Enterprise search result sorting method, model training method, device and equipment

Abstract

The invention relates to an enterprise search result sorting method, a model training method, devices and equipment. The sorting method comprises the steps of obtaining a search word, and performing preliminary retrieval based on the search word to obtain a plurality of preliminary retrieval enterprises; processing the search term by utilizing a trained associated feature model to obtain an associated feature value of each piece of feature information associated with the search term; acquiring click rate data of each initially retrieved enterprise, and constructing a sorting model based on the click rate data and the associated feature values; scoring the initially retrieved enterprises by using the sorting model to obtain sorting scores corresponding to the initially retrieved enterprises; and sorting the initially retrieved enterprises based on the sorting scores according to a descending order mode. By adopting the method, the influence of the correlation degree and the click rate between the search terms and the feature information can be integrated in the sorted search results, so that the search results better meet the search requirements of users.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method for sorting enterprise search results, a model training method, a device and equipment. Background Art

[0002] In the field of information retrieval, there are two forms: general search and vertical search. In general search, we need to pay attention to features such as relevance, timeliness, and authority. Vertical search engines are professional searches for a certain industry. According to the user's specific search request, they deeply mine and integrate a certain type of specialized information in the website (page) library, and then return the results to the user in some form.

[0003] In the vertical field search of industrial and commercial enterprise information, there are differences from the general search method, such as paying more attention to the relevance of the enterprise to real-time hot spots, the authority of the enterprise and other characteristics.

[0004] However, traditional retrieval methods are too sticky. When the search keywords entered by the user are highly matched with the target document content, the search results tend to be documents containing all or most of the keywords, rather than considering the overall relevance of the document and the actual needs of the user. Summary of the invention

[0005] Based on this, it is necessary to provide a method for sorting enterprise search results, a model training method, an apparatus and a device that can take into account the actual search needs of users in response to the above technical problems.

[0006] In a first aspect, the present application provides a method for sorting enterprise search results, the method comprising:

[0007] Obtaining a search term, and performing a preliminary search based on the search term to obtain a plurality of preliminary searched enterprises;

[0008] Processing the search terms using the trained association feature model to obtain association feature values ​​of each feature information associated with the search terms;

[0009] Obtaining click-through rate data of each of the initially searched companies, and constructing a ranking model based on the click-through rate data and the associated feature values;

[0010] Scoring each of the initially searched enterprises using the ranking model to obtain a ranking score corresponding to each of the initially searched enterprises;

[0011] The initially searched enterprises are sorted in descending order based on the sorting scores.

[0012] In one embodiment, the obtaining of the search term and performing a preliminary search based on the search term to obtain a plurality of preliminary searched enterprises include:

[0013] Obtaining a search term, and performing word segmentation processing on the search term through a word segmentation model to obtain a plurality of keywords corresponding to the search term;

[0014] Input the plurality of keywords into a preset search engine based on vertical field information, and obtain a plurality of initial search companies through the preset search engine.

[0015] In one embodiment, the step of using the ranking model to score each of the initially searched enterprises to obtain a ranking score corresponding to each of the initially searched enterprises includes:

[0016] Get the top-k documents corresponding to each initially searched enterprise;

[0017] The ranking model is used to calculate each top-k document to obtain a ranking score corresponding to each of the initially searched companies.

[0018] In one embodiment, the characteristic information includes at least one of the company name, the field in which the company is located, the business scope, financial information, shareholder information, whether it is a listed company, and whether there is an information disclosure platform.

[0019] In one of the embodiments, the click-through rate data is determined based on the number of clicks and impressions of each of the initially searched businesses.

[0020] In a second aspect, the present application also provides a method for training an association feature model, the method comprising the following steps:

[0021] Acquire historical search records; the historical search records include historical search terms and historical search results corresponding to the historical search terms, and the historical search results include multiple historical search companies;

[0022] Acquire multiple feature information of different dimensions of the top N enterprises in the historical search results; N is an integer greater than or equal to 2;

[0023] Calculating the correlation degree between the historical search terms and each of the feature information using a correlation degree algorithm to obtain a correlation feature value corresponding to the historical search terms and each of the feature information;

[0024] Based on each of the associated feature values, training is performed using a linear model training function to obtain a trained associated feature model.

