Method and system for improving POI search in digital policing scenarios

Abstract

The present invention provides a method and system for improving POI search in a digital police scenario, including: obtaining user search information, wherein the user search information includes a query statement and an origin coordinate; introducing a police knowledge graph to segment the query statement to obtain a search term; using the distance between terms as a ranking factor, using a search engine to search for POIs for the search terms to obtain a search ranking result; and using the origin coordinate to adjust the search ranking result. The present invention optimizes the word segmenter, and by using the distance between terms to make the results returned to the user more consistent with people's ranking cognition, and also adds a spatial distance factor from the origin coordinate for the search of the POI geographic location to improve the spatial similarity of the search results of the present invention with the search content.

Description

Technical Field

[0001] The present invention relates to the field of digital policing, and in particular to a method and system for improving POI search in a digital policing scenario. Background Art

[0002] In the business scenario of digital policing, the entry and collection of a lot of information requires the assistance of corresponding POI (Point of Interest) information search to determine similar crime scenes, suspect locations, driving routes, etc. A better user experience can greatly improve the efficiency and accuracy of related entry and collection.

[0003] In the past, POI searches were based on full-text searches. By indexing the text content of POIs, such as addresses and names, users could obtain results from the index library through search statements, such as Figure 1 shown.

[0004] The whole solution is divided into two parts: index creation and index search. Index creation: extract information from relevant POI content, such as name, address, etc. Index search: parse the user's query request, search in the created index library, and return the corresponding query results.

[0005] A common solution is to store the name or address information in different fields of a database table, and then create a database index at the database level. Users can use standard query statements to perform precise or fuzzy information queries.

[0006] Another solution is to use an index engine to create a full-text index for text information, and then use a search engine to query, which is a fuzzy search, such as the traditional Lucene search engine.

[0007] The solution discussed here is based on the second method, which is to use full-text search and search engines to achieve search. Therefore, word segmentation tools must be used in both index creation and query parsing. In the index creation phase, the text content will be segmented and inverted indexes will be created. In the query phase, the query content will be segmented, and matching documents will be found according to different term indexes. Then, matching documents of different terms will be aggregated to obtain a final return result.

[0008] The search method based on the existing full-text index / search engine has the following disadvantages:

[0009] 1. The word segmentation tool itself is trained using Internet data, and does not strengthen POI-related information, resulting in unsatisfactory word segmentation results. In the business scenario of digital policing, there will be special terms, unique terms, desensitized information, etc. If it cannot be correctly identified, it will have a negative impact on both index creation and search.

[0010] 2. Ranking of search results. The more popular method now is to use the similarity between the query content and the document content to determine the corresponding ranking results. The ones with higher similarity scores are ranked first, and vice versa. The mainstream text similarity algorithm is BM25. However, due to the introduction of word segmentation, the original search content will be divided into multiple independent terms, and the similarity will also be divided into the similarity of multiple word segmentations. The similarity of word segmentation cannot fully express the similarity of the search content, and the factors of the original search content are not taken into account when merging, and the distance factor before the word segmentation appears in the same document is not taken into account, that is, the closer the word segmentation distance, the higher the similarity should be. As a result, there is a large gap between the search results and the actual desired results.

[0011] 3. When searching for POIs in special areas, their geographical location characteristics are not taken into consideration. However, digital policing is highly territorial. In business scenarios such as reporting an incident, it is usually necessary to adjust the ranking of the recommended search results based on the longitude and latitude of the adjacent base stations of the caller or the landmark POI described. Summary of the invention

[0012] In view of the defects in the prior art, the object of the present invention is to provide a method and system for improving POI search in a digital police scenario.

[0013] A method for improving POI search in a digital police scenario provided by the present invention includes:

[0014] Step S1: obtaining user search information, wherein the user search information includes a query statement and origin coordinates;

[0015] Step S2: introducing the police knowledge graph to segment the query statement to obtain search terms;

[0016] Step S3: using the distance between terms as a ranking factor, using a search engine to perform a POI search for the search terms, and obtaining a search ranking result;

[0017] Step S4: using the origin coordinates to adjust the search ranking results.

