WO2023138428A1

WO2023138428A1 - Search result sorting method, search system and computer-readable storage medium

Info

Publication number: WO2023138428A1
Application number: PCT/CN2023/071322
Authority: WO
Inventors: 殷俊杰; 周祥生; 高洪; 屠要峰; 钟斌
Original assignee: 中兴通讯股份有限公司
Priority date: 2022-01-19
Filing date: 2023-01-09
Publication date: 2023-07-27
Also published as: CN116501951A

Abstract

Disclosed in the present application are a search result sorting method, a search system and a computer-readable storage medium. The search result sorting method comprises: acquiring a search request, wherein the search request comprises keyword information and a user identifier (S110); acquiring at least two target search results according to the keyword information (S120); and determining user preference information according to the user identifier, and determining the order of the at least two target search results according to the user preference information (S130).

Description

Search result sorting method, search system, computer readable storage medium

Cross References to Related Applications

This application is based on a Chinese patent application with application number 202210059836.3 and a filing date of January 19, 2022, and claims the priority of this Chinese patent application. The entire content of this Chinese patent application is hereby incorporated by reference into this application.

technical field

The present application relates to but not limited to the technical field of data processing, and in particular relates to a method for sorting search results, a search system, and a computer-readable storage medium.

Background technique

With the development of information technology, search processing technology is becoming more and more mature, becoming the main entrance for users to find information. The current search method can filter search results matching the search requirement information from the database of the search system or the Internet according to the search keywords and present them to the user. However, there are differences among users, and different users submitting the same search keywords may have different search requirements, and the target search results obtained by the current search methods are difficult to meet the personalized search requirements of users.

Contents of the invention

The following is an overview of the topics described in detail in this article. This summary is not intended to limit the scope of the claims.

Embodiments of the present application provide a method for sorting search results, a search system, and a computer-readable storage medium.

In a first aspect, an embodiment of the present application provides a method for ranking search results, including: obtaining a search request, the search request including keyword information and user identification; obtaining at least two target search results according to the keyword information; determining user preference information according to the user identification, and determining the ranking of at least two target search results according to the user preference information.

In a second aspect, an embodiment of the present application provides a search system, including: a memory, a processor, and a computer program stored in the memory and operable on the processor, and the processor implements the method for sorting search results as described in any one of the embodiments of the first aspect when executing the computer program.

In a third aspect, the embodiments of the present application further provide a computer-readable storage medium storing computer-executable instructions, the computer-executable instructions being used to execute the method for ranking search results as described in any one embodiment of the first aspect.

Additional features and advantages of the application will be set forth in the description which follows, and, in part, will be obvious from the description, or may be learned by practice of the application. The objectives and other advantages of the application will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.

Description of drawings

The accompanying drawings are used to provide a further understanding of the technical solution of the present application, and constitute a part of the specification, and are used together with the embodiments of the present application to explain the technical solution of the present application, and do not constitute a limitation to the technical solution of the present application.

FIG. 1 is a flow chart of steps of a method for sorting search results provided by an embodiment of the present application;

FIG. 2 is a flow chart of steps for obtaining user preference information provided by another embodiment of the present application;

FIG. 3 is a flow chart of steps for obtaining search result types provided by another embodiment of the present application;

FIG. 4 is a flow chart of steps of a method for sorting search results provided in another embodiment of the present application;

Fig. 5 is a flow chart of the steps of training the first sorting model provided by another embodiment of the present application;

FIG. 6 is a flow chart of steps for obtaining training data provided by another embodiment of the present application;

Fig. 7 is a flow chart of steps for sorting target search results before inputting target search results into the first sorting model provided by another embodiment of the present application;

FIG. 8 is a flow chart of steps for filtering target search results provided by another embodiment of the present application;

Fig. 9 is a block diagram of a search system provided by another embodiment of the present application;

Fig. 10 is a flowchart of steps of a ranking model training method provided by another embodiment of the present application;

Fig. 11 is a flow chart of the steps of the search result sorting method provided by another embodiment of the present application;

Fig. 12 is a flow chart of steps for marking document information provided by another embodiment of the present application;

Fig. 13 is a schematic structural diagram of a search system provided by another embodiment of the present application.

Detailed ways

In order to make the purpose, technical solution and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the embodiments described here are only used to explain the present application, not to limit the present application.