[0025] In a third aspect, the present application further provides a device for sorting enterprise search results, the device comprising:

[0026] A preliminary search module, used to obtain search terms, and perform a preliminary search based on the search terms to obtain a plurality of preliminary searched enterprises;

[0027] A feature value determination module, used to process the search term using the trained associated feature model to obtain associated feature values ​​of each feature information associated with the search term;

[0028] A ranking model building module, used to obtain click-through rate data of each of the initially searched enterprises, and build a ranking model based on the click-through rate data and the associated feature value;

[0029] A scoring module, used to score each of the initially searched enterprises using the ranking model to obtain a ranking score corresponding to each of the initially searched enterprises;

[0030] The sorting module is used to sort each of the initially searched enterprises in descending order based on the sorting scores.

[0031] In a fourth aspect, the present application further provides a training device for an association feature model, the device comprising:

[0032] A history acquisition module, used to acquire historical search records; the historical search records include historical search terms and historical search results corresponding to the historical search terms, and the historical search results include multiple historical search companies;

[0033] A feature information acquisition module, used to acquire multiple feature information of different dimensions of the top N enterprises in the historical search results; N is an integer greater than or equal to 2;

[0034] A correlation degree determination module, used to calculate the correlation degree between the historical search terms and each of the feature information using a correlation degree algorithm, and obtain a correlation feature value corresponding to the historical search terms and each of the feature information;

[0035] The training module is used to perform training based on each of the associated feature values ​​using a linear model training function to obtain a trained associated feature model.

[0036] In a fifth aspect, the present application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the steps of the above-mentioned enterprise search result ranking method or the associated feature model training method.

[0037] In a sixth aspect, the present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned enterprise search result ranking method or the associated feature model training method.

[0038] In a seventh aspect, the present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the above-mentioned enterprise search result ranking method, or the training method of the associated feature model.

[0039] In the above-mentioned enterprise search result sorting method, a preliminary search is first performed based on the search terms input by the user to obtain multiple initial search enterprises, and then the search terms are processed using the trained associated feature model to obtain associated feature values. After obtaining the click-through rate of each initial search enterprise, a sorting model is constructed based on the click-through rate data and the associated feature value, and each initial search enterprise is scored through the sorting model. The sorting is directly performed according to the ranking score of each enterprise to obtain the sorted search results. In the sorted search results, the correlation between the search terms and the feature information and the influence of the click-through rate are combined, so that the sorted search results are more in line with the user's search needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments or the related technical descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0041] Figure 1 is an application environment diagram in one embodiment;

[0042] Figure 2 A schematic diagram of a process of sorting enterprise search results in one embodiment;

[0043] Figure 3 A flowchart of a method for sorting enterprise search results in another embodiment;

[0044] Figure 4 A schematic diagram of a flow chart of a method for training an associated feature model in one embodiment;

[0045] Figure 5 It is a structural block diagram of a device for sorting enterprise search results in one embodiment;

[0046] Figure 6 is a structural block diagram of a training device for an associated feature model in one embodiment;

[0047] Figure 7 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0049] The enterprise search result sorting method and the associated feature model training method provided in the embodiments of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. The terminal 102 can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented with an independent server or a server cluster consisting of multiple servers.

[0050] The BM25 (Best Matching 25) algorithm is a weighted algorithm for information retrieval, widely used in search engines and recommendation systems. It is based on a probabilistic retrieval model and is designed to calculate the relevance score between documents and queries. However, for vertical fields, such as enterprise search, the NM25 algorithm has some defects, including excessive stickiness: when the search keywords entered by the user are highly matched with the target document content (that is, the keyword stickiness is high), the BM25 algorithm may overemphasize the keyword frequency and document length, resulting in search results that tend to contain all or most of the keywords, rather than considering the overall relevance of the document and the actual needs of the user; few keywords and many hits: when there are few keywords, a large number of documents may contain these keywords. At this time, relying solely on the BM25 algorithm is likely to result in a large number of documents with high similarity, and it is impossible to accurately distinguish which documents are more valuable to the user's query; lack of semantic understanding: BM25 does not consider the contextual association of word meanings and potential semantic information, which may cause more relevant results that are not completely consistent with the keywords to be buried in some scenarios.

