Information search and sorting method, device, computer equipment and storage medium

Abstract

The present application provides an information search and ranking method, apparatus, computer device, and storage medium, which can be applied to the fields of artificial intelligence or smart transportation, and is used to solve the problem of low search and ranking accuracy. The method includes: obtaining a search keyword; using a trained target search and ranking model to select multiple target guidance information that matches the search keyword from various candidate guidance information, wherein each target guidance information is associated with corresponding target formal information; using the target search and ranking model to obtain corresponding electronic resource prediction values based on the multiple target guidance information; and determining the arrangement order of the multiple target guidance information based on the respective electronic resource prediction values, thereby increasing the accuracy of the arrangement order of the multiple target guidance information.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to an information search and sorting method, apparatus, computer equipment, and storage medium. Background Art

[0002] With the continuous development of technology, the device can not only determine the matching search content for the search keyword, but also determine the arrangement order of the search content according to preset rules.

[0003] In related art, a search ranking method is to determine the order of each search content based on the degree of matching between the search content and the search keyword.

[0004] When the above method is used and the search keyword is relatively simple, more search contents with the same degree of matching as the search keyword will be generated, making it impossible to accurately determine the arrangement order of each search content based on the matching degree, thereby seriously affecting the sorting accuracy of each search content.

[0005] For example, in an education scenario, many courses fall into the same category. Searching for a keyword representing a category will retrieve all courses within that category. Since all retrieved courses have the same degree of match with the search keyword, they are ranked in the same order. Therefore, when faced with all courses in the same order, the target user needs to view each course individually; otherwise, they will not be able to accurately retrieve the course of interest.

[0006] It can be seen that under the related art, the accuracy of the search and ranking process is low. Summary of the Invention

[0007] The embodiments of the present application provide an information search and ranking method, apparatus, computer device, and storage medium to solve the problem of low search and ranking accuracy.

[0008] In a first aspect, a method for information search and ranking is provided, comprising:

[0009] Get search keywords;

[0010] Using the trained target search ranking model, a plurality of target guidance courses matching the search keyword are selected from each candidate guidance course, wherein each target guidance course is associated with a corresponding target formal course;

[0011] The target search ranking model is used to obtain corresponding electronic resource prediction values based on the multiple target-guided courses, wherein each electronic resource prediction value represents: the number of units of electronic resources obtained through each target object when promoting the associated target formal course to each target object based on the corresponding target-guided course within a first promotion time range starting at the current time;

[0012] An arrangement order of the plurality of goal-oriented courses is determined based on the predicted values of the respective electronic resources.

[0013] In a second aspect, an information search and sorting device is provided, comprising:

[0014] Acquisition module: used to obtain search keywords;

[0015] A processing module is configured to use a trained target search ranking model to select a plurality of target guidance courses that match the search keyword from each candidate guidance course, wherein each target guidance course is associated with a corresponding target formal course;

[0016] The processing module is further configured to: adopt the target search ranking model to obtain corresponding electronic resource prediction values based on the multiple target-guided courses, wherein each electronic resource prediction value represents: the number of units of electronic resources obtained by each target object when promoting the associated target formal course to each target object based on the corresponding target-guided course within a first promotion time range starting at the current time;

[0017] The processing module is further configured to determine an arrangement order of the plurality of target-guided courses based on the predicted values of the respective electronic resources.

[0018] Optionally, the target search ranking model is obtained by training the processing module using the following method:

[0019] Obtaining each sample data, wherein each sample data includes each sample guided course and respective historical guided data of each sample guided course, each historical guided data being used to represent relevant data generated when the corresponding sample guided course is promoted at a historical time;

[0020] Based on the sample data, the search ranking model to be trained is subjected to multiple rounds of iterative training until the training objectives are met, and the trained target search ranking model is output.

[0021] Optionally, the processing module in each round of iterative training is specifically configured to:

[0022] Get sample keywords;

[0023] Adopting a search ranking model, selecting a plurality of training guidance courses that match the sample keywords from the sample guidance courses, and obtaining corresponding electronic resource training values based on the plurality of training guidance courses, wherein each training guidance course is associated with a corresponding formal training course;

[0024] determining electronic resource sample values for each of the plurality of training guidance courses based on the respective historical guidance data of the plurality of training guidance courses;

[0025] Based on the error between each electronic resource training value and each electronic resource sample value, a training loss of the search ranking model is determined, and parameters are adjusted based on the training loss.

[0026] Optionally, the processing module is specifically configured to:

[0027] For the multiple training guidance courses, perform the following operations respectively:

[0028] Determining, based on historical guidance data of the training guidance course, a total number of first samples of electronic resources obtained when promoting the associated formal training course to each target object based on the training guidance course within the first promotion time range starting at the current time;

[0029] Obtaining the number of sample units of the electronic resource based on a ratio between the total number of the first samples and the number of the course guidance objects of each target object;

[0030] An electronic resource sample value of the training guide course is determined based on the number of sample units.

[0031] Optionally, the processing module is specifically configured to:

[0032] determining, based on historical guidance data of the training induction course, a second total number of samples of electronic resources obtained from formal training courses associated with the training induction course within a second promotion time range starting at the current time, wherein the second promotion time range is shorter than the first promotion time range;

[0033] The electronic resource sample value of the training guide course is determined based on a weighted sum of the sample unit quantity and the second total sample quantity.

[0034] Optionally, the processing module is specifically configured to:

[0035] For the multiple training guidance courses, perform the following operations respectively:

[0036] Determining, based on historical guidance data of the training induction course, a total number of third samples of electronic resources obtained when promoting the associated formal training course based on the training induction course to each target object within the first promotion time range ending at the current time;

[0037] Determining, based on the historical guidance data of the training guidance course, the number of electronic resource users of the target objects who use the electronic resources for the training guidance course among the target objects within the first promotion time range with the current time as the end time;

[0038] When it is determined that the ratio between the third total number of samples and the number of electronic resource users is greater than a ratio threshold, the electronic resource sample value of the training guidance course is determined based on the third total number of samples.

[0039] Optionally, the processing module is specifically configured to:

[0040] Determining, based on historical guidance data of the training guidance course, a fourth total number of samples of electronic resources obtained based on the training guidance course within a second promotion time range with the current time as the end time;

[0041] The electronic resource sample value of the training guide course is determined based on a weighted sum of the third total sample quantity and the fourth total sample quantity.

[0042] According to a third aspect, a computer program product is provided, comprising a computer program, which implements the method according to the first aspect when executed by a processor.

[0043] According to a fourth aspect, a computer device is provided, comprising:

[0044] a memory for storing program instructions;

[0045] The processor is configured to call the program instructions stored in the memory and execute the method described in the first aspect according to the obtained program instructions.

[0046] In a fifth aspect, a computer-readable storage medium is provided, wherein the storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the method described in the first aspect.

[0047] In an embodiment of the present application, after obtaining multiple target-guided courses that match the search keyword, the order of arrangement of the multiple target-guided courses is determined based on the electronic resource prediction values corresponding to each of the multiple target-guided courses. The electronic resource prediction value can represent the use of electronic resources by the target object for the corresponding target formal course within a first promotion time range starting from the current time, that is, the popularity of the corresponding target formal course. Therefore, by determining the order of arrangement of the multiple target-guided courses based on the electronic resource prediction value, the target formal courses with high popularity can be accurately ranked in a higher order and the target formal courses with low popularity can be ranked in a lower order, avoiding the situation where the order of arrangement of the target formal courses corresponding to the same category or the same course publishing end cannot be determined when the search keyword is a category or a video publishing end. At the same time, the target object can preferentially determine the target-guided course of interest from the target formal courses with high popularity, so that the target object can quickly determine the target-guided course of interest based on the accurate ranking order.