It should be noted that although the functional modules are divided in the schematic diagram of the device, and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the flow chart. The terms "first", "second" and the like in the specification, claims or the above-mentioned drawings are used to distinguish similar objects, and not necessarily used to describe a specific order or sequential order.

The present application provides a method for sorting search results, a search system, and a computer-readable storage medium, wherein the method for sorting search results includes: obtaining a search request, the search request including keyword information and a user ID; acquiring at least two target search results according to the keyword information; determining user preference information according to the user ID, and determining the ranking of at least two target search results according to the user preference information. According to the solution provided by the embodiment of the present application, the user preference information is determined according to the user identification, and the ranking of the search results is determined according to the user preference information. Compared with the current technical solution for matching and sorting the target search results based only on keyword information, the technical solution of the present application can improve the matching degree between the target search results and the search needs, thereby meeting the personalized search needs of users.

The embodiments of the present application will be further described below in conjunction with the accompanying drawings.

As shown in Figure 1, Figure 1 is a flow chart of the steps of a method for sorting search results provided by an embodiment of the present application. The method for sorting search results includes but is not limited to the following steps:

In step S110, a search request is obtained, and the search request includes keyword information and user identification.

It can be understood that the search request includes the keyword information input by the user to the user interface of the search system and the user identification corresponding to the input keyword information, which can provide a data basis for obtaining target search results and user preference information.

It should be noted that those skilled in the art are familiar with methods for acquiring keyword information and user identifiers, and this embodiment of the present application does not make any limitations here.

It should be noted that the embodiment of the present application does not limit the content of the user identification, which may be the Internet Protocol (Internet Protocol Address, IP) address of the user terminal device that initiates the search request, or the user identification number (Identity document, ID) that initiates the search request, and those skilled in the art can select according to the actual situation.

Step S120, acquiring at least two target search results according to the keyword information.

It should be noted that the embodiment of the present application does not limit the method of obtaining target search results based on keywords. The target search results can be obtained from the database according to the searched keyword information. Each candidate search result in the database has a mapping relationship with the keyword information. When new keyword information is obtained, the target search result corresponding to the new keyword information is obtained from the database according to the mapping relationship; it can also be understood that after the keyword information of the search request is obtained, the keyword information is matched with all candidate search results in the database, so as to obtain the target search result that matches the keyword information. Yes, the matching method of keyword information and candidate search results is well known to those skilled in the art, and may be implemented through text matching, which is not limited in this embodiment of the present application.

It can be understood that the obtained at least two target search results can be combined with user identifiers to provide a data basis for sorting the target search results.

Step S130, determining user preference information according to the user identifier, and determining rankings of at least two target search results according to the user preference information.

It can be understood that the user preference information can be obtained from the user preference information database corresponding to the user identifier according to the user identifier. The user preference information in the user preference information database is obtained according to the historical search data corresponding to the user identifier, such as analyzing and processing the historical browsing information and click information. When it is detected that the number of clicks on the historically browsed document A exceeds a preset threshold, the tag information of the document A is obtained, and the tag information of the document A is determined as the user preference information.

It can be understood that the user preference information can represent the interest information of the user corresponding to the user identifier. The technical solution of the present application is based on the matching and sorting of the target search results according to the keyword information in the related art, and adds the user preference information as a consideration parameter of the sorting process, so that the sorting of the target search results presented to the user can be more in line with the user's personalized search needs.

In addition, referring to FIG. 2, in one embodiment, step S130 in the embodiment shown in FIG. 1 also includes but is not limited to the following steps:

Step S210, obtaining historical search data from the log center according to the user identification;

Step S220, determining the search result type according to the historical search data;

Step S230, determining distribution information of search result types within at least one preset time period, and performing normalization processing on the distribution information to obtain user preference information.

It can be understood that the historical search data is obtained from the log center according to the user identification, and the search result type is determined according to the historical search data. The search result type is the category label of the historical browsing result in the historical search data. By determining the distribution information of the search result type in at least one preset time period, and normalizing the distribution information, the user preference information can be obtained, which can improve the real-time performance of the user preference information, enhance the ability to describe the user's browsing preference, thereby improving the matching degree between the target search result and the search demand, so as to meet the personalized search needs of the user.