[0051] In an exemplary embodiment, Figure 2 As shown, a method for sorting enterprise search results is provided, and the method is applied to Figure 1 The server 104 in the example is used as an example to illustrate the method, which includes the following steps 202 to 206. Among them:

[0052] Step 202, obtaining a search term, and performing a preliminary search based on the search term to obtain a plurality of preliminary searched enterprises.

[0053] Exemplarily, the search term is a search term input by a user. Optionally, the preliminary search method can be based on keyword matching, semantic analysis or user behavior.

[0054] Keyword matching refers to the search engine using algorithms to match the search terms entered by users with the keywords in the webpage content to provide relevant results. Semantic analysis refers to understanding the intent and context of the search terms to provide more relevant results, rather than just word-for-word matching. User behavior-based refers to considering the user's click behavior, search history and preferences to optimize the relevance of the results.

[0055] Step 204: Process the search terms using the trained association feature model to obtain association feature values ​​of each feature information associated with the search terms.

[0056] For example, the associated feature model may be a linear model, which is obtained based on a linear model training algorithm. The associated feature value can reflect the degree of association between the search term and the feature information.

[0057] Exemplarily, the characteristic information may be characteristics of different dimensions of an enterprise. Optionally, the different dimensions of an enterprise include business scope, annual revenue, major shareholders, shareholding ratio, shareholder background, listing status (whether listed), business model, market share, main products or services, intellectual property rights, litigation record, target market, major customers, market positioning, development strategy, long-term goals, investment plans, etc.

[0058] Furthermore, the characteristic information includes at least one of the company name, the field in which the company is located, the business scope, financial information, shareholder information, whether it is a listed company, and whether there is an information disclosure platform.

[0059] Exemplarily, the feature information is cleaned, standardized or encoded to facilitate computational processing.

[0060] Step 206, obtaining click-through rate data of each initially searched enterprise, and constructing a ranking model based on the click-through rate data and associated feature values.

[0061] For example, click-through rate data is an indicator used to measure the effectiveness of advertisements, links or search results in attracting users to click on them, indicating the proportion of users who click on a certain link or advertisement.

[0062] The sorting model may be a support vector machine (SVM) model, which can generate a corresponding SVM feature vector for each enterprise document.

[0063] SVM is a supervised learning model, mainly used for classification and regression analysis tasks. In classification problems, the basic idea of ​​SVM is to find an optimal hyperplane (for linearly separable data sets) or nonlinear decision boundary (after mapping the input space to a high-dimensional feature space through a kernel function), which can maximize the interval between its nearest sample points (i.e., support vectors), thereby achieving the maximum ability to distinguish different categories of data.

[0064] Step 208, using the ranking model to score each of the initially searched enterprises, and obtaining a ranking score corresponding to each of the initially searched enterprises.

[0065] The score reflects the degree of association between the search terms and each feature information, and is corrected by click-through rate data. The ranking score reflects the feature information of interest to users of each initially searched enterprise.

[0066] Step 210, sorting the initially searched enterprises in descending order based on the sorting scores.

[0067] Exemplarily, after obtaining the descending order results of each initially searched enterprise, the sorted results are finally returned to the user to ensure that the most relevant enterprise is displayed first in the search results.

[0068] In the above-mentioned enterprise search result sorting method, a preliminary search is first performed based on the search terms input by the user to obtain multiple initial search enterprises, and then the search terms are processed using the trained associated feature model to obtain associated feature values. After obtaining the click-through rate of each initial search enterprise, a sorting model is constructed based on the click-through rate data and the associated feature value, and each initial search enterprise is scored through the sorting model. The sorting is directly performed according to the ranking score of each enterprise to obtain the sorted search results. In the sorted search results, the correlation between the search terms and the feature information and the influence of the click-through rate are combined, so that the sorted search results are more in line with the user's search needs.

