A method and system for ranking analysis of multi-path recall based on semantic retrieval

Abstract

This invention provides a method and system for ranking and analyzing multi-path recall based on semantic retrieval, belonging to the field of big data service technology. The method includes: obtaining keywords input by a user; obtaining recall results for each path based on the keywords; obtaining the user's interest features; constructing a first sensitivity tree for the user based on the interest features; obtaining the user's identity features; obtaining a second sensitivity tree corresponding to the identity features; fusing the first and second sensitivity trees to obtain a comprehensive sensitivity tree for the user; and ranking the recall results based on the comprehensive sensitivity tree. This method and system, by combining the user's interest features and identity features within the system, ranks the results of multi-path recall, thus improving the retrieval accuracy and efficiency within the power system.

Description

Technical Field

[0001] This invention relates to the field of big data service technology, and specifically to a ranking analysis method and system based on semantic retrieval and multi-path recall. Background Technology

[0002] Multi-path retrieval is a common method used by search engines. It mainly involves semantic analysis based on keywords or phrases entered by users and recalling documents using data from multiple dimensions to obtain accurate search results.

[0003] Existing multi-channel recall methods mainly rely on big data analysis, which can achieve accurate feedback when the database system is large and the sample size is sufficient.

[0004] The power information and communication management system is a closed system within the power grid used to manage power grid data. When facing user searches, it lacks the necessary conditions (large data volume and sufficient sample size) upon which commonly used multi-channel recall methods rely, making these methods inadequate. Summary of the Invention

[0005] The purpose of this invention is to provide a ranking analysis method and system based on semantic retrieval for multi-path recall, which can adapt to the retrieval accuracy requirements within the power system.

[0006] To achieve the above objectives, embodiments of the present invention provide a ranking analysis method for multi-path recall based on semantic retrieval, comprising:

[0007] Get the keywords entered by the user;

[0008] The recall results for each path are obtained based on the keywords;

[0009] Obtain the user's interest characteristics;

[0010] Construct a first sensitivity tree for the user based on the aforementioned interest characteristics;

[0011] Obtain the user's identity characteristics;

[0012] Obtain the second sensitivity tree corresponding to the identity features;

[0013] The first and second sensitivity trees are merged to obtain the user's comprehensive sensitivity tree;

[0014] The recall results are sorted according to the comprehensive sensitivity tree.

[0015] Optionally, the recall results for each path are obtained based on the keywords, including:

[0016] The keywords are expanded to obtain multiple synonyms;

[0017] The database is matched and searched based on the keywords and synonyms to obtain the recall results;

[0018] The keywords and synonyms are encoded according to a preset word library to obtain the corresponding keyword codes;

[0019] The keyword encoding is input into a preset language model to obtain the corresponding vector representation;

[0020] The similarity query is performed in the database based on the vector representation to obtain the recall result.

[0021] Optionally, the user's interest characteristics are obtained, including:

[0022] Retrieve the user's historical query records;

[0023] The historical query records are categorized and statistically analyzed to determine the count value for each underlying index category corresponding to the historical query records;

[0024] Statistical analysis is performed on the underlying index categories to determine the concentration of the underlying index categories;

[0025] Determine whether the concentration is greater than or equal to a preset concentration threshold;

[0026] If the concentration is determined to be greater than or equal to the concentration threshold, the underlying index classification with a count value greater than the preset count value threshold is taken as the interest feature;

[0027] If the concentration is determined to be less than the concentration threshold, the historical query records are divided into multiple sub-records according to a preset time interval.

[0028] Iterate through each of the sub-records and determine the count value of the underlying index category included in each sub-record;

[0029] Each count value is updated according to its distance from the current time to obtain the comprehensive count value of the underlying index classification;

[0030] Sort the comprehensive count values ​​from largest to smallest, and select the top few bottom index categories with the largest comprehensive count values ​​as the interest features.

[0031] Optionally, statistics are performed on the underlying index categories to determine the centrality of the underlying index categories, including:

[0032] The concentration is calculated according to formula (1):

[0033]

[0034] Where C is the concentration, N is the number of underlying index categories in the historical query records, and x i The count value for the i-th underlying index category. This is the average count value.

[0035] Optionally, each count value is updated according to its distance from the current time to obtain a comprehensive count value for the underlying index classification, including:

[0036] Update the overall count value according to formula (2):

[0037]

[0038] Where, x i ′ Let x be the total count value, l be the distance, and x be the distance. il This is the count value of the i-th underlying index category with a distance of l.