[0048] Furthermore, the electronic resource prediction value can also represent whether the target object uses the electronic resources for the target formal course when promoting the associated target formal course to each target object within the first promotion time range starting from the current time, that is, the promotion success rate. Therefore, by determining the arrangement order of multiple target guided courses based on the electronic resource prediction value, the target formal courses with high promotion success rates can be accurately arranged in an earlier order, and the target formal courses with low promotion success rates can be accurately arranged in a later order. Compared with the method of sorting based on click-through rate or exposure rate, the embodiment of the present application can avoid the same object viewing a certain guided course multiple times, or clicking on a certain guided course multiple times, resulting in a higher click-through rate or exposure rate of the guided course, but in fact the guided course did not successfully promote the formal course to the object, so that the object uses electronic resources for the formal course, thereby improving the accuracy of information search sorting. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 This is an application scenario of the information search and ranking method provided in the embodiments of the present application;

[0050] Figure 2a A schematic diagram of the principle of the information search and sorting method provided in the embodiment of the present application Figure 1 ;

[0051] Figure 2b A schematic diagram of a process for the information search and sorting method provided in the embodiment of the present application Figure 1 ;

[0052] Figure 3a A second schematic diagram of a principle of the information search and sorting method provided in an embodiment of the present application;

[0053] Figure 3b A second flow chart of the information search and sorting method provided in an embodiment of the present application;

[0054] Figure 4a A third flow chart of the information search and ranking method provided in an embodiment of the present application;

[0055] Figure 4b A third schematic diagram of a principle of the information search and ranking method provided in an embodiment of the present application;

[0056] Figure 5a A fourth schematic diagram of a principle of the information search and ranking method provided in an embodiment of the present application;

[0057] Figure 5b A fifth schematic diagram of a principle of the information search and sorting method provided in an embodiment of the present application;

[0058] Figure 6 A schematic diagram of the structure of the information search and sorting device provided in the embodiment of the present application Figure 1 ;

[0059] Figure 7 A second structural diagram of the information search and sorting device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0061] Some of the terms used in the embodiments of the present application are explained below to facilitate understanding by those skilled in the art.

[0062] (1) Search sorting:

[0063] Sorting is the process of sorting a collection of disorganized content based on certain characteristics to create a sequence of content. Search sorting is when a user enters a keyword into an interactive interface and the system returns sorted search results that match the associated keywords.

[0064] With the research and progress of artificial intelligence technology, artificial intelligence technology has been studied and applied in many fields, such as common smart homes, smart wearable devices, virtual assistants, smart speakers, smart marketing, unmanned driving, automatic driving, drones, robots, smart medical care, smart teaching, Internet of Vehicles, automatic driving, smart transportation, etc. I believe that with the development of technology, artificial intelligence technology will be applied in more fields and play an increasingly important role.

[0065] The embodiments of the present application relate to the field of artificial intelligence (AI) and the field of cloud technology. The embodiments of the present application can be applied to various scenarios such as smart transportation and assisted driving.

[0066] Artificial intelligence (AI) refers to the theories, methods, techniques, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, to perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a comprehensive technology within computer science that studies the design principles and implementation methods of various machines, attempting to understand the essence of intelligence and produce new intelligent machines that can respond in ways similar to human intelligence, enabling them to perceive, reason, and make decisions.

[0067] Artificial intelligence is a comprehensive discipline covering a wide range of fields, including both hardware and software technologies. The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation interaction systems, mechatronics, and other technologies. The software technologies of artificial intelligence mainly include computer vision technology, speech processing technology, natural language processing technology, machine learning / deep learning, autonomous driving, smart transportation, and other major directions. With the development and progress of artificial intelligence, artificial intelligence has been able to be researched and applied in many fields, such as common smart homes, smart customer service, virtual assistants, smart speakers, smart marketing, smart wearable devices, unmanned driving, autonomous driving, drones, robots, smart healthcare, Internet of Vehicles, autonomous driving, smart transportation, and other fields. It is believed that with the further development of future technology, artificial intelligence will be applied in more fields and play an increasingly important role. The solutions provided in the embodiments of this application involve technologies such as deep learning and augmented reality of artificial intelligence, which are further illustrated by the following embodiments.

[0068] Machine learning is a multidisciplinary interdisciplinary subject that involves probability theory, statistics, approximation theory, convex analysis, algorithmic complexity theory and other disciplines. It specializes in studying how computers can simulate human learning behavior to acquire new knowledge or skills, reorganize existing knowledge structures, and enable computers to continuously improve their own performance.

[0069] Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications span all areas of AI. At the core of machine learning is deep learning, a technology that enables machine learning. Machine learning generally includes techniques such as deep learning, reinforcement learning, transfer learning, inductive learning, artificial neural networks, and self-learning. Deep learning includes technologies such as convolutional neural networks (CNNs), deep belief networks, recurrent neural networks, autoencoders, and generative adversarial networks.

[0070] Cloud computing is a computing model that distributes computing tasks across a resource pool consisting of a large number of computers, enabling various application systems to access computing power, storage space, and information services as needed. The network that provides these resources is called the "cloud." To users, these resources appear infinitely scalable and can be accessed at any time, used on demand, expanded at any time, and paid for on a pay-per-use basis.

[0071] As a provider of cloud computing infrastructure, a cloud computing resource pool (referred to as a cloud platform, generally referred to as an Infrastructure as a Service (IaaS) platform) is established. Various types of virtual resources are deployed in the resource pool for external customers to choose and use. The cloud computing resource pool mainly includes: computing devices (virtualized machines, including operating systems), storage devices, and network devices.

[0072] Based on logical functional divisions, the Platform as a Service (PaaS) layer can be deployed on top of the IaaS layer, and the Software as a Service (SaaS) layer can be deployed on top of the PaaS layer. SaaS can also be deployed directly on top of IaaS. PaaS is a platform for software execution, such as databases and web containers. SaaS is a variety of business software, such as web portals and text messaging apps. Generally speaking, SaaS and PaaS are upper layers relative to IaaS.

[0073] It should be noted that in the embodiments of this application, terms such as "first," "second," or "third" are used to distinguish nouns and do not indicate a sequential order of nouns. For example, in the first and second promotional time ranges described later, "first" and "second" are used only to distinguish the two promotional time ranges and do not indicate a sequential order of the two promotional time ranges.

[0074] The following is a brief introduction to the application fields of the information search and ranking method provided in the embodiments of the present application.

[0075] With the continuous development of technology, the device can not only determine the matching search content for the search keyword, but also determine the arrangement order of the search content according to preset rules.

[0076] In related art, a search ranking method is to determine the order of each search content based on the degree of matching between the search content and the search keyword.

[0077] When the above method is used and the search keyword is relatively simple, more search contents with the same degree of matching as the search keyword will be generated, making it impossible to accurately determine the arrangement order of each search content based on the matching degree, thereby seriously affecting the sorting accuracy of each search content.

[0078] For example, in an education scenario, many courses fall into the same category. Searching for a keyword representing a category will retrieve all courses within that category. Since all retrieved courses have the same degree of match with the search keyword, they are ranked in the same order. Therefore, when faced with all courses in the same order, the target user needs to view each course individually; otherwise, they will not be able to accurately retrieve the course of interest.

[0079] It can be seen that under the related technology, the accuracy of the search and ranking process is low.

[0080] In order to solve the problem of low accuracy in course search ranking, the present application proposes an information search ranking method. After obtaining the search keyword, the method uses a trained target search ranking model to select multiple target guided courses that match the search keyword from various alternative guided courses, wherein each target guided course is associated with a corresponding target formal course. After obtaining multiple target guided courses, the target search ranking model is used to obtain corresponding electronic resource prediction values based on the multiple target guided courses, wherein each electronic resource prediction value represents: within the first promotion time range with the current time as the starting time, when promoting the associated target formal course to each target object based on the corresponding target guided course, the number of units of electronic resources obtained through each target object. After obtaining each electronic resource prediction value, the arrangement order of the multiple target guided courses is determined based on each electronic resource prediction value.

[0081] In an embodiment of the present application, after obtaining multiple target-guided courses that match the search keyword, the order of arrangement of the multiple target-guided courses is determined based on the electronic resource prediction values corresponding to each of the multiple target-guided courses. The electronic resource prediction value can represent the use of electronic resources by the target object for the corresponding target formal course within a first promotion time range starting from the current time, that is, the popularity of the corresponding target formal course. Therefore, by determining the order of arrangement of the multiple target-guided courses based on the electronic resource prediction value, the target formal courses with high popularity can be accurately ranked in a higher order and the target formal courses with low popularity can be ranked in a lower order, avoiding the situation where the order of arrangement of the target formal courses corresponding to the same category or the same course publishing end cannot be determined when the search keyword is a category or a video publishing end. At the same time, the target object can preferentially determine the target-guided course of interest from the target formal courses with high popularity, so that the target object can quickly determine the target-guided course of interest based on the accurate ranking order.