[0018] Preferably, in step S1:

[0019] The corresponding coordinates are obtained by using the mobile phone base station information obtained by the user when the alarm is sounded or the landmark location provided by the user, and these coordinates are used as the origin coordinates.

[0020] Preferably, in step S2:

[0021] The entries in the police knowledge graph include: entries of address parts of speech in documents extracted from POI data, police-specific entries, police professional entries, and regional POI entries extracted from text information in the police enhanced data source, and these entries are aggregated and saved in the police knowledge graph;

[0022] The police knowledge graph is used to optimize the Lucene tokenizer for token segmentation.

[0023] Preferably, in step S3:

[0024] Associating the search terms to obtain associated terms, if the associated terms appear in a document, the ranking of the document is improved; wherein the associated terms are standard address networks constructed based on the one-standard-three-real data of standard addresses, actual population, actual houses and actual units in the police knowledge graph;

[0025] Using a refined ranking layer in the search; wherein the refined ranking layer is used to calculate the similarity between the search terms and the target terms; the higher the similarity, the higher the search ranking;

[0026] For queries using a Chinese character abbreviation as the smallest unit, a single Chinese character or a single symbol is used as the distance between terms; the closer the distance between terms, the higher the search ranking.

[0027] Preferably, in step S4:

[0028] By adjusting the number of Geohash code bits to circle squares of different sizes, POI data points of different distances are obtained, and then the distance between the POI data point and the origin coordinates is used to adjust the search ranking result; the closer the distance, the higher the search ranking.

[0029] According to the present invention, a system for improving POI search in a digital police scenario includes:

[0030] Module M1: obtaining user search information, wherein the user search information includes a query statement and origin coordinates;

[0031] Module M2: introduce the police knowledge graph to segment the query statement to obtain search terms;

[0032] Module M3: using the distance between terms as a ranking factor, using a search engine to perform a POI search for the search terms, and obtaining a search ranking result;

[0033] Module M4: Using the origin coordinates, adjust the search ranking results.

[0034] Preferably, in the module M1:

[0035] The corresponding coordinates are obtained by using the mobile phone base station information obtained by the user when the alarm is sounded or the landmark location provided by the user, and these coordinates are used as the origin coordinates.

[0036] Preferably, in the module M2:

[0037] The entries in the police knowledge graph include: entries of address parts of speech in documents extracted from POI data, police-specific entries, police professional entries, and regional POI entries extracted from text information in the police enhanced data source, and these entries are aggregated and saved in the police knowledge graph;

[0038] The police knowledge graph is used to optimize the Lucene tokenizer for token segmentation.

[0039] Preferably, in the module M3:

[0040] Associating the search terms to obtain associated terms, if the associated terms appear in a document, the ranking of the document is improved; wherein the associated terms are standard address networks constructed based on the one-standard-three-real data of standard addresses, actual population, actual houses and actual units in the police knowledge graph;

[0041] Using a refined ranking layer in the search; wherein the refined ranking layer is used to calculate the similarity between the search terms and the target terms; the higher the similarity, the higher the search ranking;

[0042] For queries using a Chinese character abbreviation as the smallest unit, a single Chinese character or a single symbol is used as the distance between terms; the closer the distance between terms, the higher the search ranking.

[0043] Preferably, in the module M4:

[0044] By adjusting the number of Geohash code bits to circle squares of different sizes, POI data points of different distances are obtained, and then the distance between the POI data point and the origin coordinates is used to adjust the search ranking result; the closer the distance, the higher the search ranking.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] 1. First, in terms of word segmentation, the present invention uses the word segmentation tool LAC. The word segmentation results increase the recognition of parts of speech, which can better identify the address-related content in the POI data. The dedicated terms, special terms, and special information terms extracted from the police enhanced data source, and the police knowledge graph constructed with these contents can help us improve the recognition accuracy when using the Lucene word segmenter. At the same time, Lucene can use the same word segmenter when searching and when building an index word segmentation index. Therefore, when searching, it can also accurately identify the address dimension and special dimension word segmentation that users are most concerned about in the search content.