[0039] Optionally, constructing a user's first sensitivity tree based on the interest features includes:

[0040] Clustering is performed on each underlying index category according to the index category above each underlying index category;

[0041] Determine whether the number of underlying index categories included in each category after the clustering operation is greater than or equal to the minimum threshold;

[0042] If it is determined that the number of the underlying index categories included in each category after the clustering operation is greater than or equal to the minimum threshold, the first sensitive tree is constructed according to the correspondence of each category after the clustering operation.

[0043] If it is determined that the number of underlying index categories included in any category after the clustering operation is greater than or equal to the minimum threshold, the clustering basis of the clustering operation is modified to the upper-level index category of the current clustering basis, the clustering operation is executed again, and the process returns to the step of determining whether the number of underlying index categories included in each category after the clustering operation is greater than or equal to the minimum threshold.

[0044] Optionally, obtaining the user's identity features includes:

[0045] Obtain the user's current identity information;

[0046] Based on the identity information, retrieve the commonly used access index category corresponding to the identity information from the database;

[0047] The commonly used access indexes are categorized as the identity features.

[0048] Optionally, the first and second sensitivity trees are fused to obtain the user's comprehensive sensitivity tree, including:

[0049] Determine the union of the first sensitive tree and the second sensitive tree;

[0050] Starting from the top level of the first sensitive tree and the second sensitive tree, determine whether each valid index category of the first sensitive tree and the second sensitive tree is the same;

[0051] If the valid index categories are the same, the weight of the corresponding index category in the union is increased by 1 to update the union;

[0052] The union is used as the comprehensive sensitivity tree.

[0053] Optionally, the recall results are sorted according to the comprehensive sensitivity tree, including:

[0054] Traverse the document corresponding to each of the recall results and find the corresponding position in the comprehensive sensitivity tree;

[0055] If the document can be found, determine whether the found location is a valid index category;

[0056] If the found location is a valid index category, the weight of the corresponding index category will be used as the ranking weight of the recall results.

[0057] If the found location is an invalid index category, the upper-level index category of the found location is used as the tag for the document, and the document is added to the list to be sorted;

[0058] Set the sorting weight of all documents in the unsorted list to the minimum sorting weight of documents in the effective index classification.

[0059] Update the sort weight of each document in the unsorted list based on the tags of all documents therein;

[0060] If the document cannot be found, determine whether the document is a single document;

[0061] If there is only one document, the ranking weight of the document is set to the minimum ranking weight of the documents that can be found.

[0062] When there are multiple documents, the maximum value of the ranking weight of the document is set to the minimum value of the ranking weight of the document that can be searched, and the remaining documents are updated according to the keyword relevance of the document that cannot be searched.

[0063] Each document is sorted according to the stated sorting weight.

[0064] On the other hand, the present invention also provides a ranking analysis system for multi-way recall based on semantic retrieval, the system including a processor configured to perform any of the methods described above.

[0065] Through the above technical solutions, the embodiments of the present invention provide a ranking analysis method and system for multi-path recall based on semantic retrieval. The method and system combine the user's interest characteristics and identity characteristics within the system to rank the results of multi-path recall, thereby improving the retrieval method based on multi-path recall and enhancing the retrieval accuracy and efficiency within the power system.

[0066] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0067] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings:

[0068] Figure 1 This is a flowchart of a multi-path recall ranking analysis method based on semantic retrieval according to an embodiment of the present invention;

[0069] Figure 2 This is a flowchart of a method for obtaining recall results according to an embodiment of the present invention;

[0070] Figure 3 This is a flowchart of a method for acquiring interest features according to an embodiment of the present invention;

[0071] Figure 4 This is a flowchart of a method for obtaining a first sensitivity tree according to an embodiment of the present invention;

[0072] Figure 5 This is an example diagram of a first sensitive tree and a second sensitive tree according to an embodiment of the present invention;

[0073] Figure 6 This is a flowchart of sorting recall results based on a comprehensive sensitivity tree according to one embodiment of the present invention. Detailed Implementation

[0074] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0075] It should be noted that the acquisition, transmission, storage, use, and processing of data in the technical solution of this application all comply with the relevant provisions of national laws and regulations. In the embodiments of this application, certain existing industry solutions such as software, components, and models may be mentioned. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solution of this application, and do not imply that the applicant has already used or necessarily used such solutions.

