Search prompt word determination method, system and information processing method

By employing federated learning on both the client and server sides, and leveraging data anonymization and local learning models, the problem of providing personalized search suggestions while protecting user privacy is solved, achieving secure and effective search term recommendation.

CN113032819BActive Publication Date: 2024-11-12TAOBAO CHINA SOFTWARE
View PDF 8 Cites 0 Cited by

Patent Information

Application Number
CN201911251642.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2019-12-09
Publication Date
2024-11-12
Estimated Expiration
2039-12-09

AI Technical Summary

Technical Problem

Existing technologies struggle to provide personalized and secure search term suggestions while protecting user privacy, and traditional methods cannot balance the search performance of personalized search terms with the security of non-personalized search terms.

Method used

By employing a client-server federated learning approach, model parameters and candidate suggestions are received and aggregated. The local learning model is then used for ranking and recommendation, ensuring data anonymization and providing personalized search suggestions while protecting user privacy.

Benefits of technology

It achieves the effect of providing personalized search suggestions while protecting user privacy, and at the same time improves the security of search term recommendations and avoids the risk of data leakage.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113032819B_ABST
    Figure CN113032819B_ABST
Patent Text Reader

Abstract

The present application discloses a method, system and information processing method for determining search prompt words. The method includes: a client device receives a model parameter and a first candidate prompt word set sent by a server, wherein the first candidate prompt word set includes candidate prompt words selected by the server from the candidate prompt word set based on data of a specified type uploaded by the client; the client device updates a local learning model according to the model parameter, and determines a ranking result of the candidate prompt words in the first candidate prompt word set based on the updated local learning model; the client device determines a target prompt word from the first candidate prompt word set based on the ranking result; and displays the target prompt word.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular, to a method and system for determining search prompt words and an information processing method. Background Art

[0002] In recommendation systems such as e-commerce product recommendations, search advertising, and information flow recommendations, due to the needs of recommendation algorithms, the system often collects a large amount of user information, such as user search records, user click records, user age, gender, location, etc., and provides a unique identifiable number to upload to the enterprise server to help the server-side train the model and make more accurate recommendations.

[0003] With the development of the Internet, user data privacy protection has become an increasingly important issue facing Internet companies. Traditional personalized search term suggestions need to collect users' search and click histories and save them in the cloud, which creates risks for large-scale data leakage. Storing user data in a uniquely identifiable form on the server poses a challenge to user data privacy protection, while non-personalized search term suggestions are less effective. In addition, in order to protect user data privacy, users' personal data must be uploaded and used only after explicit authorization by the user, which also brings uncertainty to data collection on the server side.

[0004] In order to allow enterprise service providers to better provide privacy protection for users while using user data, the following related technologies have emerged: 1. Differential privacy technology; 2. Acquiring data through shared models without sharing plaintext data between retailers; 3. Performing model prediction on the terminal, with the server only performing parameter aggregation. However, these methods are difficult to adapt to product-level search.

[0005] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0006] The embodiments of the present application provide a method, system and information processing method for determining a search prompt word, so as to at least solve the technical problem of not being able to simultaneously take into account the search effect of personalized search words and the security of non-personalized search word suggestion solutions.

[0007] According to one aspect of an embodiment of the present application, a method for determining a search prompt word is provided, comprising: a client device receiving a model parameter and a first candidate prompt word set sent by a server, the first candidate prompt word set including candidate prompt words selected by the server from the candidate prompt word set based on data of a specified type uploaded by the client; the client device updating a local learning model according to the model parameters, and determining a ranking result of the candidate prompt words in the first candidate prompt word set based on the updated local learning model; the client device determining a target prompt word from the first candidate prompt word set based on the ranking result; and displaying the target prompt word.

[0008] According to another aspect of an embodiment of the present application, a method for determining a search prompt word is also provided, including: a server obtains model parameters of a learning model in different client devices, and the different client devices run the same type of search engine provided by the server; the server aggregates the model parameters corresponding to the different client devices to obtain target model parameters; the server receives data of a specified type reported by the client device, and selects candidate prompt words from a candidate prompt word set based on the data of the specified type to obtain a first candidate prompt word set; the server sends the first candidate prompt word set and the target model parameters to the client device.