It should be noted that the embodiment of the present application does not limit the rules for dividing the preset time periods and the number of preset time periods. It can be divided into four preset time periods according to 1 week, 1 month, 3 months, and 6 months. Those skilled in the art can choose according to the actual situation.

It is understandable that the distribution information represents the number of clicks on the browsed document corresponding to the search result type within the preset time period, and the historical search data in the log center will be continuously updated. When the number of clicks on the browsed document within the preset time period changes, the value of the user preference information in different preset time periods will also change accordingly. However, if the number of clicks on the category tag browsed document is accumulated according to the time dimension, the cumulative value will become larger and larger, causing the long-term interest weight to be too high, thereby weakening the timeliness of user interest. In order to ensure the timeliness of user preference information, the embodiment of the present application proposes a method for updating user preference information. The formula for updating user preference information can be expressed as: score _update =λ×score _old +(1-λ)score _new , wherein, score _update is the user preference information in the preset time period after updating; score _old is the user preference information in the preset time period before updating; score _new is the user preference information in the current preset time period, and the value of score _new is as follows The formula determines:

Among them, click _i is the number of times the user clicks on the browsing document corresponding to the i-th search result type within the current preset time period, L is a collection of search result types; λ is the attenuation factor, and the value range is [0, 1], and the value of λ is determined by the following formula:

Wherein, c is the number of clicks of the user within the current preset time period, and α _t is a preset attenuation coefficient.

It should be noted that, the embodiment of the present application does not limit the manner of normalizing the distribution information, and those skilled in the art can select according to the actual situation.

In addition, referring to FIG. 3 , in one embodiment, step S220 in the embodiment shown in FIG. 2 also includes but is not limited to the following steps:

Step S310, obtaining the search result identifier from the historical search data;

Step S320, obtaining historical search results from the search result database according to the search result identifier, and determining the search result type of the historical search results.

It can be understood that the search result database stores all the candidate search result resources of the search system, and is used to match the keyword information in the search request, so as to filter out target search results that match the keyword information. The candidate search results stored in the search result database carry search result type information. Obtaining the search result identifier from the historical search data, obtaining the historical search result from the search result database through the search result identifier, and determining the search result type of the historical search result can provide a data basis for obtaining user preference information.

In addition, referring to FIG. 4 , in one embodiment, step S130 in the embodiment shown in FIG. 1 also includes but is not limited to the following steps:

Step S410, obtaining a pre-trained first ranking model;

Step S420, input keyword information, user preference information, historical search results and target search results into the first ranking model to obtain sorted target search results.

It should be noted that the embodiment of the present application does not involve the improvement of the first ranking model, nor does it limit the type of ranking model applied. A click-through rate prediction model based on deep learning or machine learning can be selected, such as the DeepFM model, Wide&Deep model, LambdaMart model, etc. Those skilled in the art can choose according to the actual situation.

It can be understood that inputting keyword information, user preference information, historical search results, and target search results into the first ranking model can enable the first ranking model to sort the target search results according to the keyword information, user preference information, and historical search results according to preset rules, and adding user preference information and historical search data as consideration parameters for sorting processing can make the ranking of the target search results presented to the user more in line with the user's personalized search needs.

It should be noted that this embodiment of the present application does not limit the data processing method of inputting keyword information, user preference information, historical search results, and target search results into the first ranking model. Data may be extracted from keyword information, user preference information, and historical search results according to preset rules for splicing to obtain feature data, and then the feature data and target search results are input into the pre-trained first ranking model, so that the first ranking model can sort the target search results according to the feature data, thereby improving the matching degree of search results and search requirements.

In addition, referring to FIG. 5, in one embodiment, the obtaining of the first sorting model in the embodiment shown in FIG. 4 includes but is not limited to the following steps:

Step S510, obtaining training data according to keyword information, user preference information and historical search results;

In step S520, a preset second ranking model is obtained, and the second ranking model is trained according to the training data to obtain the first ranking model.

It can be understood that the training data obtained according to keyword information, user preference information and historical search results, and the first ranking model obtained by training the preset second ranking model according to the training data can provide a data basis for determining the ranking of target search results, so as to better meet the personalized search needs of users.

It should be noted that the embodiment of the present application does not limit the training method of the second ranking model, and those skilled in the art can adjust the training parameters of the model according to the actual situation.