[0082] Furthermore, the electronic resource prediction value can also represent whether the target object uses the electronic resources for the target formal course when promoting the associated target formal course to each target object within the first promotion time range starting from the current time, that is, the promotion success rate. Therefore, by determining the arrangement order of multiple target guided courses based on the electronic resource prediction value, the target formal courses with high promotion success rates can be accurately arranged in an earlier order, and the target formal courses with low promotion success rates can be accurately arranged in a later order. Compared with the method of sorting based on click-through rate or exposure rate, the embodiment of the present application can avoid the same object viewing a certain guided course multiple times, or clicking on a certain guided course multiple times, resulting in a higher click-through rate or exposure rate of the guided course, but in fact the guided course did not successfully promote the formal course to the object, so that the object uses electronic resources for the formal course, thereby improving the accuracy of information search sorting.

[0083] The following describes the application scenarios of the information search and ranking method provided by this application.

[0084] Please refer to Figure 1 , which is a schematic diagram of an application scenario of the information search and ranking method provided in this application. The application scenario includes a client 101 and a server 102. The client 101 and the server 102 can communicate with each other. The communication method can be to use wired communication technology, such as communicating via a network cable or a serial port cable; or to use wireless communication technology, such as communicating via Bluetooth or wireless fidelity (WIFI) and other technologies, without specific limitation.

[0085] The client 101 generally refers to a device that can provide search keywords to the server 102 and can also display the sorting order, such as a terminal device, a third-party application that the terminal device can access, or a web page that the terminal device can access. Terminal devices include but are not limited to mobile phones, computers, intelligent voice interaction devices, intelligent transportation equipment, smart appliances, vehicle-mounted terminals, etc. The server 102 generally refers to a device that can generate a sorting order based on the obtained search keywords, such as a terminal device or a server. The server includes but is not limited to a cloud server, a local server, or an associated third-party server. Both the client 101 and the server 102 can use cloud computing to reduce the occupation of local computing resources; cloud storage can also be used to reduce the occupation of local storage resources.

[0086] As an embodiment, the client 101 and the server 102 can be the same device, and there is no specific limitation. In the embodiment of the present application, the client 101 and the server 102 are different devices as an example for introduction.

[0087] The following is based on Figure 1 , taking the server as the main body, the information search and sorting method provided in the embodiment of this application is specifically introduced.

[0088] Please refer to Figure 2a Figure 1 is a schematic diagram of a method for information search and ranking. After obtaining a search keyword, the server can use a target search ranking model to select multiple target-guided courses that match the search keyword and predict the corresponding electronic resource values for these target-guided courses. Based on the predicted values for each electronic resource, the server determines the order in which these target-guided courses are ranked.

[0089] Please refer to Figure 2b , which is a flow chart of the information search and sorting method provided in an embodiment of the present application.

[0090] S201, obtaining a search keyword.

[0091] The server can receive the search keyword sent by the client, and the client can obtain the search keyword in response to the editing operation triggered by the target object on the display interface. The client sends the search keyword to the server, so that the server receives the search keyword sent by the client; the server can also recognize the search keyword based on the voice instructions of the target object; the server can also generate corresponding search keywords based on the gesture operation of the target object, so that the server obtains the search keyword, etc. There is no specific limit to the method of obtaining the search keyword.

[0092] S202 , using the trained target search ranking model, select multiple target guidance courses that match the search keyword from various candidate guidance courses.

[0093] After obtaining the search keyword, the server may use a trained target search ranking model to perform search ranking processing, and first select multiple target guided courses that match the search keyword from various candidate guided courses.

[0094] Before using the trained target search ranking model to perform search ranking processing, the server may first obtain the trained target search ranking model. The server may directly receive the trained target search ranking model sent by other devices, or may train the to-be-trained search ranking model to obtain the trained target search ranking model before using the trained target search ranking model, etc., without limitation.

[0095] The following is an introduction to the method of training the search ranking model to be trained. Please refer to Figure 3a , which is a schematic diagram of a principle for training a search ranking model to be trained. After obtaining each sample data, the server can perform multiple rounds of iterative training on the search ranking model based on each sample data. When the training objectives are met, the server outputs the trained target search ranking model.

[0096] Please refer to Figure 3b , which is a flowchart of training a search ranking model to be trained.

[0097] S301, obtaining each sample data.

[0098] The server can read the historical guidance data for each sample guidance course and generate each sample data based on each sample guidance course and its own historical guidance data. The server can also receive each sample data sent by other devices, without limitation. Each sample data includes each sample guidance course and its own historical guidance data. Each historical guidance data is used to represent the relevant data generated when the corresponding sample guidance course promoted the associated sample formal course at a historical time.

[0099] As an embodiment, to ensure the validity of each sample data, each sample data may be pre-processed. For example, the server may determine whether there is any missing data in each sample data, such as missing historical guidance data, incomplete historical guidance data, missing sample guidance courses, or incomplete sample guidance course information. If there is any missing data in the sample data, the server may discard the sample data or supplement the missing sample data based on other non-missing sample data to ensure that each sample data has no missing data.

[0100] For another example, the server can determine whether there is data duplication in each sample data, such as containing two identical sample data, etc., then the server can deduplicate the redundant data so that every two sample data in each sample data are different data.

[0101] For example, the server can determine whether there are any abnormal numerical data in the historical guidance data of a sample guided course in each sample data. For example, in the historical guidance data of a sample guided course, the number of viewers of the sample guided course is generally in the range of [10, 100], but there is a time point where the number of viewers is 5 million. In this case, the data may be abnormal. The server can modify the abnormal data based on other normal data or discard the sample data, so that no abnormal numerical data is present in each sample data.

[0102] For another example, the server can determine whether there are sample data pairs with high correlation in each sample data. For example, in two sample data, the sample guidance courses are the same, but the historical guidance data are different. Then it can be determined that the two sample data have a high correlation. They may be historical guidance data obtained from the same sample guidance course in different time periods. Therefore, the server can merge the two sample data into one sample data to avoid the existence of sample data pairs with high correlation in each sample data.

[0103] As an embodiment, in order to improve the accuracy of the trained target search ranking model, positive samples and negative samples can be used to train the search ranking model, so that in the process of training the search ranking model, the search ranking model can learn more useful features in positive samples and less useless features in negative samples, and can also achieve the purpose of identifying positive features by learning negative features in negative samples.

[0104] After obtaining each sample data, the server can classify the sample data into positive sample data and negative sample data. The server can calculate a quality score for each sample data and classify the sample data into positive sample data and negative sample data based on the quality score. The following describes the process of calculating the quality score of one sample data as an example. The process of calculating the quality scores of other sample data is similar and is not further described here.

[0105] Based on the historical guidance data of the sample guided course in the sample data, the server can determine the number of subjects who were exposed to the sample guided course presented to the target subject; the number of subjects who clicked on the sample guided course and viewed the detailed introduction of the sample guided course by the target subject; the number of subjects who registered for the sample guided course and viewed the sample guided course by the target subject; and the number of subjects who converted from the target subject to using electronic resources for the sample formal course associated with the sample guided course within a first promotion time range starting at a specified time. The server can determine a first score for the sample guided course at a specified time based on a weighted sum of the number of subjects who were exposed, the number of subjects who clicked, the number of subjects who registered, and the number of subjects who converted.

[0106] When calculating the weighted sum of the number of exposures, clicks, registrations, and conversions, the weight of the number of exposures can be set to be smaller than the weight of the number of clicks, smaller than the weight of the number of registrations, and smaller than the weight of the number of conversions. From the course perspective, the quality score of the sample-guided course is measured based on the target audience's behavior statistics, such as viewing, clicking, or registration. Furthermore, the quality score of the sample-guided course is measured based on the target audience's use of electronic resources for the associated sample formal courses. This measures the subject's interest in the sample-guided course from multiple perspectives.

[0107] The weight of the number of conversion objects is set to the maximum, and the success rate of the sample-guided course in promoting the sample formal course is taken as the most important influencing factor. Therefore, the objects can be promoted according to the sample-guided course, and the guided course promotion with the highest possibility of using electronic resources for the associated sample formal courses can be ranked in the front. To a certain extent, the objects can quickly find the courses they are interested in according to the accurate arrangement order.