[0009] According to another aspect of an embodiment of the present application, a method for determining a search prompt word is also provided, including: a client device receives a candidate prompt word set sent by a server; the client device inputs the candidate prompt word set into a learning model for analysis to obtain a scoring index for the candidate prompt word; the client device determines a target prompt word from the candidate prompt word set based on the scoring index; and displays the target prompt word.

[0010] According to another aspect of an embodiment of the present application, a system for determining a search prompt word is also provided, including: a client device and a server, wherein the server is used to obtain a first candidate prompt word set by selecting a prompt word from a candidate prompt word set based on specified type data uploaded by the client device, and send model parameters to the client device; the client device is used to receive the model parameters and the first candidate prompt word set; update a local learning model according to the model parameters, and determine a ranking result of the candidate prompt words in the first candidate prompt word set based on the updated local learning model; determine a target prompt word from the first candidate prompt word set based on the ranking result; and display the target prompt word.

[0011] According to another aspect of an embodiment of the present application, a system is also provided, including: a processor; and a memory, connected to the processor, for providing the processor with instructions for processing the following processing steps: receiving model parameters and a first candidate prompt word set sent by a server, wherein the first candidate prompt word set includes candidate prompt words selected by the server from the candidate prompt word set based on the specified type of data uploaded by the client; updating a local learning model according to the model parameters, and determining a ranking result of the candidate prompt words in the first candidate prompt word set based on the updated local learning model; determining a target prompt word from the first candidate prompt word set based on the ranking result; and displaying the target prompt word.

[0012] According to another aspect of an embodiment of the present application, another method for determining a search prompt word is provided, including: a client device receives a first candidate prompt word set from a server, wherein the first candidate prompt word set is determined by candidate prompt words selected by the server from a preset candidate prompt word set based on a first learning model; the client device determines a target prompt word from the first candidate prompt word set based on a local second learning model; and displays the target prompt word.

[0013] According to another aspect of an embodiment of the present application, another method for determining a search prompt word is provided, comprising: a client device receives a first candidate prompt word set from a server, wherein the candidate prompt words in the first candidate prompt word set are arranged in a predetermined order; the client device determines a target prompt word based on the sorting result of the candidate prompt words in the first candidate prompt word set; and displays the target prompt word.

[0014] In the embodiment of the present application, the client device receives the model parameters and the first candidate prompt word set sent by the server, the first candidate prompt word set includes the candidate prompt words selected by the server from the candidate prompt word set based on the specified type of data uploaded by the client; the client device updates the local learning model according to the model parameters, and determines the ranking results of the candidate prompt words in the first candidate prompt word set based on the updated local learning model; the client device determines the target prompt word from the first candidate prompt word set based on the ranking result; and displays the target prompt word. The above scheme is adopted to achieve the purpose of not directly using the user's personalized data to provide search suggestions to the user, thereby achieving the technical effect of protecting the user's privacy when providing search suggestions to the user based on the user data, and further solving the technical problem of not being able to take into account the search effect of personalized search words and the security of non-personalized search word suggestion schemes at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0016] Figure 1a is a flow chart of an information processing method according to an embodiment of the present application;

[0017] Figure 1b It is a hardware structure block diagram of a computer terminal according to a method for determining a search prompt word in an embodiment of the present application;

[0018] Figure 2a is a flow chart of a method for determining a search prompt word according to an embodiment of the present application;

[0019] Figure 2b is a schematic diagram of an application scenario of a method for determining a search prompt word according to an embodiment of the present application;

[0020] Figure 3 is a flowchart of another method for determining a search prompt word according to an embodiment of the present application;

[0021] Figure 4 is a flowchart of another method for determining a search prompt word according to an embodiment of the present application;

