, comment viewpoints, keywords of location semantic relationship, and topic words formed in Step 2 as the recommended keywords for geographical entities. According to the recommended keywords selected by the user, match the corresponding type of geographical entity semantic network, and generate the geographical entity semantic network. Furthermore, push the geographical entities most relevant to the user's interests through the newly generated geographical entity semantic network; The method for obtaining the first geographical entity semantic network in Step 2 is specifically as follows: Step 2.1, based on the TF—TDF algorithm, extract the keywords in each comment data, and perform word frequency statistics on the keywords extracted from all comment data; Step 2.2, establish a co-occurrence relationship network of keywords with word frequency greater than the preset threshold in the comment data, and calculate the intensity value of the core keywords of the co-occurrence relationship network; Step 2.3, screen out the core focus words according to the word frequency quantity size of the keywords in all comment data and the intensity value of the core keywords in the co-occurrence relationship network; Step 2.4, use geographical entities as the network nodes of the geographical entity semantic network. If a certain core focus word appears in the respective comment data of two different geographical entities, it is considered that there is a complex network connection edge between these two geographical entities; Step 2.5, based on the complex network theory, construct the first geographical entity semantic network through the relationship between network nodes and complex network connection edges; The method for obtaining the second geographical entity semantic network in Step 2 is specifically as follows: Step 3.1, based on the online API of Baidu natural language processing, extract the comment viewpoints in each comment data. Each comment viewpoint is composed of an attribute word and a descriptive word; Step 3.2, further refine and summarize the comment viewpoint list according to the combination characteristics of the attribute words and descriptive words of the comment viewpoints; Step 3.3, based on the complex network theory, use geographical entities as the network nodes of the geographical entity semantic network. If a certain comment viewpoint in the comment viewpoint list appears in the respective comment data of two different geographical entities, it is considered that there is a complex network connection edge between these two geographical entities; Step 3.4, based on the complex network theory, construct the second geographical entity semantic network through the relationship between network nodes and complex network connection edges; The method for obtaining the third geographical entity semantic network in Step 2 is specifically as follows: Step 4.1, extract all geographical entities in each comment data; Step 4.2: Construct keywords representing location semantic relationships and their corresponding weight values I, where the weight value I > 1; Step 4.3: Based on natural language semantic analysis technology, extract the location semantic connection relationships between all pairs of geographical entities in each comment data, that is: If there are no keywords representing location semantic relationships as described in Step 4.2 between two geographical entities in a comment data, it is considered that there is a complex network connection edge between these two geographical entities, and the weight of the edge is 1; If there are keywords representing location semantic relationships as described in Step 4.2 between two geographical entities in a comment data, it is considered that there is a complex network connection edge between these two geographical entities, and the weight of the edge is the weight value corresponding to the keyword representing the location semantic relationship as described in Step 4.2; among them, if there are multiple keywords representing location semantic relationships as described in Step 4.2, the keyword with the maximum weight value is used to determine the network edge connection relationship; Step 4.4: Based on complex network theory, use geographical entities as network nodes of the geographical entity semantic network, and then construct the third geographical entity semantic network through the relationships between the complex network connection edges determined in Step 4.3; The method for obtaining the fourth geographical entity semantic network in Step 2 is specifically as follows: Step 5.1: Based on the online API of Baidu natural language processing, extract the topic words of each comment data; Step 5.2: Summarize the topic lists of all comment data according to the topic words; Step 5.3: Based on complex network theory, use geographical entities as network nodes of the geographical entity semantic network. If a certain topic in the topic list appears in the respective comment data of two different geographical entities, it is considered that there is a complex network connection edge between these two geographical entities; Step 5.4: Based on complex network theory, construct the fourth geographical entity semantic network through the relationships between network nodes and complex network connection edges.