[0069] In an exemplary embodiment, a search term is obtained, and a preliminary search is performed based on the search term to obtain a plurality of initial search enterprises, including: obtaining the search term, segmenting the search term through a word segmentation model to obtain a plurality of keywords corresponding to the search term; inputting the plurality of keywords into a preset search engine based on vertical field information, and obtaining a plurality of initial search enterprises through the preset search engine.

[0070] Exemplarily, the word segmentation model is used to segment continuous character strings into meaningful keywords, and use the keywords as index terms input into a search database.

[0071] Exemplarily, the preset search engine may be a business information search engine.

[0072] In this embodiment, the search terms input by the user are first segmented to obtain keywords, and then the keywords are input into the vertical search engine of business information, which can avoid the influence of irrelevant or weakly relevant information and improve the search accuracy.

[0073] In one of the embodiments, a ranking model is used to score each initially searched enterprise to obtain a ranking score corresponding to each initially searched enterprise, including: obtaining the top-k documents corresponding to each initially searched enterprise; and calculating each top-k document using the ranking model to obtain a ranking score corresponding to each initially searched enterprise.

[0074] Exemplarily, among the multiple initially searched enterprises obtained after the initial search, a corresponding top-k document is generated for each enterprise, and the top-k documents reflect the ranking method of each enterprise in the initial search.

[0075] In this embodiment, a ranking model is used to perform refined ranking based on the top-k documents to achieve personalized ranking that better meets the actual needs of users.

[0076] In one embodiment, the click-through rate data is determined based on the number of clicks and impressions of each initially searched company. The calculation formula for the click-through rate is the number of clicks divided by the number of impressions, where the number of clicks is the number of times a user actually clicks on an ad or link, and the number of impressions is the total number of times the company information is displayed to users.

[0077] In another embodiment, Figure 3 As shown, a method for sorting enterprise search results is provided, including the following steps 302 to 314. Among them:

[0078] Step 302, obtain the search term, perform word segmentation on the search term through the word segmentation model, and obtain multiple keywords corresponding to the search term.

[0079] Step 304, input multiple keywords into a preset search engine based on vertical field information, and obtain multiple initial search companies through the preset search engine.

[0080] Step 306, use the trained associated feature model to process the search terms to obtain associated feature values ​​of each feature information associated with the search terms; the feature information includes at least one of the company name, the field in which the company is located, the business scope, financial information, shareholder information, whether it is a listed company, and whether there is an information disclosure platform.

[0081] Step 308, obtaining click-through rate data of each initially searched enterprise, and constructing a ranking model based on the click-through rate data and associated feature values.

[0082] Step 310, obtaining the top-k documents corresponding to each initially searched enterprise.

[0083] Step 312, using the ranking model to calculate each top-k document, to obtain the ranking score corresponding to each initially searched enterprise.

[0084] Step 314, sorting each initially searched enterprise in descending order based on the sorting score.

[0085] See also Figure 4 In an exemplary embodiment, a method for training a correlation feature model includes the following steps:

[0086] Step 402, obtaining historical search records; the historical search records include historical search terms and historical search results corresponding to the historical search terms, and the historical search results include multiple historical search companies.

[0087] Exemplarily, user search records within a preset time period in the past may be obtained, such as search records in the past three months.

[0088] Illustratively, multiple historical search results are also sorted in descending order of relevance.

[0089] Step 404, obtaining multiple feature information of different dimensions of the top N enterprises in the historical search results; N is an integer greater than or equal to 2.

[0090] Exemplarily, multiple feature information of different dimensions may include at least one of the company name, the field in which the company is located, the business scope, financial information, shareholder information, whether it is a listed company, and whether there is an information disclosure platform.

[0091] Furthermore, the feature information is cleaned, standardized or encoded so as to be input into the learning model.