[0022] Figure 5 is a schematic diagram of a system for determining a search prompt word according to an embodiment of the present application;

[0023] Figure 6 is a flowchart of another method for determining a search prompt word according to an embodiment of the present application;

[0024] Figure 7 is a flowchart of another method for determining a search prompt word according to an embodiment of the present application;

[0025] Figure 8 is a schematic diagram of a client device according to an embodiment of the present application;

[0026] Fig. 9 is a schematic diagram of another client device according to an embodiment of the present application;

[0027] Fig.10 is a schematic diagram of another system for determining search prompt words according to an embodiment of the present application;

[0028] Fig.11 is a schematic diagram of a device for determining a search prompt word according to an embodiment of the present application;

[0029] Fig.12is a schematic diagram of another device for determining a search prompt word according to an embodiment of the present application;

[0030] Fig.13 is a schematic diagram of another device for determining a search prompt word according to an embodiment of the present application;

[0031] Fig.14 is a schematic diagram of another device for determining a search prompt word according to an embodiment of the present application;

[0032] Fig.15 is a schematic diagram of another device for determining a search prompt word according to an embodiment of the present application;

[0033] Fig.16 is a schematic diagram of an information processing device according to an embodiment of the present application; and

[0034] Fig.17 It is a hardware structure block diagram of a computer terminal according to an embodiment of the present application. DETAILED DESCRIPTION

[0035] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution 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. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0036] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0037] First, some nouns or terms that appear in the description of the embodiments of the present application are subject to the following explanations:

[0038] Model parameters: are configuration variables within the model whose values ​​can be estimated from the data. The values ​​of model parameters define the model that can be used. They are estimated or learned from the data, usually not set manually by the programmer, and are saved as part of the learned model. They are usually summarized from past training data.

[0039] Differential privacy algorithm: A method for protecting data privacy by adding random perturbations.

[0040] Federated learning algorithm: a machine learning algorithm that performs large-scale distributed training with linkage between client and server.

[0041] General Data Protection Regulation: A regulation on user data protection issued and implemented by the European Union.

[0042] In the related art, there are two ways to determine the search word prompt: traditional personalized search word suggestions and non-personalized search word suggestions. Among them, the personalized search word suggestion solution: saves user search records and click records in the cloud, has the risk of large-scale data leakage, and will be subject to increasingly stringent privacy protection laws; while the non-personalized search word suggestion solution, although the privacy security is better, the search word recommendation effect is poor.

[0043] To solve the above problems, in the embodiment of the present application, the client and the server are used to perform federated learning to determine the search prompt words. The specific principle is as follows: Figure 1a As shown, the following steps are included: S11, the client device receives the first data from the server, wherein the first data is determined by the server after analyzing the specified type of data based on the first learning model; S12, the client device processes the first data based on the local second learning model to obtain the second data; S13, outputs the second data. For example, the communication data sent to the server side is anonymized on the client side, and at the same time, the server side uses data aggregation to eliminate the sample data noise caused by the differential privacy technology to ensure data availability. Based on the above principles, the present application provides the following embodiments to explain in detail the process of determining the search prompt word.

[0044] Example 1

[0045] According to an embodiment of the present application, an embodiment of a method for determining a search prompt word is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0046] The method embodiment provided in the first embodiment of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1b The hardware structure block diagram of a computer terminal (or mobile device) for implementing a method for determining a search prompt word is shown. Figure 1b As shown, the computer terminal 10 (or mobile device 10) may include one or more (only one is shown in the figure) processors 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device for communication functions. It can be understood by those skilled in the art that Figure 1b The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1b More or fewer components as shown, or with Figure 1b Different configurations shown.

[0047] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the method for determining the search prompt word in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, the method for determining the search prompt word of the above-mentioned application is realized. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0048] The transmission device is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computer terminal 10. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0049] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 (or mobile device).