It should be noted that the embodiment of the present application does not limit the way of obtaining training data according to keyword information, user preference information and historical search results. It may be the method steps shown in FIG. 6. Referring to FIG. 6, in one embodiment, the acquisition of training data in the embodiment shown in FIG.

Step S610, determining the first matching data according to the keyword information and the historical search results, the first matching data represents the matching score between the keyword information and the historical search results;

Step S620, associating the first matching data, historical search results and user preference information to obtain training data.

It can be understood that the matching score between the keyword information and the historical search results is determined, and the first matching data, historical search results and user preference information are associated to obtain the training data. The training data obtained through the above embodiment is subjected to data preprocessing and feature extraction, which can improve the data utilization rate of the training data in training the second ranking model.

In addition, referring to FIG. 7 , in one embodiment, before step S420 in the embodiment shown in FIG. 4 , it also includes but is not limited to the following steps:

Step S710, determining second matching data according to the keyword information and the target search result, where the second matching data represents the matching score between the keyword information and the target search result;

Step S720, sorting the target search results according to the second matching data to obtain the sorted target search results.

It can be understood that before inputting the keyword information, user preference information, historical search results and target search results into the first ranking model, the second matching data is determined according to the keyword information and the target search results, and the target search results are initially sorted according to the second matching data, and the sorted target search results are then input into the first ranking model along with the keyword information, user preference information and historical search results, which can improve the accuracy of the first ranking model and make the target search results processed by the first ranking model more in line with the individual needs of users.

In addition, referring to FIG. 8 , in one embodiment, after step S130 in the embodiment shown in FIG. 1 , it also includes but is not limited to the following steps:

Step S810, filter the sorted target search results.

It can be understood that, after the target search results are sorted, the embodiment of the present application proposes to filter the sorted target search results, which can effectively avoid duplicate data in the target search result list, or the target search results in the sorted target search result list have illegal content, or there are target search results for which the user corresponding to the user identifier does not have browsing authority, so as to improve the matching degree of the target search results and search requirements, thereby meeting the personalized search needs of users.

It should be noted that the embodiment of the present application does not limit the operation mode of filtering target search results. The preset filtering rules can be defined according to the needs of actual business scenarios to obtain the matching score between each target search result, and filter and filter the target search results according to the matching score. The user ID that has browsing authority for the target search result is matched and screened according to the current user ID and the authority verification information of each target search result to obtain the target search result that meets the browsing authority of the current user ID. Those skilled in the art can adjust and select according to the actual situation.

In addition, in order to describe the data synchronization method provided by the present application in more detail, the technical solutions of the present application are described below with three embodiments.

Referring to FIG. 9 , FIG. 9 is a block diagram of a search system provided by another embodiment of the present application. The search system 900 includes functional modules and data units. The functional modules include: a user interface module 901, a document recall module 902, a secondary sorting module 903, a document filtering module 904, an automatic document labeling module 905, a data preprocessing module 906, and a model training module 907; The functions of each module are described:

The user interface module 901, the function of the user interface module 901 is to interact with the user through the front-end page, the user inputs the search keyword information through the user interface module 901, the user interface module 901 transmits the search request to the background related module, finally receives the target search result list delivered by the background module, and displays it to the user on the front-end page. In addition, the user interface module 901 will record the user's historical search record data, including the search request submitted by the user, the search result information returned by the system, and the user's browsing information on the documents in the search result, etc., and transmit these information to the log center 908 for storage.

The document recall module 902, the function of the document recall module 902 is to initially match the search keyword information input by the user with all the document information in the document database 910, recall a small number of candidate documents, and provide a data basis for obtaining target search results that meet the user's personalized search needs.

Secondary ranking module 903. The function of the secondary ranking module 903 is to reorder the candidate documents recalled by the document recall module 902 through the ranking model. The secondary sorting module 903 includes a trained secondary sorting model. The model can choose a click-through rate prediction model based on deep learning or machine learning, such as the DeepFM model, Wide&Deep model, LambdaMart model, etc. The secondary sorting receives the various features delivered by the data preprocessing module 906, inputs them into the secondary sorting model, outputs the scoring of candidate documents, and sorts, and passes the sorted document list to the downstream document filtering module 904.