[0092] Step 406: Calculate the correlation degree between the historical search terms and each feature information using a correlation degree algorithm to obtain the correlation feature value corresponding to the historical search terms and each feature information.

[0093] The association degree algorithm is an algorithm used to measure and discover the association degree between items (or features) in the data. Exemplarily, the association degree algorithm used in this embodiment can be an Apriori algorithm, and after the association feature value is calculated, it is saved in the feature.txt file.

[0094] Step 408: Perform training using a linear model training function based on each associated feature value to obtain a trained associated feature model.

[0095] Exemplarily, the linear model training function provided by the linear.h library is used to train the associated feature model according to the data in the feature file feature.txt to obtain a trained associated feature model.

[0096] Furthermore, the trained association feature model is loaded into the retrieval system.

[0097] In this embodiment, after the association feature model is constructed, it is trained using a linear model to obtain a trained association feature model. In traditional business searches, when the search keywords entered by the user are highly matched with the target document content, the BM25 algorithm may overemphasize keyword frequency and document length, causing search results to favor documents containing all or most of the keywords, rather than considering the overall relevance of the document and the actual needs of the user. When there are fewer keywords, a large number of documents may contain these keywords. At this time, relying solely on the BM25 algorithm is likely to result in a large number of documents with high similarity, and it is impossible to accurately distinguish which documents are more valuable to the user's query.

[0098] Therefore, this embodiment introduces learning to rank, using rich features such as user click data and browsing behavior for deep learning and training to achieve personalized ranking that better meets the actual needs of users. In addition to text content, factors such as enterprise attributes, user behavior data, and external evaluation can also be included in the ranking factors to enhance the comprehensive judgment ability of the ranking system. In large-scale data sets, hierarchical retrieval or two-stage retrieval strategies are adopted, firstly, rough sorting is performed through efficient retrieval algorithms such as BM25, and then refined sorting is performed with the help of more complex models.

[0099] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0100] Based on the same inventive concept, the embodiment of the present application also provides a device for sorting enterprise search results for implementing the enterprise search result sorting method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations of one or more embodiments of the device for sorting enterprise search results provided below can refer to the limitations of the enterprise search result sorting method above, and will not be repeated here.

[0101] In an exemplary embodiment, Figure 5 As shown, a device 500 for ranking enterprise search results is provided, comprising: a preliminary inspection module 502, a feature value determination module 504, a ranking model construction module 506, a scoring module 508 and a ranking module 510, wherein:

[0102] The initial search module 502 is used to obtain search terms and perform a preliminary search based on the search terms to obtain a plurality of initial search companies.

[0103] The feature value determination module 504 is used to process the search term using the trained associated feature model to obtain associated feature values ​​of each feature information associated with the search term.

[0104] The ranking model building module 506 is used to obtain the click-through rate data of each initially searched enterprise and build a ranking model based on the click-through rate data and the associated feature values.

[0105] The scoring module 508 is used to score each initially searched enterprise using the ranking model to obtain a ranking score corresponding to each initially searched enterprise.

[0106] The sorting module 510 is used to sort the initially searched enterprises in descending order based on the sorting scores.

[0107] In an exemplary embodiment, the initial inspection module 502 is also used to: obtain search terms, segment the search terms through a word segmentation model, and obtain multiple keywords corresponding to the search terms; input the multiple keywords into a preset search engine based on vertical field information, and obtain multiple initial search companies through the preset search engine.

[0108] In an exemplary embodiment, the scoring module 508 is also used to: obtain the top-k documents corresponding to each initially searched enterprise; calculate each top-k document using a ranking model to obtain a ranking score corresponding to each initially searched enterprise.

[0109] Based on the same inventive concept, the embodiment of the present application also provides a training device for an associated feature model for implementing the training method for the associated feature model involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the embodiments of the training device for one or more associated feature models provided below can refer to the limitations of the training method for the associated feature model above, and will not be repeated here.