[0050] Under the above operating environment, this application provides Figure 2a The method for determining the search prompt word shown in the figure includes:

[0051] S21, the client device receives the model parameters and the first candidate prompt word set sent by the server, wherein the first candidate prompt word set includes candidate prompt words selected by the server from the candidate prompt word set based on the specified type of data uploaded by the client. The model parameters include but are not limited to weights and bias parameter values ​​in the machine learning model.

[0052] It should be noted that the server is used to provide a search engine to the client device, and the user searches for information based on the search engine on the client device.

[0053] The model parameters sent by the server are obtained by aggregating the model parameters reported by each client device under the same search engine. In this way, the client device can obtain relatively reliable model parameters. Aggregation includes but is not limited to combining multiple model parameters, or taking the average value of each model parameter, or taking the weighted average of each model parameter.

[0054] The candidate prompt words in the first candidate prompt word set are selected by the server from the candidate prompt word set maintained locally. Specifically, the selection conditions are obtained according to the specified type of data uploaded by the client, and the candidate prompt words are selected from the candidate prompt word set according to the selection conditions, wherein the candidate words in the candidate prompt word set can be related recommended candidate words stored locally by the client after the user enters the search word. The specified type of data may include anonymous data, which is data obtained by processing the collected personalized data of the user using a privacy protection algorithm. The above selection conditions include but are not limited to: click-through rate, etc. In addition, it should be noted that when selecting candidate prompt words, in addition to selecting using the above method, the server can also select using a machine learning model, using coarse-grained user features and related search word exposure and click logs to train machine learning models such as logistic regression, wherein the coarse-grained user features include: age, gender, purchasing power and other features.

[0055] In addition, in some embodiments of the present application, when the client device uploads data to the server, the granularity of the data uploaded can be configured according to user needs, that is, one type of data can be selected for upload from coarse-grained data (i.e., anonymized data) and fine-grained data (i.e., personalized data), for example: the client device receives a selection instruction from the user; based on the selection instruction, the data granularity corresponding to the selection instruction is determined; and then the data of the corresponding granularity is uploaded based on the selected data granularity. There are multiple ways to implement the client device receiving the user's selection instruction, such as directly entering the instruction in the command input box, and another example: the client device displays an interactive interface, in which options for selecting coarse-grained and fine-grained data are provided; the data granularity finally selected is determined based on the option selected by the user.

[0056] S22: The client device updates the local learning model according to the model parameters, and determines a ranking result of the candidate prompt words in the first candidate prompt word set based on the updated local learning model.

[0057] There are many ways to obtain the ranking results, for example: the client device inputs the candidate prompt words in the first candidate prompt word set into the local learning model for analysis to obtain the scoring index of the candidate prompt words; the client device sorts the candidate prompt words in the first candidate prompt word set based on the scoring index to obtain the ranking result; wherein the scoring index is used to score the candidate words and recommend the candidate words to the user based on the scoring result. The local learning model needs to be continuously updated to accurately train the scoring index. As described above, the local learning model can be updated based on the model parameters from the server.

[0058] Among them, when using the local learning model to determine the ranking results, it can be divided into two stages: 1) Use the personalized data (fine-grained features) of the user stored locally to determine whether the target prompt word is clicked (selected): user ID, user search word record, city, current location, age, gender, purchasing power and other features, and the search word prefix entered by the user, the search word list exposed by the system, the search word clicked by the user and other data to train the local learning model. After the training is completed, the local learning model can output the output result of whether the target prompt word is clicked, that is, determine whether the target prompt word is selected. 2) Count the output results used to indicate whether the search word is selected, and determine the frequency of use of the target prompt word based on the statistical results; determine the above-mentioned scoring index based on the frequency of use, for example, the mapping relationship between the frequency of use and the scoring index can be used to determine the scoring index corresponding to the current frequency of use, wherein the higher the frequency of use, the larger the scoring index; or, the frequency of use can be directly used as the scoring index.

[0059] S23: The client device determines a target prompt word from the first candidate prompt word set based on the scoring index.

[0060] For example, the candidate prompt words in the first candidate prompt word set are sorted based on the scoring index to obtain a sorting result, and the top N candidate prompt words in the sorting result are used as target prompt words.