The document filtering module 904, the function of the document filtering module 904 is to filter the document list output by the secondary sorting module 903 based on certain filtering rules, such as removing duplicate documents in the document list, document auditing, that is, filtering operations such as removing illegal documents and documents that the user has no authority, and finally passing the filtered document list to the user interface module 901.

An automatic document labeling module 905. The function of the automatic document labeling module 905 is to perform automatic document labeling on documents in the database of the search system 900 through text classification technology in the field of natural language processing. The principle is to construct a labeling system that can accurately describe all document categories based on domain expert knowledge; then manual labeling of some documents by labelers who are familiar with documents and labeling systems to generate high-quality manual labeling data; then use the manual labeling data as training data, and use supervised learning methods to train text classification models; finally, use the trained text classification model to label all documents in the database, and perform corresponding operations on newly entered documents to ensure that all documents in the document database 910 of the search system 900 have their corresponding documents. Documentation tab.

The data preprocessing module 906, the function of the data preprocessing module 906 is divided into two parts: online reasoning scene and offline scene. In the online reasoning scenario, the data preprocessing module 906 receives the keyword information transmitted by the user interface module 901, the user preference information extracted from the user preference information database 909, and the corresponding information of the candidate documents in the document database 910, and performs feature extraction and preprocessing to generate the input features required by the online secondary ranking model; in the offline scenario, the data preprocessing module 906 receives the historical search data extracted from the log center 908 and the corresponding user preference information and document information, and performs feature extraction and preprocessing in the same way as in the online reasoning scenario. The processed model input features are combined with the document tags in the historical search data to form training data and sent to the model training module 907 for model training.

Model training module 907. The function of the model training module 907 is to use the training data to train the sorting model and update the model after offline training and offline testing to the online secondary sorting module 903.

The log center 908, the function of the log center 908 is to store all the search record data generated by the users of the search system 900 accessing the search system 900, including the search keyword information submitted by the user, the target search results returned by the system, and the browsing information of the documents in the search results by the user (whether to click and browse, browsing time, etc.) and other information. In addition, in the offline scenario, the log center 908 is responsible for providing user historical search data for subsequent modules to generate training data for the ranking model.

The user preference information database 909. The function of the user preference information database 909 is to maintain user preference information and update it regularly. The principle is to calculate the category label distribution of the clicked documents in the corresponding time dimension according to the user's historical click document data, and perform normalization processing respectively to obtain the user preference information of the current user in a given time dimension.

The document database 910, the function of the document database 910 is to store all the document resources of the search system 900, including document related information data, including document related information data, such as document title, document text, document tags and so on.

It should be noted that the method steps of Embodiment 1, Embodiment 2, and Embodiment 3 are applied to the search system 900 shown in FIG. 9 .

Embodiment 1, referring to FIG. 10, FIG. 10 is a flow chart of the steps of a ranking model training method provided by another embodiment of the present application. The ranking model training method includes the following steps:

In step S1010, the user interface module 901 transmits the historical search data to the log center 908 for storage. The historical search data includes the search keyword information input by the user, the search result list displayed to the user by the search system 900, and the user's click and browse information on documents in the search result list.

Step S1020, the data preprocessing module 906 obtains historical search data from the log center 908, and the format of the historical search data of the log center 908 can be as shown in Table 1:

Table 1 List of historical search data

Step S1030, the data preprocessing module 906 obtains the user identifier from the historical search data, and extracts user preference information from the user preference information database 909 according to the user identifier. The user preference information represents the preference score of the document tag corresponding to the historical search document within the preset time period. For example, assuming that the document tag system in the automatic document tagging module 905108 has three tags (such as three categories of news, technical literature, and rules and regulations), and the preset time period takes four dimensions of 5 days, 10 days, 30 days, and 60 days.

Step S1040, the data preprocessing module 906 obtains the search result document identifier from the historical search data, and obtains the corresponding document information from the document database 910 according to the search result document identifier. The document information includes the title of the document, the text of the document, the category label of the document, and the historical click volume information of the document.