2. A geographical entity recommendation method according to claim 1, wherein, each geographical entity in the geographical entity list in Step 1 has longitude and latitude coordinate values, and the geographical entity list is obtained through the following method: Step 1.1: Each geographical entity in the study area has a unique number and name, and capture the comment data of each geographical entity; Step 1.2: Map each captured comment data to the geographical entity to which it belongs and establish a one-to-one correspondence.

3. A geographical entity recommendation method according to claim 1, wherein, After generating the first to fourth geographical entity semantic networks, calculate various index values of network nodes for each geographical entity semantic network respectively; Based on the obtained various index values of network nodes, conduct geographical entity clustering analysis, and perform spatial visualization display on geographical entities according to the clustering results.

4. A geographical entity recommendation method according to claim 1, wherein, The specific steps for calculating the intensity value of the core keyword in Step 2.2 specifically include: Step 2.2.1: Based on complex network theory, for all comment data, using the core keywords as network nodes, if two core keywords appear in the same comment, it is considered that there is a complex network connection edge between these two core keywords, and thus a keyword complex network is established. Step 2.2.2: For the keyword complex network established in Step 2.2.1, calculate the degree centrality, weighted proximity centrality, weighted degree centrality, weighted betweenness centrality, and eigenvector centrality index values of the network nodes. Step 2.2.3: Based on the entropy value method, calculate the comprehensive evaluation value of the network nodes for all the indicators obtained in Step 2.2.2; this comprehensive evaluation value is the intensity value of the core keywords. The specific calculation steps for refining and summarizing the comment view list in Step 3.2 are as follows: Step 3.2.1: Merge the attribute words with similar semantics in the comment views. Step 3.2.2: Merge the descriptive words with similar semantics in the comment views. Step 3.2.3: Based on the results of the merging processes in Steps 3.2.1 and 3.2.2, remove the duplicate comment views with the same attribute words and descriptive words in the comment views, and thus obtain the final comment view list.

5. A geographical entity recommendation method according to claim 1, wherein, In Step 4.1, extract all geographical entities in each comment data, including the geographical entity to which the comment data belongs, regardless of whether the comment data includes the geographical entity to which the comment data belongs.

6. A geographical entity recommendation method according to claim 1, wherein, The method for updating the semantic network of the m-th geographical entity: Reconstruct the semantic networks of the first to fourth geographical entities respectively according to the new comment data, or only construct the semantic network of the m-th geographical entity for the new comment data; among them, the network node index results of the newly constructed semantic network of the m-th geographical entity are superimposed and calculated with the network node index results of the original semantic network of the m-th geographical entity to update the network node index results of the semantic network of the m-th geographical entity; m = 1, 2, 3, 4.

7. A geographical entity recommendation method for a geographical entity semantic network according to claim 1, wherein, Use the core focus words, comment views, keywords of location semantic relationships, and topic words formed in Step 2 as the recommended keywords for geographical entities; Construct a geographical entity recommendation system, which provides an interactive interface and provides click and query functions for the recommended keywords on the system interface. When the user clicks or queries a certain recommended keyword, the system automatically displays the list of geographical entities most relevant to the recommended keyword and visualizes it in the geographical space.

8. A geographical entity recommendation method for a geographical entity semantic network according to claim 7, wherein, The implementation method of the geographical entity recommendation system is as follows: Step A.1: When the user clicks or queries a certain recommended keyword, the geographical entity recommendation system determines which category among the semantic networks of the first to fourth geographical entities the recommended keyword belongs to. Step A.2: According to the type of the geographical entity semantic network determined in Step A.1, only use the user's click or query of a certain recommended keyword as the judgment condition for determining the complex network connection edges in a certain type of geographical entity semantic network determined in Step A.1; Step A.3: Generate the corresponding geographical entity semantic network according to the requirements of the generation steps of a certain type of geographical entity semantic network determined in Step A.1; Step A.4: Calculate various index values of the network nodes in the geographical entity semantic network generated in Step A.

3. According to the order of the values of any type of index of the network nodes, screen the geographical entities and complete the spatial visualization display of the geographical entities.

Images