[0076] like Figure 1 The diagram shows a flowchart of a multi-path recall ranking analysis method based on semantic retrieval according to an embodiment of the present invention. Figure 1 In this method, the steps may include:

[0077] In step S10, the keywords input by the user are obtained;

[0078] In step S11, the recall results for each path are obtained based on the keywords;

[0079] In step S12, the user's interest characteristics are obtained;

[0080] In step S13, the user's first sensitivity tree is constructed based on interest features;

[0081] In step S14, the user's identity features are obtained;

[0082] In step S15, the second sensitive tree corresponding to the identity features is obtained;

[0083] In step S16, the first and second sensitive trees are merged to obtain the user's comprehensive sensitive tree;

[0084] In step S17, the recall results are sorted according to the comprehensive sensitivity tree.

[0085] In such Figure 1 In the method shown, step S10 can obtain keywords input by the user. These keywords can be single keywords directly entered by the user in the search component or extracted from keyword phrases containing keywords. Furthermore, if the user input is a keyword phrase containing keywords, step S10 can also first perform semantic analysis on the keyword phrase to determine the keywords within the keyword phrase based on contextual relevance.

[0086] Step S11 can be used to obtain the recall results for each path based on keywords. The method for obtaining the recall results for each path can be of various forms known to those skilled in the art. In one example of the present invention, the method for obtaining the recall results may include, for example... Figure 2 The method shown. In this Figure 2 In this context, the method for obtaining the recall result may include the following steps:

[0087] In step S20, the keywords are expanded to obtain multiple synonyms;

[0088] In step S21, a matching search is performed in the database based on keywords and synonyms to obtain recall results;

[0089] In step S22, the keywords and synonyms are encoded according to a preset word library to obtain the corresponding keyword codes;

[0090] In step S23, the keyword encoding is input into a preset language model to obtain the corresponding vector representation;

[0091] In step S24, a similarity query is performed in the database based on the vector representation to obtain the recall results.

[0092] Step S12 can be used to obtain the user's interest features. These interest features can be used to represent the index categories the user is interested in. The specific method for obtaining these interest features can be of various forms known to those skilled in the art, including but not limited to accessing the user's favorites records, historical access records, etc. In one example of the present invention, the method for obtaining these interest features may include, for example... Figure 3 The steps shown are described. Figure 3 In this context, the method for obtaining this interest feature may include the following steps:

[0093] In step S30, the user's historical query records are retrieved;

[0094] In step S31, the historical query records are classified and statistically analyzed to determine the count value of each underlying index category corresponding to the historical query records;

[0095] In step S32, statistics are performed on the underlying index categories to determine the concentration of the underlying index categories;

[0096] In step S33, it is determined whether the concentration is greater than or equal to a preset concentration threshold;

[0097] In step S34, if the concentration is determined to be greater than or equal to the concentration threshold, the underlying index classification with a count value greater than the preset count value threshold is used as the interest feature.

[0098] In step S35, if the concentration is determined to be less than the concentration threshold, the historical query records are divided into multiple sub-records according to a preset time interval.

[0099] In step S36, each sub-record is traversed to determine the count value of the underlying index category included in each sub-record;

[0100] In step S37, each count value is updated according to its distance from the current time to obtain the comprehensive count value of the underlying index classification;

[0101] In step S38, the top few bottom index categories with larger comprehensive count values ​​are selected as the features of interest, sorted in descending order of comprehensive count values.

[0102] In such Figure 3 In the method shown, step S30 can be used to obtain the user's historical query records. These historical query records may include multiple query records, each of which is represented as the last link accessed by the user after entering a single keyword. Therefore, in one example of the present invention, when obtaining the user's historical query records, step S30 cannot directly crawl all of the user's historical access records in the server. Instead, it searches for each time the user enters a keyword and selects the last link accessed as a query record in the historical query records.

[0103] Step S31 can be used to classify and statistically analyze historical query records to determine the count value of each underlying index category corresponding to the historical query record. Specifically, in this embodiment, step S31 can be used to count the underlying index categories according to the position of the document corresponding to the link of each query record in the index.