In step S1050, the data preprocessing module 906 performs feature extraction, preprocessing, and splicing operations according to the data obtained from the log center 908, the user preference information database 909, and the document database 910 respectively, that is, historical search data, user preference information, and document information, to generate training data. The format of the training data can be as shown in Table 2:

样本标识Sample ID	用户标识User ID	文档标识document identification	特征向量Feature vector	样本标签sample label
00	User101User101	Doc 103Doc 103	[0.1,0.3,...][0.1,0.3,...]	11
11	User101User101	Doc 101Doc 101	[1.0,0.4,...][1.0,0.4,...]	00

Table 2 Training data list

Among them, the sample identification is the serial number of the training sample; the user identification and the document identification are the unique identification numbers of the user and the document respectively; the feature vector is spliced from the features extracted from the input data, including matching features, user preference information and document features. , generally use whether the user clicks on the corresponding document as the sample label, 1 is clicked, and 0 is not clicked. The data preprocessing module 906 integrates all training samples, generates training data and sends it to the model training module 907 .

Step S1060 , the model training module 907 uses the training data obtained from the data preprocessing module 906 to train the ranking model, and the trained ranking model is used to update the ranking model in the model training module 907 .

Embodiment 2, refer to FIG. 11 . FIG. 11 is a flowchart of steps of a method for sorting search results provided in another embodiment of the present application. The method for sorting search results includes the following steps:

Step S1110, the user interface module 901 interacts with the user of the search system 900 through the front-end page, and the user submits his keyword information through the search dialog box on the page. After clicking the search button, the user interface module 901 passes the keyword information to the document recall module 902.

In step S1120, the document recall module 902 performs preliminary matching according to the keyword information and all document information in the document database 910, for example, using the matching score between the keyword information and the document title, sorting and recalling some documents according to the scores as the candidate target document list, and sending the target document list to the secondary ranking module.

Step S1130, the data preprocessing module 906 performs feature extraction, preprocessing, and splicing operations according to the keyword information transmitted by the user interface module 901, the user preference information provided by the user preference information database 909, and the document information in the candidate document list provided by the document database 910, to generate a feature vector with the same format as the training data, and pass it to the secondary ranking module 903.

In step S1140, the secondary ranking module 903 inputs the feature vectors delivered by the data preprocessing module 906 into the pre-trained ranking model to generate the ranking scores of the documents in the candidate document list, and reorder them according to the ranking scores to generate a sorted target document list.

In step S1150, the document filtering module 904 filters the sorted target document list according to preset filtering rules. The filtering rules generally include removing duplicate documents, removing illegal content documents, removing documents that the user does not have permission to browse, etc., and the filtered document list is passed to the user interface module 901 as a target search result list.

In step S1160, the user interface module 901 displays the target search result list returned by the document filtering module 904 to the user on the page, and the user can click to browse the corresponding document according to their needs.

Embodiment 3, referring to FIG. 12 , FIG. 12 is a flow chart of the steps of labeling document information provided by another embodiment of the present application. The application scenario of the step of labeling document information is the data interaction between the automatic document labeling module 905 and the document database 910. The steps of labeling document information are as follows:

In step S1210, the automatic document labeling module 905 acquires a pre-labeled document data set, and the document data set carries document tags.

In step S1220, the automatic document labeling module 905 obtains a preset text classification model, and trains the text classification model according to the document data set to obtain a document labeling model.

Step S1230, the automatic document labeling module 905 obtains document information from the document database 910, inputs the document information into the document labeling model, and obtains document information carrying document labels.

Step S1240 , the automatic document labeling module 905 stores the document information carrying the document label into the document database 910 .

In addition, referring to FIG. 13 , FIG. 13 is a schematic structural diagram of a search system 1300 provided by another embodiment of the present application. An embodiment of the present application also provides a search system 1300 .

The processor 1320 and the memory 1310 may be connected through a bus or in other ways.

The non-transitory software programs and instructions required to realize the search result sorting method of the above-mentioned embodiment are stored in the memory 1310, and when executed by the processor 1320, the search result sorting method in the above-mentioned embodiment is executed, for example, the method step S110 to method step S130 in Fig. 1 described above, the method step S210 to method step S230 in Fig. 2 , the method step S310 to method step S320 in Fig. 3 , and the method step S410 to method step in Fig. 4 S420, method step S510 to method step S520 in FIG. 5 , method step S610 to method step S620 in FIG. 6 , method step S710 to method step S720 in FIG. 7 , method step S810 in FIG. 8 .