[0108] The server can also continue to determine the number of electronic resources generated by the sample formal courses associated with the sample guiding courses within the second promotion time range with the specified time as the starting time based on the historical guiding data of the sample guiding courses in the sample data, that is, the total number of electronic resources used by each target object for the sample formal courses; the server can also determine the course publishing end of the sample guiding course, and the total number of electronic resources generated by the sample formal courses associated with all sample guiding courses published under the category to which the sample guiding course belongs within the third promotion time range with the specified time as the starting time.

[0109] The server may determine a second score for the sample guided course at the designated time based on a weighted sum of the number of electronic resources generated within a second promotion time range starting at the designated time and the total number of electronic resources generated within a third promotion time range starting at the designated time, where the second promotion time range is shorter than the third promotion time range.

[0110] When calculating the weighted sum, the server can set the weight of the number of electronic resources generated within the second promotion time range starting at the specified time to be greater than the weight of the total number of electronic resources generated within the third promotion time range starting at the specified time. From the institutional side, on the one hand, from a micro perspective, consider the number of electronic resources generated by the sample formal courses associated with the sample guided courses in the recent period relative to the specified time; on the other hand, from a macro perspective, consider the total number of electronic resources generated by all sample formal courses on the course publishing end in the recent period relative to the specified time. If the total number of electronic resources generated by all sample formal courses is large, it means to a certain extent that there are more courses in all sample formal courses that are of interest to the subject. If the number of electronic resources generated by a certain sample formal course is large, it means that the sample formal course is very likely to be a course of interest to the subject.

[0111] Since the second promotion time range is smaller than the third promotion time range, the server may use the number of electronic resources generated by the sample formal courses associated with the sample guided courses in the recent period relative to the specified time as an important influencing factor. If the sample formal courses have generated a large number of electronic resources in the recent period relative to the specified time, it means that the sample formal courses are very likely to be the courses that the subject is interested in in the recent period relative to the specified time.

[0112] After obtaining the first score and the second score of the sample guided course at the specified time, the server may determine the quality score of the sample guided course at the specified time based on a weighted average of the first score and the second score.

[0113] Among them, the specified time can be any time, for example, the specified time is midnight on Monday of each week; for example, the specified time is six o'clock in the afternoon of every day; for example, the specified time is the time when the amount of data corresponding to the same time in each historical guidance data is greater than the data amount threshold, etc., and there is no specific restriction.

[0114] S302: Based on each sample data, the search ranking model to be trained is subjected to multiple rounds of iterative training until the training target is met, and the trained target search ranking model is output.

[0115] After obtaining each sample data, the server can perform multiple rounds of iterative training on the search ranking model to be trained based on each sample data. If the training objectives are not met, the server can adjust the model parameters of the search ranking model and continue to train the search ranking model with the adjusted model parameters based on each sample data until the training objectives are met. The server then outputs the search ranking model as the trained target search ranking model.

[0116] The following is an example of the process of one round of iterative training. The process of each round of iterative training is similar and will not be repeated here. Figure 4a , which is a flowchart of the process of one round of iterative training.

[0117] S401, obtaining sample keywords.

[0118] The server can extract keywords from each sample guide course contained in each sample data, and deduplicate the obtained keywords to obtain each sample keyword; the server can also determine the sample keywords based on the category to which each sample guide course belongs; the server can also first determine the similarity between each two sample guide courses, divide each sample data into multiple similar sample data sets, and extract corresponding sample keywords for each sample guide course in the same similar sample data set, etc., without specific restrictions.

[0119] S402 , using a search ranking model, selects a plurality of training guidance courses that match the sample keywords from each sample guidance course, and obtains corresponding electronic resource training values based on the plurality of training guidance courses.

[0120] After obtaining the sample keyword, the server may adopt a search ranking model to select multiple training guidance courses that match the sample keyword from each sample guidance course, and obtain corresponding electronic resource training values based on the multiple training guidance courses.

[0121] When selecting multiple training guidance courses that match the sample keyword, the server can match the sample keyword with each sample guidance course from multiple perspectives. The server can determine the course name similarity between the sample keyword and the sample guidance course; determine the course publishing end similarity between the sample keyword and the course publishing end name of the sample guidance course; and determine the course subject similarity between the sample keyword and the course subject name of the sample guidance course.

[0122] After obtaining the course name similarity, course publishing end similarity, and course subject similarity, the server can determine the similarity between the sample keyword and the sample guided courses based on a weighted sum of the course name similarity, course publishing end similarity, and course subject similarity. The server selects the sample guided courses whose similarity exceeds a similarity threshold as the multiple training guided courses that match the sample keyword.

[0123] S403 : Determine electronic resource sample values for each of the plurality of training guidance courses based on the respective historical guidance data of the plurality of training guidance courses.

[0124] Before executing S404 and determining the training loss of the search ranking model based on the error between each electronic resource training value and each electronic resource sample value, the server may first determine the electronic resource sample values of each of the multiple training guidance courses based on their respective historical guidance data.

[0125] The following takes a training guide course as an example to introduce the process of determining the electronic resource sample value of the training guide course. The process of determining the electronic resource sample value for each training guide course is similar and will not be repeated here.

[0126] There are multiple methods for determining the electronic resource sample value of a training guide course. Two of them are described below as examples. Multiple methods can also be used in combination without specific limitation.

[0127] Method 1:

[0128] The electronic resource sample value of the training guidance course is determined based on the number of sample units of electronic resources obtained by each target object when the associated training formal course is promoted to each target object based on the training guidance course within the first promotion time range starting from the current time.

[0129] Based on the historical guidance data of the training guidance course, the server can determine the total number of first samples of electronic resources obtained when promoting the associated formal training course to each target object based on the training guidance course within a first promotion time range starting from the current time. Based on the training guidance course, promoting the associated formal training course to each target object can be that the target object views the course introduction of the training guidance course and then signs up to watch the training guidance course, and the target object achieves the purpose of promoting the associated formal training course to the target object by watching the training guidance course; or it can be that the target object clicks to view the course introduction of the training guidance course while browsing each training guidance course, and the purpose of promoting the associated formal training course to the target object is achieved, etc., without specific limitation. The target object who clicks to view the course introduction of the training guidance course can include objects that have not signed up to watch the training guidance course, and can also include objects that have signed up to watch the training guidance course, etc.

[0130] Based on the training guidance course, when promoting the associated formal training course to each target object, the total number of the first sample of electronic resources obtained can be the total number of electronic resources used by each target object for the formal training course; it can also be the total number of electronic resources used by each target object for the training guidance course and the formal training course, etc., without specific restrictions.

[0131] After obtaining the first total sample quantity, the server can obtain the number of sample units of the electronic resource based on the ratio between the first total sample quantity and the number of course guide objects for each target object. The server can determine the electronic resource sample value of the training guide course based on the number of sample units. The number of course guide objects for each target object can be the unique visitor (UV) of the training guide course, for example, the number of target objects who registered to view the training guide course, or the number of target objects who clicked to view the course description of the training guide course.

[0132] The ratio between the total number of first samples and the number of course guidance objects for each target object can be the ratio between the total number of first samples and the number of target objects who sign up to watch the training guidance course; it can also be the ratio between the total number of first samples and the number of target objects who click to view the course introduction of the training guidance course, etc., without specific limitation.

[0133] The electronic resource sample value of the training guidance course is determined by the success rate of the formal training course associated with the promotion of the training guidance course within the first promotion time range starting from the current time. Therefore, the higher the success rate of the training guidance course, the larger the corresponding electronic resource sample value, and the lower the success rate of the training guidance course, the smaller the corresponding electronic resource sample value.

[0134] The first promotion period can be 14 days. Typically, when a training guide course promotes its associated formal training course to a target audience, the target audience will decide within 14 days whether to use the electronic resources for the formal training course. Therefore, by analyzing the success rate of promoting the formal training course within the next 14 days, we can accurately identify recently popular formal training courses and improve search ranking accuracy.