[0104] Step S32 can be used to perform statistics on the underlying index classification to determine the concentration of the underlying index classification. The calculation of this concentration can take various forms known to those skilled in the art. In one example of the present invention, considering that concentration is a parameter used to measure the dispersion of the count values ​​of the underlying index classification, the concentration can be calculated using the following formula (1):

[0105]

[0106] Where C is the concentration, N is the number of underlying index categories in the historical query records, and x i The count value for the i-th underlying index category. This is the average count value.

[0107] Step S33 can be used to determine whether the concentration is greater than or equal to a preset concentration threshold. If the concentration is greater than the concentration threshold, it means that the user consistently accesses a portion of the underlying index categories regardless of the time. Therefore, the underlying index categories with count values ​​greater than or equal to the count value threshold can be directly used as interest features, i.e., step S34.

[0108] Conversely, this indicates that the user's access is relatively dispersed. Since a more dispersed underlying index classification is not conducive to subsequent sorting and filtering, further processing based on the time series is required through steps 35 to S38. Specifically, in this embodiment, in step S35, if the concentration is less than the concentration threshold, the historical query records are divided into multiple sub-records according to a preset time interval. Step S36 is used to traverse each sub-record and determine the count value of the underlying index classification included in each sub-record, thereby determining the user's access habits in different time periods. Considering that the further away a user's sub-record is from the current time, the less effective it is in reflecting the user's access habits, the comprehensive count value cannot be directly represented by simply adding the count values ​​of each sub-record. In one example of the present invention, considering the influence of time distance, the comprehensive count value can be calculated using the following formula (2):

[0109]

[0110] Where, x i ′ This is the combined count value, where l is the distance and x is the total count value. il This is the count value of the i-th underlying index category with a distance of l.

[0111] Step S38 is used to sort the underlying index categories in descending order of their comprehensive count values ​​and select the top few categories with larger comprehensive count values ​​as the features of interest. The method for determining the number of underlying index categories to be selected can be various forms known to those skilled in the art, such as by pre-setting the number. In one example of the present invention, since it is necessary to select the most representative categories, but not all underlying index categories can be selected, the difference in count values ​​between adjacent underlying index categories after sorting is predetermined. Then, the selection is performed from left to right. When the difference exceeds a preset threshold, the selection can be stopped, the underlying index categories after the difference are discarded, and the underlying index categories before the difference are selected as the features of interest.

[0112] Step S13 is used to construct a first sensitivity tree for the user based on interest features. This first sensitivity tree can represent the distribution characteristics of the user's index classification based on interest features. The specific method for obtaining this first sensitivity tree can be of various forms known to those skilled in the art. In one example of the present invention, the method for obtaining the first sensitivity tree may include, for example... Figure 4 The steps shown are described. Figure 4 In this context, the method for obtaining the first sensitive tree may include the following steps:

[0113] In step S40, clustering is performed on each underlying index category according to the parent index category of each underlying index category;

[0114] In step S41, it is determined whether the number of underlying index categories included in each category after the clustering operation is greater than or equal to the minimum threshold.

[0115] In step S42, if it is determined that the number of underlying index categories included in each category after the clustering operation is greater than or equal to the minimum threshold, a first sensitive tree is constructed according to the correspondence of each category after the clustering operation.

[0116] In step S43, if it is determined that the number of underlying index categories included in any category after the clustering operation is greater than or equal to the minimum threshold, the clustering basis of the clustering operation is modified to the upper-level index category of the current clustering basis, the clustering operation is executed again, and the process returns to the step of determining whether the number of underlying index categories included in each category after the clustering operation is greater than or equal to the minimum threshold, that is, the process returns to step S41.

[0117] Step S14 can be used to obtain the user's identity features. These identity features represent the user's access habits regarding their identity information within the current system. Unlike interest features, since there are many users with the same identity information in the same system, the number of samples available is also large. In this case, it is sufficient to directly obtain the user's frequently accessed index classification as the identity feature.

[0118] Step S15 can be used to obtain the second sensitive tree corresponding to the identity features. The method of obtaining the second sensitive tree can be similar to that of obtaining the first sensitive tree, and the method of obtaining the first sensitive tree has been described in detail above, so it will not be repeated here.

[0119] Step S16 can be used to fuse the first sensitive tree and the second sensitive tree to obtain the user's comprehensive sensitive tree. Specifically, in one example of this invention, the example diagram of the first sensitive tree and the second sensitive tree can be as follows: Figure 5 As shown. In this Figure 5In the first and second sensitive trees, there are multiple index categories. The index category number is marked with an asterisk (*) to indicate that it is a valid index category, which means that the user is interested in that index category. Conversely, it is marked with an asterisk (*) to indicate that the user may be interested in the upper-level index category of that index category.