[0110] In an exemplary embodiment, Figure 6 As shown, a training device 600 for a correlation feature model is provided, comprising: a history acquisition module 602, a feature information acquisition module 604, a correlation degree determination module 606 and a training module 608, wherein:

[0111] The history record acquisition module 602 is used to acquire the historical search record; the historical search record includes the historical search terms and the historical search results corresponding to the historical search terms, and the historical search results include multiple historical search companies.

[0112] The feature information acquisition module 604 is used to obtain multiple feature information of different dimensions of the top N enterprises in the historical search results; N is an integer greater than or equal to 2.

[0113] The correlation degree determination module 606 is used to calculate the correlation degree between the historical search terms and each feature information using a correlation degree algorithm to obtain the correlation feature value corresponding to the historical search terms and each feature information.

[0114] The training module 608 is used to perform training based on each associated feature value using a linear model training function to obtain a trained associated feature model.

[0115] Each module in the above-mentioned enterprise search result ranking device and / or the training device of the associated feature model can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0116] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 7 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a method for sorting enterprise search results or a method for training an associated feature model is implemented.

[0117] The display unit of the computer device is used to form a visually visible picture, and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball, or touchpad provided on the computer device housing, or an external keyboard, touchpad, or mouse, etc.

[0118] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0119] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0120] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0121] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0122] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0123] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0124] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for sorting enterprise search results, characterized in that: The method comprises: Obtaining a search term, and performing a preliminary search based on the search term to obtain a plurality of preliminary searched enterprises; Processing the search terms using the trained association feature model to obtain association feature values ​​of each feature information associated with the search terms; Obtaining click-through rate data of each of the initially searched companies, and constructing a ranking model based on the click-through rate data and the associated feature values; Scoring each of the initially searched enterprises using the ranking model to obtain a ranking score corresponding to each of the initially searched enterprises; The initially searched enterprises are sorted in descending order based on the sorting scores.

2. The method according to claim 1, characterized in that The obtaining of the search terms and performing a preliminary search based on the search terms to obtain a plurality of preliminary searched enterprises include: Obtaining a search term, and performing word segmentation processing on the search term through a word segmentation model to obtain a plurality of keywords corresponding to the search term; Input the plurality of keywords into a preset search engine based on vertical field information, and obtain a plurality of initial search companies through the preset search engine.

3. The method according to claim 1, characterized in that: The step of using the ranking model to score each of the initially searched enterprises to obtain a ranking score corresponding to each of the initially searched enterprises includes: Get the top-k documents corresponding to each initially searched enterprise; The ranking model is used to calculate each top-k document to obtain a ranking score corresponding to each of the initially searched enterprises.

4. The method according to claim 1, characterized in that The characteristic information includes at least one of the company name, the field in which the company is located, the business scope, financial information, shareholder information, whether it is a listed company, and whether there is an information disclosure platform.

5. The method according to claim 1, characterized in that The click-through rate data is determined based on the number of clicks and the number of impressions of each of the initially searched companies.

6. A method for training an association feature model, characterized in that: The method comprises the following steps: Acquire historical search records; the historical search records include historical search terms and historical search results corresponding to the historical search terms, and the historical search results include multiple historical search companies; Acquire multiple feature information of different dimensions of the top N enterprises in the historical search results; N is an integer greater than or equal to 2; Calculating the correlation degree between the historical search terms and each of the feature information using a correlation degree algorithm to obtain a correlation feature value corresponding to the historical search terms and each of the feature information; Based on each of the associated feature values, training is performed using a linear model training function to obtain a trained associated feature model.

7. A device for sorting enterprise search results, characterized in that: The device comprises: A preliminary search module, used to obtain search terms, and perform a preliminary search based on the search terms to obtain a plurality of preliminary searched enterprises; A feature value determination module, used to process the search term using the trained associated feature model to obtain associated feature values ​​of each feature information associated with the search term; A ranking model building module, used to obtain click-through rate data of each of the initially searched enterprises, and build a ranking model based on the click-through rate data and the associated feature value; A scoring module, used to score each of the initially searched enterprises using the ranking model to obtain a ranking score corresponding to each of the initially searched enterprises; The sorting module is used to sort each of the initially searched enterprises in descending order based on the sorting scores.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images