[0061] In some embodiments of the present application, the scoring index of the target prompt word can be obtained from multiple dimensions. For example, different types of data corresponding to the prompt word are input into the local learning model, and multiple scoring indexes are output. At this time, the scores of each scoring index can be weighted averaged to determine the final scoring index of the target prompt word. In other embodiments of the present application, different types of data can be input into the local learning model, and a final scoring index can be output using the internal processing of the local learning model.

[0062] S24, display the target cue word.

[0063] It should be noted that the target prompt word is the search prompt word recommended to the user, and the user can search for information based on the target prompt word.

[0064] Specifically, the target prompt word may be displayed in a display interface of the client device, for example, the target prompt word may be displayed in a drop-down list of a search box for selection by the user.

[0065] In some embodiments of the present application, the above-mentioned specified type data can be determined in the following manner: the client device obtains personalized data of the target object, wherein the personalized data includes at least search behavior information of the target object and identity information used to uniquely identify the target object; the client device uses a privacy protection algorithm to process the personalized data to obtain non-personalized data, and uses the non-personalized data as the specified type data.

[0066] In one embodiment, the target object is a user, and the personalized data of the target object is the fine-grained feature data of the user. The identity information used to uniquely identify the target object in the personalized data may include user ID, device ID, IP, user's word embedding vector, user's current location, user's age, user's gender and other feature information. The search behavior information in the personalized data may include currently input keywords and historical search behavior log information, specifically, it may include purchasing power obtained based on purchase data, user search term records, search term prefixes entered by the user, a list of search terms exposed by the system, search terms clicked by the user and other feature information.

[0067] It should be noted that in order to prevent excessive memory usage, the client device only stores part of the data within a preset period, for example, data within the past three months.

[0068] In addition, in order to protect the privacy of the user, after the client device obtains the personalized data of the target object, it uses a privacy protection algorithm to process the personalized data. Specifically, a differential privacy algorithm can be used, or other privacy protection noise algorithms can be used to remove the data that can uniquely identify the user after being reported to the server in the personalized data, and obtain non-personalized data. The obtained non-personalized data is reported to the server as a specified type of data. Specifically, the data that can uniquely identify the user after being reported to the server includes but is not limited to user ID, device ID, IP, user's word embedding vector and other data.

[0069] According to the above-mentioned embodiment of the present application, before the client device sends the specified type of data to the server, the method also includes: the client device trains the local learning model based on the personalized data, and counts the number of training times of the local learning model in the current statistical period; the client device compares the number of training times and a preset threshold; when the number of training times is greater than the preset threshold, the client device determines that it is allowed to send the specified type of data to the server.

[0070] It should be noted that the local learning model is used to obtain the scoring indicators of the candidate prompt words. The basic data for training the local learning model is non-personalized data. The model is continuously trained and updated with the obtained non-personalized data to obtain a model that can accurately obtain the scoring indicators. When the number of training times reaches the preset threshold, it means that the local model meets the requirements, the model parameters are uploaded to the server, and the specified type of data is allowed to be sent to the server.

[0071] According to the above-mentioned embodiment of the present application, the search behavior information includes: the historical selection record of the target object for the prompt word, the search word prefix input by the target object, and the prompt word list exposed by the server.

[0072] It should be noted that when the target object selects a prompt word, a selection record will be generated, and the search behavior information includes the user's historical selection data; when the user searches, he will first enter the prefix of the search word, and the search behavior information includes the search word prefix entered by the user; before the user enters the search word, the server can provide a prompt word list for the user to choose, and the search behavior information includes the prompt word list exposed by the server. In addition, the search behavior information also includes other data related to the user's search behavior.

[0073] According to the above-mentioned embodiment of the present application, the client device determines the target prompt word from the first candidate prompt word set based on the scoring index, including: the client device sorts the candidate prompt words in the first candidate prompt word set based on the scoring index, and takes the top N candidate prompt words as the target prompt words.