[0135] In one embodiment, after determining the number of sample units of the electronic resource, the server may further determine, based on historical training data for the training guidance course, a second total number of samples of the electronic resource obtained from the formal training courses associated with the training guidance course within a second promotional time range starting at the current time, where the second promotional time range is shorter than the first promotional time range. After obtaining the second total number of samples, the server may determine the electronic resource sample value for the training guidance course based on a weighted sum of the number of sample units and the second total number of samples.

[0136] The electronic resource sample value of the training induction course is determined by the success rate of the formal training course associated with the promotion of the training induction course within a first promotion time range starting from the current time, and the total number of electronic resources used by each target subject for the formal training course within a second promotion time range starting from the current time. Thus, a formal training course with a high promotion success rate and a large number of target subjects who have recently used electronic resources for the formal training course will have a larger corresponding electronic resource sample value, while a formal training course with a low promotion success rate or a small number of target subjects who have recently used electronic resources for the formal training course will have a smaller corresponding electronic resource sample value.

[0137] The second promotion period can be one day. Generally, when subjects search for a keyword and a large number of subjects use electronic resources for a particular formal course on that day, this indicates, to a certain extent, that the formal course is popular. Therefore, by calculating the total number of electronic resources generated by a formal training course within one day, we can further accurately identify recently popular formal training courses and improve the accuracy of search ranking.

[0138] As an embodiment, the first promotion time range, the second promotion time range, and the third promotion time range introduced above can use the current time at the time of training as the starting time, and after the first promotion time range, the second promotion time range, or the third promotion time range, the historical guidance data generated within the corresponding time range can be used to train the search ranking model. In order to avoid a long waiting time for training, the server can also use a specified historical time as the starting time, so that the historical guidance data generated within the first promotion time range, the second promotion time range, or the third promotion time range can be directly read to train the search ranking model. The server can also use the following method 2 to determine the electronic resource sample value of the training guidance course, and avoid the problem of long waiting time for training by estimating.

[0139] Method 2:

[0140] The electronic resource sample value of the training guidance course is determined based on the total number of electronic resources obtained when promoting the associated training formal course to each target object based on the training guidance course within the first promotion time range ending at the current time.

[0141] Based on the historical guidance data of the training guidance course, the server may determine a third total sample quantity of electronic resources obtained when promoting the associated formal training course to each target object based on the training guidance course within a first promotion time range ending at the current time. After determining the third total sample quantity, the server may determine, based on the historical guidance data of the training guidance course, the number of electronic resource users of the target object that used the electronic resource for the training guidance course within the first promotion time range ending at the current time. When the ratio between the third total sample quantity and the number of electronic resource users is greater than a ratio threshold, the server may determine an electronic resource sample value for the training guidance course based on the third total sample quantity.

[0142] The server can estimate the number of sample units in method one based on the ratio between the third total sample quantity and the number of electronic resource users. For example, if a training introductory course is promoted to 10 target subjects through the associated formal training course, and within a first promotion time range ending at the current time, 8 target subjects signed up for the training introductory course and 6 target subjects used electronic resources for the formal training course, then the third total sample quantity is the total number of electronic resources used by the 8 target subjects, and the number of electronic resource users is 8.

[0143] The third total number of samples of electronic resources obtained when promoting the associated formal training course to each target object based on the training guidance course within the first promotion time range ending at the current time can be obtained based on the time period of the first promotion time range ending at the current time; or it can be the average of the sum of the total number of third samples corresponding to multiple time periods based on the first time period of the first promotion time range ending at the current time, the second time period of the first promotion time range ending at the start time of the first time period, and so on; or it can be obtained by using the time period of the first promotion time range ending at the current time as a sliding window, sliding towards the historical time in units of one day, or sliding the sliding window farthest from the current time towards the sliding window ending at the current time, and obtaining the total number of electronic resources obtained when promoting the associated formal training course to each target object based on the training guidance course within multiple sliding windows as the third total number of samples. At the same time, the sum of the number of target objects that use electronic resources for the training guidance course among each target object within the multiple sliding windows is obtained as the number of electronic resource users.

[0144] Please refer to Figure 4b , taking the first promotion time range as a sliding window length, sliding from the sliding window with the current time as the end time, with the sliding time of one day as the sliding time, sliding to the historical time, taking three slides as an example, four sliding windows can be obtained. For each sliding window, the server can obtain the total number of electronic resources obtained when promoting the associated formal training courses to each target object based on the training guidance course within the sliding window. The server will take the average value of the sum of the four total quantities obtained as the third sample total number. For each sliding window, the server can also obtain the number of target objects that use electronic resources for the formal training courses within the sliding window. The server will take the average value of the sum of the four object quantities obtained as the number of objects using electronic resources.

[0145] The ratio threshold may be determined based on the ratio between the total number of electronic resources obtained within the multiple sliding windows and the total number of target objects. Therefore, when the server determines that the ratio between the total number of the third sample and the number of electronic resource users is greater than the ratio threshold, it indicates that the ratio is relatively stable within the multiple sliding windows. When the server determines that the ratio between the total number of the third sample and the number of electronic resource users is not greater than the ratio threshold, it indicates that the ratio is abnormal within the multiple sliding windows.

[0146] When the server determines that the ratio between the total number of third samples and the number of electronic resource users is not greater than the ratio threshold, the server may determine the total number of third samples and the corresponding number of electronic resource users for each sample formal course in all sample formal courses published by the course publishing end to which the training guide course belongs. The server may determine whether the distribution of the ratio of the sum of the total number of third samples corresponding to all sample formal courses to the sum of the number of electronic resource users corresponding to all sample formal courses is stable based on whether the ratio is greater than the course publishing end ratio threshold. If the ratio is greater than the course publishing end ratio threshold, the server may use the ratio to estimate the number of sample units in method one.

[0147] If it is not greater than the ratio threshold of the course publishing end, the server can determine whether the distribution of the ratio is stable based on whether the ratio of the sum of the total number of third samples corresponding to all sample formal courses and the sum of the number of electronic resource users corresponding to all sample formal courses is greater than the category ratio threshold. If it is greater than the category ratio threshold, the ratio can be used to estimate the number of sample units in method one. This can avoid the problem of a certain ratio being abnormal and causing low estimation accuracy.

[0148] As an embodiment, the process of determining the sample value of electronic resources for the training guidance course based on the third total sample quantity may involve the server determining, based on historical guidance data for the training guidance course, a fourth total sample quantity of electronic resources obtained based on the training guidance course within a second promotion time range ending at the current time. The server determines the sample value of electronic resources for the training guidance course based on a weighted sum of the third total sample quantity and the fourth total sample quantity. The fourth total sample quantity of electronic resources obtained based on the training guidance course may be the total number of electronic resources used by each target subject for the formal training course associated with the training guidance course.

[0149] As an embodiment, if the training guidance course requires the use of electronic resources to watch, then the fourth sample total number of electronic resources obtained based on the training guidance course also includes the total number of electronic resources used by each target object for the training guidance course within the second promotion time range with the current time as the end time.

[0150] S404 : Determine the training loss of the search ranking model based on the error between the training value of each electronic resource and the sample value of each electronic resource, and adjust the parameters based on the training loss.

[0151] After obtaining the electronic resource training values and sample values for each of the multiple training guide courses, the server can determine the training loss of the search ranking model based on the error between each electronic resource training value and each electronic resource sample value. If the training loss does not meet the training objective, the server can adjust the model parameters of the search ranking model based on the Bayesian optimization method and continue to train the search ranking model based on the sample data.

[0152] If the training loss meets the training target, the server can output the search ranking model to obtain the trained target search ranking model.

[0153] As an embodiment, if the training loss meets the training goal, the server can further verify the trained search ranking model based on the historical guidance data of the recently obtained guidance courses. The server can determine the electronic resource verification value of each of the multiple target guidance courses that match the search keyword based on the above method one or method two, and verify the trained search ranking model based on the error between the electronic resource prediction value of each of the multiple target guidance courses predicted by the trained search ranking model and the electronic resource verification value of the corresponding target guidance course. If the error is large, it means that the trained search ranking model has not met the use standard and needs to continue training. If the error is small, it means that the trained search ranking model has met the use standard and can be put into use as the target search ranking model.