[0047] 2. Secondly, after obtaining the data through search, Lucene will calculate the similarity between the word segmentation and the document. The high version uses BM25 by default. Among the two assumptions of the premise of this formula, Assumption 2: Whether the word appears in the document is independent. Therefore, this method does not take into account the influence of the distance between word segmentations, the ranking of word segmentations, and the position of word segmentations in the document on the document similarity. The present invention uses the distance between terms and the position of word segmentations in the document as factors affecting the final search result ranking, so that the results returned to the user are more in line with people's ranking cognition. At the same time, the special terms and special terms provided in the digital police knowledge graph can also help us make associations during the retrieval process, so as to help us adjust the corresponding ranking order when associating term groups.

[0048] 3. In addition, the present invention also adds a spatial distance factor from the origin coordinates for the search of POI geographic locations. The origin coordinates are obtained in specific police business scenarios according to the coordinates of the mobile phone base station of the caller, or the description of surrounding key landmarks to determine and influence the ranking order of corresponding recommended information, so as to improve the spatial similarity between the search results of the present invention and the search content. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0050] Figure 1 It is a schematic diagram of the principle of full-text search in the prior art.

[0051] Figure 2 This is a schematic diagram of the principle of optimizing the word segmenter in the present invention.

[0052] Figure 3 It is a schematic diagram of the principle of comprehensively optimizing search ranking results in the present invention.

[0053] Figure 4 Schematic diagram of the principle of correcting search ranking results through Geohash. DETAILED DESCRIPTION

[0054] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0055] A method for improving POI search in a digital police scenario provided by the present invention includes:

[0056] Step S1: Obtain user search information, wherein the user search information includes query statements and origin coordinates; in step S1: use the mobile phone base station information obtained by the user when the alarm is triggered or the landmark location provided by the user to obtain the corresponding coordinates, and use these coordinates as the origin coordinates.

[0057] Step S2: Introduce the police knowledge graph to segment the query statement to obtain search terms; in step S2: the terms in the police knowledge graph include: extracting address terms in documents from POI data, extracting police-specific terms, police professional terms, and regional POI terms from text information in the police enhanced data source, and aggregating these terms and saving them in the police knowledge graph; using the police knowledge graph to optimize the Lucene word segmenter for word segmentation. Specifically, use a third-party word segmentation tool to extract address terms in documents from POI data, obtain text information from the police enhanced data source, and then use a third-party word segmentation tool to extract professional and unique terms specific to policing, aggregate the two terms and merge them and save them in the police knowledge graph, and in subsequent cycles, introduce the police knowledge graph during word segmentation to improve the word segmentation effect. Figure 2 As shown in the figure, the initialization information of the police knowledge graph includes digital police-specific terms, professional terms, and regional POI terms. These data need to be gathered through police experts, national standards, etc.

[0058] Step S3: Use the distance between terms as a ranking factor, use a search engine to perform a POI search for the search terms, and obtain a search ranking result. The present invention optimizes the similarity calculation in the search. After a single term obtains a similarity score, a weight coefficient of the physical distance between the segmented terms is added. The ranking adjustment rules between terms are as follows: For example, when searching for ABC, after segmentation, three terms A, B, and C are obtained. If the document contains the three terms A, B, and C; then the greater the distance value between the terms, the lower the ranking will be. For example, the minimum distance of ABC is 0, and the distance of ADBC ​​is 1, then it is larger than the distance of ABC, and ABC ranks higher than ADBC. And the closer ABC appears in the document, the higher the ranking. The order in which the terms appear will also affect the ranking, ABC>ACB>CAB>CBA. If the document does not contain all the three terms A, B, and C, then the more it contains, the higher the ranking will be.

[0059] In addition, in the step S3, the present invention associates the search terms to obtain associated terms, and if the associated terms appear in the document, the ranking of the document is improved; wherein the associated terms are standard address networks constructed by one standard and three real data based on standard addresses, actual population, actual houses and actual units in the police knowledge graph. Specifically, by using the police-related knowledge graph accumulated in the indexing stage, the ranking of the results of the associated phrases will also be adjusted accordingly. For example, when some special terms are associated, we will push up the ranking of the document. The police knowledge graph contains a standard address network constructed by one standard and three real data based on standard addresses, actual population, actual houses and actual units. The present invention will use the query terms that obtain a high similarity in the standard address network as associated terms, for example, the homophone "Tiantongyuan" is associated with "Tiantongyuan South District", "Tiantongyuan North District" and so on. The present invention will perform limited association expansion on some shorter terms.