[0120] Step S16 can be as follows: First, compare the underlying index categories of each first sensitive tree and second sensitive tree to determine if there are any identical underlying index categories. Then, if identical underlying index categories are found, increment the weight of the corresponding upper-level index category by 1. Based on this example, in one example of the present invention, the method for obtaining the comprehensive sensitive tree can be as follows: First, determine the union of the first sensitive tree and the second sensitive tree. Then, starting from the top level of each of the first and second sensitive trees, determine whether each effective index category of the first and second sensitive trees is the same. If the effective index categories are the same, increment the weight of the corresponding index category in the union by 1 to update the union. Finally, output the union as the comprehensive sensitive tree.

[0121] Finally, step S17 can be used to sort the recall results according to the comprehensive sensitivity tree. Specifically, in one example of the present invention, the method for sorting the recall results according to the comprehensive sensitivity tree may include, for example... Figure 6 The steps shown are described. Figure 6 In this context, the method for ranking the recall results based on the comprehensive sensitivity tree may include the following steps:

[0122] In step S50, the document corresponding to each recall result is traversed, and the corresponding position is found in the comprehensive sensitivity tree;

[0123] In step S51, if the document can be found, it is determined whether the found location is a valid index category;

[0124] In step S52, if the found location is a valid index category, the weight of the corresponding index category is used as the ranking weight of the recall results.

[0125] In step S53, if the found position is an invalid index category, the upper-level index category of the found position is used as the document's tag, and the document is added to the list to be sorted;

[0126] In step S54, the sorting weight of all documents in the list to be sorted is set to the minimum sorting weight of documents in the effective index classification.

[0127] In step S55, the ranking weight of each document in the list to be ranked is updated based on the tags of all documents in the list. Specifically, in this example, the ranking weight can be updated using the following formula (3):

[0128]

[0129] Where, λ i ′ For the updated sorting weights, λ i c represents the sorting weights before the update. i Let be the tag of the i-th document, and s be the number of items in the list to be sorted.

[0130] In step S56, if the document cannot be found, it is determined whether the document is a single document;

[0131] In step S57, if there is only one document, the ranking weight of the document is set to the minimum ranking weight of the documents that can be found.

[0132] In step S58, when there are multiple documents, the maximum ranking weight of a document is set to the minimum ranking weight of the documents that can be retrieved, and the remaining documents are updated based on the keyword relevance of the documents that cannot be retrieved. Specifically, the keyword relevance can be the ranking order of the recall results based on the keyword search. To ensure that the documents in this part can meet this order during sorting, the ranking weight of the documents can be set to decrease by a fixed value starting from the maximum value. This fixed value can be 0.1 times the maximum value.

[0133] In step S59, each document is sorted according to its sorting weight.

[0134] On the other hand, the present invention also provides a ranking analysis system for multi-way recall based on semantic retrieval, the system including a processor configured to perform any of the methods described above.

[0135] Through the above technical solutions, the embodiments of the present invention provide a ranking analysis method and system for multi-path recall based on semantic retrieval. The method and system combine the user's interest characteristics and identity characteristics within the system to rank the results of multi-path recall, thereby improving the retrieval method based on multi-path recall and enhancing the retrieval accuracy and efficiency within the power system.