[0154] As an embodiment, after obtaining a trained target search ranking model, the server can compare the ranking order of multiple target-guided courses that match the search keyword using the trained target search ranking model and not using the trained target search ranking model, using any search keyword, to determine the estimated value of the multiple target-guided courses. According to the second method described above, for each of the multiple target-guided courses, the server can determine the total number of electronic resources obtained when promoting the associated target formal course based on the target-guided course to each target object within a first promotion time range ending at the current time. Simultaneously, the server can determine the number of target objects within each target object that use electronic resources for the target-guided course within the first promotion time range ending at the current time. If the server determines that the ratio between the total number of electronic resources and the number of objects is greater than a course ratio threshold, the server uses the sum of the ratio between the total number of electronic resources and the number of objects and the total number of electronic resources obtained by the target formal course within a second promotion time range ending at the current time as the estimated value of the target-guided course. Based on the sum of the estimated values of the multiple target-guided courses, the estimated value of the ranking order of the multiple target-guided courses is determined.

[0155] When the server determines that the ratio between the total number of electronic resources and the number of objects is not greater than the course ratio threshold, it determines the total number of electronic resources obtained when promoting all target formal courses associated with the target guided course based on the course publishing end within the first promotion time range ending at the current time. At the same time, it determines the number of target objects that use electronic resources for all target guided courses published by the course publishing end in each target object within the first promotion time range ending at the current time. When it is determined that the ratio between the total number of electronic resources and the number of objects is greater than the publishing end ratio threshold, the ratio between the total number of electronic resources and the number of objects and the sum of the total number of electronic resources obtained by all target formal courses published by the course publishing end within the second promotion time range ending at the current time are used as the estimated value of the target guided course.

[0156] When the server determines that the ratio between the total number of electronic resources and the number of objects is not greater than the ratio threshold of the publishing end, it determines the total number of electronic resources obtained when promoting all target formal courses associated with the course category to which the target guided course belongs within the first promotion time range ending at the current time. At the same time, it determines the number of target objects that use electronic resources for all target guided courses under the course category among the target objects within the first promotion time range ending at the current time. The ratio between the total number of electronic resources and the number of objects and the sum of the total number of electronic resources obtained by all target formal courses under the course category within the second promotion time range ending at the current time are used as the estimated value of the target guided course.

[0157] Please refer to Table 1 to calculate the total number of first samples of electronic resources obtained when promoting the associated formal training courses to each target object based on the training guidance course within the first promotion time range starting at the current time. Then, based on the ratio between the first sample total number and the number of course guidance objects for each target object, obtain the number of sample units of the electronic resources. Simultaneously, calculate the total number of second samples of electronic resources obtained based on the formal training courses associated with the training guidance course within the second promotion time range starting at the current time. The server may compare the number of sample units, the number of second sample totals, and the sum of the number of sample units and the second sample total.

[0158] Table 1

[0159]

[0160]

[0161] Among them, from the three perspectives of the size of the number of sample units, the total size of the second samples, and the size of the sum of the number of sample units and the total size of the second samples, the method using the embodiments of the present application can obtain a more accurate arrangement order than the method not using the embodiments of the present application, thereby improving the accuracy of search sorting.

[0162] Refer to Table 2. If viewing the targeted course also requires the use of electronic resources, the server can also calculate the click-through rate (CTR) of the targeted course within the first promotional timeframe starting from the current time, to obtain the CTR improvement percentage for the targeted course. The server can also calculate the conversion rate (CVR) of the targeted course within the first promotional timeframe starting from the current time, to obtain the CVR improvement percentage.

[0163] The server can also calculate the CTR of the target official course associated with the target induction course within the first promotion time range starting from the current time, and obtain the CTR improvement percentage. The server can also calculate the CVR of the target official course within the first promotion time range starting from the current time, and obtain the CVR improvement percentage.

[0164] Table 2

[0165]

[0166] Among them, from multiple perspectives such as the target-guided course CTR, CVR, and the target formal course CTR, CVR, the method using the embodiments of the present application can obtain a more accurate arrangement order than the method not using the embodiments of the present application, thereby improving the accuracy of search sorting.

[0167] S203: Using a target search ranking model, corresponding electronic resource prediction values are obtained based on multiple target-guided courses.

[0168] After the server obtains multiple target-guided courses, it can use a target search ranking model to obtain corresponding electronic resource prediction values based on the multiple target-guided courses. Each electronic resource prediction value represents: within the first promotion time range starting from the current time, when promoting the associated target formal course to each target object based on the corresponding target-guided course, the number of units of electronic resources obtained by each target object. Based on the corresponding target-guided course, the associated target formal course is promoted to each target object. This can be done by the target object viewing the course introduction of the target-guided course and then signing up to watch the target-guided course. The target object achieves the purpose of promoting the associated target formal course to the target object by watching the target-guided course; or the target object clicks to view the course introduction of the target-guided course while browsing each target-guided course, thereby achieving the purpose of promoting the associated target formal course to the target object, etc. There is no specific limitation. The target object who clicks to view the course introduction of the target-guided course can include objects that have not signed up to watch the target-guided course, and can also include objects that have signed up to watch the target-guided course, etc.

[0169] The unit quantity represents the ratio between the first target total quantity and the number of objects of each target object. Specifically, when promoting the associated target formal course to each target object based on the target-guided course, the first target total quantity of electronic resources obtained can be the total number of electronic resources used by each target object for the target formal course, or the total number of electronic resources used by each target object for both the target-guided course and the target formal course, etc., without limitation. The number of objects of each target object can be the exposure UV of the target-guided course, for example, the number of target objects who signed up to watch the target-guided course, or the number of target objects who clicked to view the course description of the target-guided course, etc.

[0170] The ratio between the total number of first targets and the number of objects of each target object can be the ratio between the total number of first targets and the number of target objects that have signed up to watch the target-guided course; it can also be the ratio between the total number of first targets and the number of target objects that click to view the course introduction of the target-guided course, etc., without specific limitation.

[0171] S204: Determine the arrangement order of the multiple target-guided courses based on the predicted values of the electronic resources.

[0172] After obtaining the predicted values of each electronic resource, the server can determine the order of the multiple target-guided courses based on the predicted values of each electronic resource. The server can arrange each target-guided course in descending order according to the predicted value of the electronic resource to obtain the order of the multiple target-guided courses; or it can arrange each target-guided course in descending order according to the predicted value of the electronic resource to obtain the order of the multiple target-guided courses, etc., without limitation.

[0173] The following is an example introduction to the information search and sorting method provided in the embodiments of the present application.

[0174] After each guided course and its associated official course are launched, each subject can browse, view details, or watch the guided course and official course. After each guided course and its associated official course are launched, corresponding historical guidance data can be generated based on the operation records of each subject. The server can train a search ranking model based on the corresponding historical guidance data of each guided course and its associated official course, and put the trained target search ranking model into use, so that the server can accurately determine the ranking order of multiple target guided courses for the search keywords provided by each subject.

[0175] Please refer to Figure 5a , which is a schematic diagram of the principles of information search and ranking. The search and ranking model can include a feature extraction module and a prediction module. The server can obtain each sample data based on the historical guidance data of the guided courses that ended 14 days prior to the current time. The server uses the feature extraction module of the search and ranking model to obtain the characteristics of each sample guided course. Based on the characteristics of each sample guided course, the server selects multiple target guided courses from each sample guided course that match the sample keywords.

[0176] The server uses a prediction module of the search ranking model to determine the sample values of electronic resources for each of the multiple target-guided courses based on the historical guidance data of each of the multiple target-guided courses. Simultaneously, the server uses the search ranking model to obtain predicted values of electronic resources for each of the multiple target-guided courses. The server also includes a training module that determines a training loss for the search ranking model based on the error between each predicted value of the electronic resource and the corresponding sample value of the electronic resource. If, based on the training loss, it is determined that the training objective is not met, the model parameters of the search ranking model are adjusted, and the search ranking model with the adjusted model parameters is continuously trained based on the sample data until, based on the obtained training loss, it is determined that the training objective is met, thereby obtaining a trained search ranking model.

[0177] The server may include a verification module. After obtaining the trained search ranking model, the verification module may use historical guidance data of sample-guided courses within 14 days with the current time as the end time, and historical guidance data of sample-guided courses within 14 days with the start time 14 days ago as the end time, etc., to verify the trained search ranking model.