[0074] It should be noted that the candidate prompt words in the first candidate prompt word set are sorted according to the determined evaluation index and the selected sorting algorithm. Specifically, the sorting method can be a point-to-point method, a pair method and a list method. After sorting, the top candidate prompt words are used as target prompt words and displayed for user selection.

[0075] In addition, after displaying the target keyword, the displayed target prompt word is saved, and information such as which target prompt words the user clicks is stored, providing a data basis for subsequent collection of personalized data.

[0076] According to the above-mentioned embodiment of the present application, the model parameters received by the client device are obtained by aggregating the model parameters of the local learning model of the client device and the model parameters of the learning models of other client devices. The other client devices share the same type of search engine provided by the server with the client device.

[0077] It should be noted that the server provides a search engine to multiple clients. The server receives model parameters of the learning model uploaded by multiple clients of the same type of search engine, and aggregates the model parameters. Specifically, the federated average update algorithm can be used for aggregation, or other distributed algorithms can be used for aggregation. The aggregated model parameters are used as the model parameters sent to the client device to update the local learning model of the client device.

[0078] like Figure 2b As shown, in some optional embodiments of the present application, the client device 2 displays a search box 20 to the user, wherein after the user enters the keyword (i.e., search word) "parameter", the client device requests the latest model parameters from the server 3, wherein the latest model parameters are obtained by the server based on the aggregation of model parameters uploaded by the client device 2 and other client devices communicating with the server; the client device updates the local learning model based on the received latest model parameters, and uses the updated local learning model to determine the candidate prompt word set corresponding to the current keyword, and uses the candidate prompt word set to select the candidate prompt word set from the candidate prompt word set. Figure 2a The scheme shown determines and displays the target prompt words, that is, displays search prompt words such as "what is parameter", "parametric equation" and "parametric equation derivative" in the drop-down box of the search box 20; after determining the target prompt word, the client device 2 can provide feedback on the final output result of the local learning model, and adjust the model parameters of the local learning model according to the feedback result; upload the adjusted model parameters to the server 3. When the client device 2 uploads the model parameters, it needs to be anonymized to obtain anonymous data, and the anonymous data is sent to the server 3.

[0079] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0080] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0081] Example 2

[0082] This application provides Figure 3 The method for determining the search hint word shown. Figure 3 It is a flowchart of a method for determining a search prompt word according to the second embodiment of the present application.

[0083] S31, the server obtains model parameters of learning models in different client devices, and the different client devices run the same type of search engine provided by the server.

[0084] It should be noted that the server is used to provide a search engine to the client, and the user searches for information based on the search engine on the client. The server receives model parameters of the learning model uploaded by multiple clients of the same type of search engine.

[0085] S32, the server aggregates the model parameters corresponding to different client devices to obtain target model parameters.

[0086] Specifically, a federated average update algorithm may be used for aggregation, or other distributed algorithms may be used for aggregation, and the model parameters obtained by aggregation are used as model parameters sent to the client device to update the local learning model of the client device.

[0087] S33: The server receives the designated type data reported by the client device, and selects a candidate prompt word from the candidate prompt word set based on the designated type data to obtain a first candidate prompt word set.

[0088] It should be noted that the designated type of data is data obtained by processing the collected personalized data of the user using a privacy protection algorithm, wherein the personalized data at least includes the search behavior information of the target object and the identity information used to uniquely identify the target object.

[0089] The candidate words in the candidate prompt word set may be related recommended candidate words stored locally on the client after the user inputs a search word. The candidate prompt words are selected from the candidate prompt word set according to the designated type data uploaded by the client to obtain the first candidate prompt word set.

[0090] S34, the server sends the first candidate prompt word set and the target model parameters to the client device.

[0091] Specifically, the first candidate prompt word set and the target model parameters are sent to the client device to update the local learning model according to the model parameters, and the candidate prompt words in the first candidate prompt word set are input into the local learning model for analysis to obtain scoring indicators of the candidate prompt words, so as to score the candidate words and recommend the candidate words to the user according to the scoring results.