[0136] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0137] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0138] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0139] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0140] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0141] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0142] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0143] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0144] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A ranking analysis method based on semantic retrieval multi-path recall, characterized in that, The method includes: Get the keywords entered by the user; The recall results for each path are obtained based on the keywords; Obtain the user's interest characteristics; Construct a first sensitivity tree for the user based on the aforementioned interest characteristics; Obtain the user's identity characteristics; Obtain the second sensitivity tree corresponding to the identity features; The first and second sensitivity trees are merged to obtain the user's comprehensive sensitivity tree; The recall results are sorted according to the comprehensive sensitivity tree; Obtaining the user's interest characteristics includes: Retrieve the user's historical query records; The historical query records are categorized and statistically analyzed to determine the count value for each underlying index category corresponding to the historical query records; Statistical analysis is performed on the underlying index categories to determine the concentration of the underlying index categories; Determine whether the concentration is greater than or equal to a preset concentration threshold; If the concentration is determined to be greater than or equal to the concentration threshold, the underlying index classification with a count value greater than the preset count value threshold is taken as the interest feature; If the concentration is determined to be less than the concentration threshold, the historical query records are divided into multiple sub-records according to a preset time interval. Iterate through each of the sub-records and determine the count value of the underlying index category included in each sub-record; Each count value is updated according to its distance from the current time to obtain the comprehensive count value of the underlying index classification; Sort the comprehensive count values ​​from largest to smallest, and select the top few bottom index categories with the largest comprehensive count values ​​as the interest features; Constructing a user's first sensitivity tree based on the aforementioned interest features includes: Clustering is performed on each underlying index category according to the index category above each underlying index category; Determine whether the number of underlying index categories included in each category after the clustering operation is greater than or equal to the minimum threshold; If it is determined that the number of the underlying index categories included in each category after the clustering operation is greater than or equal to the minimum threshold, the first sensitive tree is constructed according to the correspondence of each category after the clustering operation. If it is determined that the number of underlying index categories included in any category after the clustering operation is greater than or equal to the minimum threshold, the clustering basis of the clustering operation is modified to the upper-level index category of the current clustering basis, the clustering operation is executed again, and the process returns to the step of determining whether the number of underlying index categories included in each category after the clustering operation is greater than or equal to the minimum threshold.

2. The method according to claim 1, characterized in that, The recall results for each path are obtained based on the keywords, including: The keywords are expanded to obtain multiple synonyms; The database is matched and searched based on the keywords and synonyms to obtain the recall results; The keywords and synonyms are encoded according to a preset word library to obtain the corresponding keyword codes; The keyword encoding is input into a preset language model to obtain the corresponding vector representation; The similarity query is performed in the database based on the vector representation to obtain the recall result.

3. The method according to claim 1, characterized in that, Statistical analysis is performed on the underlying index categories to determine the concentration of the underlying index categories, including: The concentration is calculated according to formula (1): ，（1） in, The concentration, The number of underlying index categories in the historical query records. For the first The count value of each underlying index category. This is the average count value.

4. The method according to claim 1, characterized in that, Each count value is updated according to its distance from the current time to obtain a comprehensive count value for the underlying index classification, including: Update the overall count value according to formula (2): ，（2） in, The total count value is... For the distance, For the first In the lowest-level index categories, the distance is The count value.

5. The method according to claim 1, characterized in that, Obtaining the user's identity characteristics includes: Obtain the user's current identity information; Based on the identity information, retrieve the commonly used access index category corresponding to the identity information from the database; The commonly used access indexes are categorized as the identity features.

6. The method according to claim 1, characterized in that, The first and second sensitivity trees are fused to obtain the user's comprehensive sensitivity tree, including: Determine the union of the first sensitive tree and the second sensitive tree; Starting from the top level of the first sensitive tree and the second sensitive tree, determine whether each valid index category of the first sensitive tree and the second sensitive tree is the same; If the valid index categories are the same, the weight of the corresponding index category in the union is increased by 1 to update the union; The union is used as the comprehensive sensitivity tree.

7. The method according to claim 6, characterized in that, The recall results are sorted according to the comprehensive sensitivity tree, including: Traverse the document corresponding to each of the recall results and find the corresponding position in the comprehensive sensitivity tree; If the document can be found, determine whether the found location is a valid index category; If the found location is a valid index category, the weight of the corresponding index category will be used as the ranking weight of the recall results. If the found location is an invalid index category, the upper-level index category of the found location is used as the tag for the document, and the document is added to the list to be sorted; Set the sorting weight of all documents in the unsorted list to the minimum sorting weight of documents in the effective index classification. Update the sort weight of each document in the unsorted list based on the tags of all documents therein; If the document cannot be found, determine whether the document is a single document; If there is only one document, the ranking weight of the document is set to the minimum ranking weight of the documents that can be found. When there are multiple documents, the maximum value of the ranking weight of the document is set to the minimum value of the ranking weight of the document that can be searched, and the remaining documents are updated according to the keyword relevance of the document that cannot be searched. Each document is sorted according to the stated sorting weight.

8. A ranking analysis system based on semantic retrieval and multi-path recall, characterized in that, The system includes a processor configured to perform the method as described in any one of claims 1 to 7.

Images