The device embodiments described above are only illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place, or may be distributed to multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, an embodiment of the present application also provides a computer-readable storage medium, the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are executed by a processor 1320 or a controller, for example, by a processor 1320 in the above-mentioned embodiment of the search system 1300, so that the above-mentioned processor 1320 can execute the method for sorting search results in the above-mentioned embodiment, for example, execute the above-described method step S110 to method step S130 in FIG. 1 and method step S210 in FIG. 2 To method step S230, from method step S310 to method step S320 in FIG. 3, from method step S410 to method step S420 in FIG. 4, from method step S510 to method step S520 in FIG. 5, from method step S610 to method step S620 in FIG. 6, from method step S710 to method step S720 in FIG. Those skilled in the art can understand that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware and an appropriate combination thereof. Some or all of the physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on computer readable media, which may include computer storage media (or non-transitory media) and communication media (or transitory media). As is well known to those of ordinary skill in the art, the term computer storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cartridges, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and that can be accessed by a computer. In addition, communication media typically embody computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism, and can include any information delivery media, as is known to those of ordinary skill in the art.

Embodiments of the present application include a search result ranking method, a search system, and a computer-readable storage medium, wherein the search result ranking method includes: obtaining a search request, the search request including keyword information and user identification; obtaining at least two target search results according to the keyword information; determining user preference information according to the user identification, and determining the ranking of at least two target search results according to the user preference information. According to the solution provided by the embodiment of the present application, the user preference information is determined according to the user identification, and the ranking of the search results is determined according to the user preference information. Compared with the current technical solution for matching and sorting the target search results based only on keyword information, the technical solution of the present application can improve the matching degree between the target search results and the search needs, thereby meeting the personalized search needs of users.

The above is a description of several implementations of the present application, but the application is not limited to the above-mentioned implementations. Those skilled in the art can also make various equivalent deformations or replacements without violating the spirit of the application. These equivalent deformations or replacements are all included within the scope defined by the claims of the application.

Claims

A method for sorting search results, comprising:

Obtain a search request, the search request includes keyword information and user identification;

Acquiring at least two target search results according to the keyword information;

User preference information is determined according to the user identifier, and ranking of at least two of the target search results is determined according to the user preference information.
The method according to claim 1, wherein said determining user preference information according to said user identifier comprises:

Acquiring historical search data from the log center according to the user identification;

determining the search result type according to the historical search data;

determining the distribution information of the search result type within at least one preset time period;

Perform normalization processing on the distribution information to obtain the user preference information.
The method according to claim 2, wherein said determining the search result type according to the historical search data comprises:

Obtain the search result identifier from the historical search data;

Acquiring historical search results from the search result database according to the search result identification, and determining the search result type of the historical search results.
The method according to claim 3, wherein said determining the ranking of at least two of said target search results according to said user preference information comprises:

Obtain the pre-trained first ranking model;

Inputting the keyword information, the user preference information, the historical search results and the target search results into the first ranking model to obtain the sorted target search results.
The method according to claim 4, wherein the first ranking model is obtained through the following training steps:

obtaining training data according to the keyword information, the user preference information and the historical search results;

A preset second ranking model is obtained, and the second ranking model is trained according to the training data to obtain the first ranking model.
The method according to claim 5, wherein said obtaining training data according to said keyword information, said user preference information and said historical search results comprises:

determining first matching data according to the keyword information and the historical search results, the first matching data characterizing the matching score between the keyword information and the historical search results;

The training data is obtained by associating the first matching data, the historical search results and the user preference information.
The method according to claim 4, wherein, before said inputting said keyword information, said user preference information, said historical search results and said target search results into said first ranking model, further comprising:

determining second matching data according to the keyword information and the target search result, where the second matching data characterizes a matching score between the keyword information and the target search result;

The target search results are sorted according to the second matching data to obtain the sorted target search results.
The method according to claim 1, wherein, after determining the ranks of at least two of the target search results according to the user preference information, further comprising:

Filtering is performed on the sorted target search results.
A search system, comprising: a memory, a processor, and a computer program stored in the memory and operable on the processor, when the processor executes the computer program, the search result sorting method according to any one of claims 1 to 8 is realized.
A computer-readable storage medium storing computer-executable instructions, the computer-executable instructions being used to execute the method for ranking search results according to any one of claims 1-8.