[0060] At the same time, the present invention uses a refined ranking layer in the search; wherein, the refined ranking layer is used to calculate the similarity between the search terms and the target terms; the higher the similarity, the higher the search ranking. Specifically, the present invention adds a refined ranking layer in the search stage, and the purpose of accessing the refined ranking layer is to solve unreasonable problems in the score ranking based on the BM25 algorithm, such as the impact of the search term's position in the document on the ranking. The refined ranking layer uses the Euclidean distance formula two-dimensional algorithm to calculate the ranking score, and combines the edit distance algorithm to calculate the similarity between the search term and the target term. Specifically as follows:

[0061]

[0062] p is the Euclidean distance between point (x1, y1) and point (x2, y2);

[0063] Taking the left side of the document as the origin, that is, (x2, y2) = (0, 0), we get the formula:

[0064]

[0065] y=lenght(doc)-index(key)+Simi(key,tag)

[0066] P is the distance from the point (x, y) to the origin;

[0067] Doc means document;

[0068] length indicates the length of the document;

[0069] index indicates the position where the first character of the search term first appears in the document;

[0070] key represents the search term;

[0071] Simi represents the edit distance similarity algorithm;

[0072] Tag indicates the term starting from index and with the length of the search term;

[0073] The length of the document minus the search term position, plus the similarity between the search term and the tag term of the same length as the search term in the document cut from the first character of the search term. The similarity is calculated using the edit distance Levenshtein algorithm; x = index (key) is the starting position of the search term in the text. Substituting into the formula, we get:

[0074]

[0075] Then p is reversed. It should be mentioned here that the document collection data needs to be aligned.

[0076] In addition, for queries using a minimum unit of a Chinese character abbreviation, a single Chinese character or a single symbol is used as the distance between terms; the closer the distance between terms, the higher the search ranking.

[0077] Lucene is a mature full-text search engine toolkit that provides a complete query engine and index engine. However, Lucene's default similarity scoring algorithm is insufficient, resulting in a need to improve the user experience during the search process.

[0078] In view of the shortcomings of Lucene's default similarity scoring algorithm, the existing mainstream technology is to apply BM25 to Lucene to optimize the scoring calculation of query results. BM25 is an extension of the binary independence model BIM, which is an algorithm used to evaluate the relevance between search terms and documents. However, this mainstream algorithm only considers the frequency of query terms in the document, but ignores the distance characteristics of query terms in the document.

[0079] In this regard, "A Lucene ranking algorithm based on improved segmentation distance" published in the Journal of Shandong Normal University (Natural Science Edition) first preprocesses the query string and document data, and then, by identifying the positions of keywords and query terms in the document, it calculates the segmentation distance between the query terms and keywords, and then derives the similarity score between the query string and the entire document.

[0080] However, the shortcoming of "a lucene ranking algorithm based on improved word segmentation distance" is that it is not suitable for the short document characteristic scenario of POI. As a document, the document intent of POI information is not "ranked in descending order of strength", especially the address information. It cannot meet the needs of police scenarios because one of the particularities of police scenarios is that many scenarios require the use of a minimum unit of a Chinese character abbreviation for query, and the algorithm is based on the "number of word segments" as the distance between terms. This particularity leads to the problem of too low a hit rate if the algorithm is used in police scenarios. In view of this particularity, the present invention makes an improvement by using a single Chinese character or a single symbol as the distance between terms.

[0081] Step S4: Use the coordinates of the origin to adjust the search ranking results. In step S4: by adjusting the number of Geohash coding bits to circle squares of different sizes, POI data points of different distances are obtained, and then the distance between the POI data points and the origin coordinates is used to adjust the search ranking results; the closer the distance, the higher the search ranking. Specifically, the corresponding coordinates are obtained by using the mobile phone base station information obtained by the user when the alarm is issued or the clear landmark location provided by the user, and the coordinates are used as the coordinate origin. By adjusting the number of Geohash coding bits to circle squares of different sizes, POI data points of different distances are obtained, and then the recommended ranking is adjusted by the distance from the coordinate origin. Most areas now use 4G base stations, and the effective radius in urban areas is generally 500m. The present invention selects 6 Geohash coding bits. Then a rectangular area of ​​1.2km*609m can be obtained, and the person who reports the alarm should theoretically fall within this rectangular area. In the police alarm service, the caller will be asked whether he is at the crime scene. If the answer is affirmative, we will first return the POI data points within the rectangular area, which will greatly improve the search efficiency and hit rate. If the caller is not at the crime scene, when he can provide a landmark, we will consider using the landmark as the origin coordinate to obtain the corresponding Geohash code to obtain the rectangular area as the condition for correcting the search results.