[0178] Since the current time during verification and training is different, the server also needs to update the features of the corresponding sample guidance courses based on the historical guidance data generated between the two current times, using the feature extraction module of the trained search ranking model. If a new guidance course is introduced during this period, the new guidance course can also be used as a sample guidance course to participate in the verification process of the search ranking model.

[0179] The server uses the trained search ranking model to obtain multiple verification guide courses that match the verification keyword from each sample guide course. The server also uses the trained search ranking model to obtain the electronic resource verification values for each of the multiple verification guide courses. Based on the second method described above, the server obtains the electronic resource sample values for each of the multiple verification guide courses. The verification module determines whether the trained search ranking model can be used as the target search ranking model based on the error between the electronic resource verification value and the electronic resource sample value. If not, the search ranking model continues to be trained based on the sample data.

[0180] Please refer to Figure 5b , the server obtains the search keyword "keyword A", and adopts the target search ranking model to select multiple target guidance courses that match "keyword A" from various alternative guidance courses, including "keyword A, from entry to mastery", "Simple and easy to understand keyword A", "Keyword A introduction" and "Understand keyword A, one lesson is enough" four target guidance courses.

[0181] The goal-guided course "Keyword A, from Beginner to Mastery" belongs to course category A, and the course publishing end is publishing end A; the goal-guided course "Easy-to-understand Keyword A" belongs to course category A, and the course publishing end is publishing end B; the goal-guided course "Introduction to Keyword A" belongs to course category B, and the course publishing end is publishing end A; the goal-guided course "Understand Keyword A, One Lesson is Enough" belongs to course category B, and the course publishing end is publishing end B.

[0182] The server uses the target search ranking model to obtain the predicted electronic resource values for the four target-oriented courses: 10.01, 8.18, 22.79, and 8.48, respectively. The server sorts the four target-oriented courses in descending order of predicted electronic resource values, resulting in the following rankings: "Introduction to Keyword A," "Keyword A, From Beginner to Mastery," "Understanding Keyword A: One Lesson Is Enough," and "Easy-to-Understand Keyword A."

[0183] In an embodiment of the present application, each target guided course is ranked according to the number of units of electronic resources obtained by each target object when promoting the associated target formal course to each target object based on the corresponding target guided course within a first promotion time range starting at the current time. Compared with the method of sorting based on click-through rate, etc., the embodiment of the present application can avoid the same object viewing a certain guided course multiple times, or clicking on a certain guided course multiple times, resulting in a high click-through rate or exposure rate of the guided course, but in fact the guided course did not successfully promote the formal course to the object, so that the object uses electronic resources for the formal course, thereby improving the accuracy of information search sorting.

[0184] Furthermore, the Bayesian optimization method is used to automatically adjust the model parameters, reducing the reliance on manual parameter adjustment experience. The second method introduced above does not require waiting for the first promotion time range to obtain electronic resource sample values, which speeds up the training of the search ranking model.

[0185] Based on the same inventive concept, the embodiment of the present application provides an information search and sorting device that can implement the functions corresponding to the aforementioned information search and sorting method. Figure 6 , the device includes an acquisition module 601 and a processing module 602, wherein:

[0186] Acquisition module 601: used to obtain search keywords;

[0187] Processing module 602: configured to use the trained target search ranking model to select multiple target guidance courses that match the search keyword from each candidate guidance course, wherein each target guidance course is associated with a corresponding target formal course;

[0188] The processing module 602 is further configured to: adopt a target search ranking model to obtain corresponding electronic resource prediction values based on a plurality of target-guided courses, wherein each electronic resource prediction value represents: the number of units of the electronic resource obtained by each target object when promoting the associated target formal course to each target object based on the corresponding target-guided course within a first promotion time range starting at the current time;

[0189] The processing module 602 is further configured to determine an arrangement order of the plurality of target-guided courses based on the predicted values of the respective electronic resources.

[0190] In a possible embodiment, the target search ranking model is obtained by training the processing module 602 using the following method:

[0191] Obtaining each sample data, wherein each sample data includes each sample guided course and respective historical guided data of each sample guided course, each historical guided data being used to represent the relevant data generated when the corresponding sample guided course is promoted in a historical time;

[0192] Based on each sample data, the search ranking model to be trained is trained for multiple rounds of iterative training until the training objectives are met, and the trained target search ranking model is output.

[0193] In a possible embodiment, each round of iterative training processing module 602 is specifically configured to:

[0194] Get sample keywords;

[0195] A search ranking model is used to select multiple training guidance courses that match the sample keywords from each sample guidance course, and corresponding electronic resource training values are obtained based on the multiple training guidance courses, wherein each training guidance course is associated with a corresponding formal training course;

[0196] Determining electronic resource sample values for each of the plurality of training guidance courses based on respective historical guidance data of the plurality of training guidance courses;

[0197] Based on the error between the training value of each electronic resource and the sample value of each electronic resource, the training loss of the search ranking model is determined, and the parameters are adjusted based on the training loss.

[0198] In a possible embodiment, the processing module 602 is specifically configured to:

[0199] For multiple guided training sessions, perform the following operations:

[0200] Determine, based on historical guidance data of the training guidance course, a total number of first samples of electronic resources obtained when promoting the associated formal training course to each target object based on the training guidance course within a first promotion time range starting from the current time;

[0201] Obtaining the number of sample units of the electronic resource based on the ratio between the total number of the first sample and the number of course guidance objects of each target object;

[0202] Determine the sample value of the electronic resource for the training guide course based on the number of sample units.

[0203] In a possible embodiment, the processing module 602 is specifically configured to:

[0204] Determining, based on historical guidance data of the training induction course, a second total number of samples of electronic resources obtained from the formal training courses associated with the training induction course within a second promotion time range starting at the current time, wherein the second promotion time range is shorter than the first promotion time range;

[0205] An electronic resource sample value for the training guide course is determined based on a weighted sum of the sample unit quantity and the second sample total quantity.

[0206] In a possible embodiment, the processing module 602 is specifically configured to:

[0207] For multiple guided training sessions, perform the following operations:

[0208] Determining, based on historical guidance data of the training induction course, a total number of third samples of electronic resources obtained when promoting the associated formal training course to each target object based on the training induction course within a first promotion time range ending at the current time;

[0209] Based on the historical guidance data of the training guidance course, determining the number of electronic resource users of the target objects who use the electronic resources for the training guidance course among the target objects within a first promotion time range with the current time as the end time;

[0210] When it is determined that the ratio of the third total sample quantity to the number of electronic resource users is greater than a ratio threshold, the electronic resource sample value of the training guidance course is determined based on the third total sample quantity.

[0211] In a possible embodiment, the processing module 602 is specifically configured to:

[0212] Determining, based on historical guidance data of the training guidance course, a fourth total number of samples of electronic resources obtained based on the training guidance course within a second promotion time range with the current time as the end time;

[0213] An electronic resource sample value of the training guidance course is determined based on a weighted sum of the third sample total quantity and the fourth sample total quantity.

[0214] Based on the same inventive concept, an embodiment of the present application provides a computer device, which may be a terminal device or a server, without specific limitation. The following describes a structure of a computer device as an example.

[0215] Please refer to Figure 7The above-mentioned information search and ranking device can be run on a computer device 700. The current version and historical versions of the data storage program and the application software corresponding to the data storage program can be installed on the computer device 700. The computer device 700 includes a processor 780 and a memory 720. In some embodiments, the computer device 700 may include a display unit 740, and the display unit 740 includes a display panel 741 for displaying a user interactive operation interface, etc.

[0216] In a possible embodiment, the display panel 741 may be configured in the form of a liquid crystal display (LCD) or an organic light-emitting diode (OLED).

[0217] The processor 780 is configured to read a computer program and then execute the method defined by the computer program. For example, the processor 780 reads a data storage program or file, thereby running the data storage program on the computer device 700 and displaying a corresponding interface on the display unit 740. The processor 780 may include one or more general-purpose processors and may also include one or more DSPs (Digital Signal Processors) to perform related operations to implement the technical solutions provided in the embodiments of the present application.