[0092] In some embodiments of the present application, when selecting candidate prompt words from the candidate prompt words based on the specified type data to obtain a first candidate prompt word set, it can be achieved in the following way: sorting the candidate prompt words in the candidate prompt word set based on the specified type data; and selecting the top N candidate prompt words as the selected candidate prompt words.

[0093] It should be noted that the selection condition is obtained according to the designated type of data uploaded by the client, and the candidate prompt words are selected from the candidate prompt word set according to the selection condition to obtain the first candidate prompt word set. For example, the selection condition such as the click rate and popularity of the candidate words can be obtained according to the designated type of data, and the candidate words in the candidate prompt word set are sorted, and the top N candidate prompt words are taken as the selected candidate prompt words to form the first candidate prompt word set; wherein, before sorting, different types of data reflecting the selection condition can also be classified using a learning model, for example: the learning model can be a learning model based on a logistic regression algorithm, etc., which can be used to classify the data uploaded by the client, and then sort the preset candidate prompt words based on the classified data. For example, the server classifies the received non-personalized data (i.e., anonymized data, such as the click rate and popularity of the candidate words), and then sorts the preset candidate prompt words based on different types of data, and takes the top N candidate prompt words as the selected candidate prompt words to form the first candidate prompt word set.

[0094] The preferred implementation methods of the embodiments of the present application can be found in the relevant descriptions in the embodiments, which will not be repeated here.

[0095] Example 3

[0096] This application provides Figure 4 The method for determining the search hint word shown. Figure 4 It is a flowchart of a method for determining a search prompt word according to Embodiment 3 of the present application.

[0097] S41, the client device receives a candidate prompt word set sent by the server.

[0098] It should be noted that the server is used to provide a search engine to the client, and the user searches for information on the client based on the search engine. The candidate words in the candidate prompt word set can be related recommended candidate words stored locally on the client after the user enters the search word.

[0099] S42, the client device inputs the candidate prompt word set into the learning model for analysis to obtain a scoring index for the candidate prompt word.

[0100] It should be noted that the learning model of the client device is used to obtain the scoring index of the candidate prompt word, and the scoring index of the candidate prompt word is used to score the candidate word and recommend the candidate word to the user according to the scoring result.

[0101] S43, the client device determines a target prompt word from the candidate prompt word set according to the scoring index.

[0102] It should be noted that there may be multiple scoring indicators, and different target prompt words may be determined from the first candidate prompt words according to different scoring indicators, or the scores of various scoring indicators may be weighted averaged to determine the target prompt word.

[0103] S44, display the target cue word.

[0104] It should be noted that the target prompt word is the search prompt word recommended to the user, and the user can search for information based on the target prompt word.

[0105] Specifically, the target prompt word may be displayed in a display interface of the client device, for example, the target prompt word may be displayed in a drop-down list of a search box for selection by the user.

[0106] Example 4

[0107] This application provides Figure 5 The system for determining the search prompt words shown. Figure 5 It is a schematic diagram of a system for determining search prompt words according to Embodiment 4 of the present application.

[0108] Specifically, it includes a server and a client for data interaction. The server includes a specified type data storage module, a non-personalized sorting model training module, a non-personalized sorting module, a personalized model parameter aggregation module, and a data storage and distribution module. The client includes a personalized data storage module, a data anonymization module, a personalized module training module, and a personalized sorting module.

[0109] Among them, on the server side, the specified type data storage module is used to receive the specified type data sent by the client. Specifically, the specified type data is the data obtained after processing the collected user's personalized data using a privacy protection algorithm, including the processed coarse-grained user feature data such as age, gender, purchasing power, and the search term prefix entered by the user, the search term list exposed by the system, the search term clicked by the user and other data, but does not include user ID, device ID, IP and other data that may uniquely identify the user.

[0110] The non-personalized sorting model training module is used to train the non-personalized sorting model using data of a specified type.

[0111] The personalized module training module is used to select candidate prompt words from the candidate prompt word set based on the specified type of data uploaded by the client, and transmit the candidate prompt words back to the client.