[0082] The present invention also provides a system for improving POI search in a digital police scenario. The system for improving POI search in a digital police scenario can be implemented by executing the step flow of the method for improving POI search in a digital police scenario, and those skilled in the art can understand the method for improving POI search in a digital police scenario as a specific implementation of the system for improving POI search in a digital police scenario.

[0083] According to the present invention, a system for improving POI search in a digital police scenario includes:

[0084] Module M1: obtaining user search information, wherein the user search information includes a query statement and origin coordinates;

[0085] Module M2: introduce the police knowledge graph to segment the query statement to obtain search terms;

[0086] Module M3: using the distance between terms as a ranking factor, using a search engine to perform a POI search for the search terms, and obtaining a search ranking result;

[0087] Module M4: Using the origin coordinates, adjust the search ranking results.

[0088] Preferably, in the module M1:

[0089] The corresponding coordinates are obtained by using the mobile phone base station information obtained by the user when the alarm is sounded or the landmark location provided by the user, and these coordinates are used as the origin coordinates.

[0090] Preferably, in the module M2:

[0091] The entries in the police knowledge graph include: entries of address parts of speech in documents extracted from POI data, police-specific entries, police professional entries, and regional POI entries extracted from text information in the police enhanced data source, and these entries are aggregated and saved in the police knowledge graph;

[0092] The police knowledge graph is used to optimize the Lucene tokenizer for token segmentation.

[0093] Preferably, in the module M3:

[0094] Associating the search terms to obtain associated terms, if the associated terms appear in a document, the ranking of the document is improved; wherein the associated terms are standard address networks constructed based on the one-standard-three-real data of standard addresses, actual population, actual houses and actual units in the police knowledge graph;

[0095] Using a refined ranking layer in the search; wherein the refined ranking layer is used to calculate the similarity between the search terms and the target terms; the higher the similarity, the higher the search ranking;

[0096] For queries using a Chinese character abbreviation as the smallest unit, a single Chinese character or a single symbol is used as the distance between terms; the closer the distance between terms, the higher the search ranking.

[0097] Preferably, in the module M4:

[0098] By adjusting the number of Geohash code bits to circle squares of different sizes, POI data points of different distances are obtained, and then the distance between the POI data point and the origin coordinates is used to adjust the search ranking result; the closer the distance, the higher the search ranking.

[0099] Those skilled in the art know that, in addition to implementing the system, device and its various modules provided by the present invention in a purely computer-readable program code, it is entirely possible to implement the same program in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers and embedded microcontrollers by logically programming the method steps. Therefore, the system, device and its various modules provided by the present invention can be considered as a hardware component, and the modules included therein for implementing various programs can also be considered as structures within the hardware component; the modules for implementing various functions can also be considered as both software programs for implementing the method and structures within the hardware component.

[0100] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A method for improving POI search in a digital police scenario, characterized in that: include: Step S1: obtaining user search information, wherein the user search information includes a query statement and origin coordinates; Step S2: introducing the police knowledge graph to segment the query statement to obtain search terms; Step S3: using the distance between terms as a ranking factor, using a search engine to perform a POI search for the search terms, and obtaining a search ranking result; Step S4: using the origin coordinates to adjust the search ranking results; In step S4: By adjusting the number of Geohash code bits to circle squares of different sizes, POI data points of different distances are obtained, and then the distance between the POI data point and the origin coordinates is used to adjust the search ranking result; the closer the distance, the higher the search ranking; In specific police business scenarios, the ranking order of recommended information is determined and influenced according to the coordinates of the mobile phone base station of the caller, or by describing the surrounding key landmarks; Select 6 Geohash code bits to get a rectangular area; in the police alarm service, ask the caller whether he is at the crime scene. If a positive answer is obtained, first return the POI data points in this rectangular area; if the caller is not at the crime scene and can provide a landmark, use the landmark as the origin coordinate to obtain the corresponding Geohash code to get the rectangular area as the condition for correcting the search results.