[0218] The memory 720 generally includes internal memory and external memory, and the internal memory can be a random access memory (RAM), a read-only memory (ROM), and a cache (CACHE), etc. The external memory can be a hard disk, an optical disk, a USB disk, a floppy disk or a tape drive, etc. The memory 720 is used to store computer programs and other data. The computer program includes an application corresponding to each client, etc. Other data may include data generated after the operating system or application is run, and the data includes system data (such as configuration parameters of the operating system) and user data. In the embodiment of the present application, program instructions are stored in the memory 720, and the processor 780 executes the program instructions stored in 720 to implement any of the information search and sorting methods discussed in the previous figure.

[0219] The display unit 740 is used to receive input digital information, character information, or contact touch operations / contactless gestures, and to generate signal input related to user settings and function control of the computer device 700. Specifically, in the embodiment of the present application, the display unit 740 may include a display panel 741. The display panel 741, such as a touch screen, can collect user touch operations on or near it (such as operations performed by the user using a finger, stylus, or any other suitable object or accessory on or on the display panel 741) and drive corresponding connected devices according to a pre-set program.

[0220] In one possible embodiment, the display panel 741 may include two parts: a touch detection device and a touch controller. The touch detection device detects the player's touch position and detects the signal generated by the touch operation, and transmits the signal to the touch controller. The touch controller receives the touch information from the touch detection device, converts it into touch point coordinates, and then sends it to the processor 780. The touch controller can also receive and execute commands sent by the processor 780.

[0221] The display panel 741 can be implemented using various types such as resistive, capacitive, infrared, and surface acoustic wave. In addition to the display unit 740, in some embodiments, the computer device 700 may further include an input unit 730. The input unit 730 may include an image input device 731 and other input devices 732. The other input devices may include, but are not limited to, one or more of a physical keyboard, function keys (such as volume control keys, power keys, etc.), a trackball, a mouse, a joystick, and the like.

[0222] In addition to the above, the computer device 700 may also include a power supply 790 for powering other modules, an audio circuit 760, a near-field communication module 770, and an RF circuit 710. The computer device 700 may also include one or more sensors 750, such as an accelerometer, a light sensor, a pressure sensor, etc. The audio circuit 760 specifically includes a speaker 761 and a microphone 762. For example, the computer device 700 can use the microphone 762 to collect the user's voice and perform corresponding operations.

[0223] As an embodiment, the number of processors 780 may be one or more, and the processor 780 and the memory 720 may be coupled or relatively independently configured.

[0224] As an example, Figure 7 The processor 780 in the embodiment may be used to implement the following Figure 6 The functions of the acquisition module 601 and the processing module 602 in .

[0225] As an example, Figure 7 The processor 780 in can be used to implement the corresponding functions of the server or terminal device discussed above.

[0226] Those skilled in the art will appreciate that all or part of the steps of the above-mentioned method embodiments may be implemented by hardware associated with program instructions, and the aforementioned program may be stored in a computer-readable storage medium. When the program is executed, the program executes the steps of the above-mentioned method embodiments. The aforementioned storage medium includes various media that can store program codes, such as mobile storage devices, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0227] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention, or the part that contributes to the existing technology, can be embodied in the form of a software product, for example, through a computer program product, which is stored in a storage medium and includes a number of instructions for enabling a computer device to execute all or part of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program code, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0228] Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is intended to include these modifications and variations.

Claims

1. An information search and ranking method, characterized in that: include: Get search keywords; Using the trained target search ranking model, a plurality of target guidance courses matching the search keyword are selected from each candidate guidance course, wherein each target guidance course is associated with a corresponding target formal course; The target search ranking model is used to obtain corresponding electronic resource prediction values based on the multiple target-guided courses, wherein each electronic resource prediction value represents: within a first promotion time range starting at the current time, when promoting the associated target formal course to each target object based on the corresponding target-guided course, the number of units of electronic resources used by each target object for the target formal course; the unit number represents: the ratio between the total number of electronic resources used by each target object and the number of objects of each target object; An arrangement order of the plurality of goal-oriented courses is determined based on the predicted values of the respective electronic resources.

2. The method according to claim 1, characterized in that The target search ranking model is trained using the following method: Obtaining each sample data, wherein each sample data includes each sample guided course and respective historical guided data of each sample guided course, each historical guided data being used to represent relevant data generated when the corresponding sample guided course is promoted at a historical time; Based on the sample data, the search ranking model to be trained is subjected to multiple rounds of iterative training until the training objectives are met, and the trained target search ranking model is output.

3. The method according to claim 2, characterized in that Based on the sample data, the search ranking model to be trained is subjected to multiple rounds of iterative training until the training target is met, and the trained target search ranking model is output, wherein each round of iterative training includes: Get sample keywords; Adopting a search ranking model, selecting a plurality of training guidance courses that match the sample keywords from the sample guidance courses, and obtaining corresponding electronic resource training values based on the plurality of training guidance courses, wherein each training guidance course is associated with a corresponding formal training course; determining electronic resource sample values for each of the plurality of training guidance courses based on the respective historical guidance data of the plurality of training guidance courses; Based on the error between each electronic resource training value and each electronic resource sample value, a training loss of the search ranking model is determined, and parameters are adjusted based on the training loss.

4. The method according to claim 3, characterized in that Determining the electronic resource sample values of the plurality of training guidance courses based on the respective historical guidance data of the plurality of training guidance courses, respectively, includes: For the multiple training guidance courses, perform the following operations respectively: Determining, based on historical guidance data of the training guidance course, a total number of first samples of electronic resources obtained when promoting the associated formal training course to each target object based on the training guidance course within the first promotion time range starting at the current time; Obtaining the number of sample units of the electronic resource based on a ratio between the total number of the first samples and the number of the course guidance objects of each target object; An electronic resource sample value of the training guide course is determined based on the number of sample units.

5. The method according to claim 4, characterized in that Determining the electronic resource sample value of the training guide course based on the number of sample units includes: determining, based on historical guidance data of the training induction course, a second total number of samples of electronic resources obtained from formal training courses associated with the training induction course within a second promotion time range starting at the current time, wherein the second promotion time range is shorter than the first promotion time range; The electronic resource sample value of the training guide course is determined based on a weighted sum of the sample unit quantity and the second total sample quantity.

6. The method according to claim 3, characterized in that Determining the electronic resource sample values of the plurality of training guidance courses based on the respective historical guidance data of the plurality of training guidance courses, respectively, includes: For the multiple training guidance courses, perform the following operations respectively: Determining, based on historical guidance data of the training induction course, a total number of third samples of electronic resources obtained when promoting the associated formal training course based on the training induction course to each target object within the first promotion time range ending at the current time; Determining, based on the historical guidance data of the training guidance course, the number of electronic resource users of the target objects who use the electronic resources for the training guidance course among the target objects within the first promotion time range with the current time as the end time; When it is determined that the ratio between the third total number of samples and the number of electronic resource users is greater than a ratio threshold, the electronic resource sample value of the training guidance course is determined based on the third total number of samples.

7. The method according to claim 6, characterized in that Determining the electronic resource sample value of the training guidance course based on the third total sample quantity includes: Determining, based on historical guidance data of the training guidance course, a fourth total number of samples of electronic resources obtained based on the training guidance course within a second promotion time range with the current time as the end time; The electronic resource sample value of the training guide course is determined based on a weighted sum of the third total sample quantity and the fourth total sample quantity.

8. An information search and sorting device, characterized in that: include: Acquisition module: used to obtain search keywords; A processing module is configured to use a trained target search ranking model to select a plurality of target guidance courses that match the search keyword from each candidate guidance course, wherein each target guidance course is associated with a corresponding target formal course; The processing module is further configured to: adopt the target search ranking model to obtain corresponding electronic resource prediction values based on the multiple target-guided courses, wherein each electronic resource prediction value represents: within a first promotion time range starting at the current time, when promoting the associated target formal course to each target object based on the corresponding target-guided course, the number of units of electronic resources used by each target object for the target formal course; the number of units represents: the ratio between the total number of electronic resources used by each target object and the number of objects of each target object; The processing module is further configured to determine an arrangement order of the plurality of target-guided courses based on the predicted values of the respective electronic resources.

9. 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 7 are implemented.

10. A computer device, characterized in that: include: a memory for storing program instructions; A processor is configured to call the program instructions stored in the memory and execute the method according to any one of claims 1 to 7 according to the obtained program instructions.

11. A computer-readable storage medium, characterized in that The storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the method according to any one of claims 1 to 7.

Images