2. The method for improving POI search in a digital police scenario according to claim 1, characterized in that: In step S1: The corresponding coordinates are obtained by using the mobile phone base station information obtained by the user when the alarm is sounded or the landmark location provided by the user, and these coordinates are used as the origin coordinates.

3. The method for improving POI search in a digital police scenario according to claim 1, characterized in that: In step S2: The entries in the police knowledge graph include: entries of address parts of speech in documents extracted from POI data, police-specific entries, police professional entries, and regional POI entries extracted from text information in the police enhanced data source, and these entries are aggregated and saved in the police knowledge graph; The police knowledge graph is used to optimize the Lucene tokenizer for token segmentation.

4. The method for improving POI search in a digital police scenario according to claim 1, characterized in that: In step S3: Associating the search terms to obtain associated terms, if the associated terms appear in a document, the ranking of the document is improved; wherein the associated terms are standard address networks constructed based on the one-standard-three-real data of standard addresses, actual population, actual houses and actual units in the police knowledge graph; Using a refined ranking layer in the search; wherein the refined ranking layer is used to calculate the similarity between the search terms and the target terms; the higher the similarity, the higher the search ranking; For queries using a Chinese character abbreviation as the smallest unit, a single Chinese character or a single symbol is used as the distance between terms; the closer the distance between terms, the higher the search ranking.

5. A system for improving POI search in digital policing scenarios, characterized in that: include: Module M1: obtaining user search information, wherein the user search information includes a query statement and origin coordinates; Module M2: introduce the police knowledge graph to segment the query statement to obtain search terms; Module M3: using the distance between terms as a ranking factor, using a search engine to perform a POI search for the search terms, and obtaining a search ranking result; Module M4: using the origin coordinates to adjust the search ranking results; In the module M4: By adjusting the number of Geohash code bits to circle squares of different sizes, POI data points of different distances are obtained, and then the distance between the POI data point and the origin coordinates is used to adjust the search ranking result; the closer the distance, the higher the search ranking; In specific police business scenarios, the ranking order of recommended information is determined and influenced according to the coordinates of the mobile phone base station of the caller, or by describing the surrounding key landmarks; Select 6 Geohash code bits to get a rectangular area; in the police alarm service, ask the caller whether he is at the crime scene. If a positive answer is obtained, first return the POI data points in this rectangular area; if the caller is not at the crime scene and can provide a landmark, use the landmark as the origin coordinate to obtain the corresponding Geohash code to get the rectangular area as the condition for correcting the search results.

6. The system for improving POI search in digital police scenarios according to claim 5, characterized in that: In the module M1: The corresponding coordinates are obtained by using the mobile phone base station information obtained by the user when the alarm is sounded or the landmark location provided by the user, and these coordinates are used as the origin coordinates.

7. The system for improving POI search in digital police scenarios according to claim 5, characterized in that: In the module M2: The entries in the police knowledge graph include: entries of address parts of speech in documents extracted from POI data, police-specific entries, police professional entries, and regional POI entries extracted from text information in the police enhanced data source, and these entries are aggregated and saved in the police knowledge graph; The police knowledge graph is used to optimize the Lucene tokenizer for token segmentation.

8. The system for improving POI search in digital police scenarios according to claim 5, characterized in that: In the module M3: Associating the search terms to obtain associated terms, if the associated terms appear in a document, the ranking of the document is improved; wherein the associated terms are standard address networks constructed based on the one-standard-three-real data of standard addresses, actual population, actual houses and actual units in the police knowledge graph; Using a refined ranking layer in the search; wherein the refined ranking layer is used to calculate the similarity between the search terms and the target terms; the higher the similarity, the higher the search ranking; For queries using a Chinese character abbreviation as the smallest unit, a single Chinese character or a single symbol is used as the distance between terms; the closer the distance between terms, the higher the search ranking.

Images