[0112] The personalized model parameter aggregation module is used to receive the model parameters processed by the privacy protection algorithm reported by the client, and use algorithms such as the federated average update algorithm to update the personalized module training module.

[0113] The data storage and distribution module is used to store the model parameters obtained after processing by the personalized model parameter aggregation module, and send update reminders to the client.

[0114] On the client side, the personalized data storage module is used to store the user's personalized data, which includes at least the search behavior information of the target object and the identity information used to uniquely identify the target object. Specifically, the identity information used to uniquely identify the target object in the personalized data may include user ID, device ID, IP, user's word embedding vector, user's current location, user's age, user's gender and other feature information. The search behavior information in the personalized data may include currently input keywords and historical search behavior log information, specifically, it may include purchasing power obtained based on purchase data, user search word records, search word prefixes entered by the user, a list of search words exposed by the system, search words clicked by the user and other feature information.

[0115] The personalized model training module is used to train a personalized sorting model based on personalized data. When the model update times reach a preset number, the obtained model is sent to the data anonymity module to report the model parameters.

[0116] The data anonymization module is used to process personalized data using a privacy protection algorithm. Specifically, a differential privacy algorithm or other privacy protection noise algorithms can be used to remove data that can uniquely identify the user after being reported to the server from the personalized data to obtain non-personalized data.

[0117] The personalized sorting module is used to use the scoring index to sort the candidate words returned by the server to obtain the sorting results, and use the first N words of the sorting results as the target prompt words, that is, the search prompt words recommended to the user. The target prompt words are displayed in the display interface of the client device, for example, the target prompt words are displayed in the drop-down list of the search box, so that the user can select the target prompt word and search for information.

[0118] Through the embodiments of the present application, privacy protection algorithms such as data de-ID and differential privacy technology are used to ensure data anonymity, and the sample data noise caused by differential privacy technology is eliminated according to the data aggregation algorithm to ensure data availability. In addition, the federated learning algorithm is used to ensure the effectiveness of the sorting model trained by the client and server, and the double-layer sorting strategy is used to ensure the effectiveness of the recommended target prompt words.

[0119] Example 5

[0120] The present application also provides a method for determining a search prompt word, such as Figure 6 As shown, the method includes:

[0121] Step S61: The client device receives a first candidate prompt word set from a server, wherein the first candidate prompt word set is determined by candidate prompt words selected by the server from a preset candidate prompt word set based on a first learning model;

[0122] The first learning model may be a learning model based on a logistic regression algorithm, etc. The first learning model may be used to classify the data uploaded by the client, and then sort the preset candidate prompt words based on the classified data. For example, the server classifies the received non-personalized data (i.e., anonymized data, such as the click rate and popularity of the candidate words), and then sorts the preset candidate prompt words based on different types of data, and takes the top N candidate prompt words as the selected candidate prompt words to form the first candidate prompt word set.

[0123] Step S62, the client device determines a target prompt word from the first candidate prompt word set based on a local second learning model;

[0124] Step S63, displaying the target prompt word.

[0125] The preferred implementation of the embodiments of the present application can refer to the relevant descriptions in embodiments 1-4. For example, the second learning model in the embodiments of the present application can be a local learning model of the client device in embodiment 1. Therefore, the specific functions or execution steps of the client device can be the same as those in embodiment 1. For another example, before the client device in the embodiments of the present application uses the second learning model to determine the target prompt word, it can also receive aggregated model parameters from the server side. The aggregated model parameters can be the model parameters uploaded by the above-mentioned client device and other client devices. Other related implementation processes will not be repeated here.

[0126] Example 6

[0127] The present application also provides a method for determining a search prompt word, such as Figure 7 As shown, the method includes:

[0128] Step S71, the client device receives a first candidate prompt word set from the server, wherein the candidate prompt words in the first candidate prompt word set are arranged in a predetermined order;

[0129] Optionally, the first candidate prompt word set is determined by the candidate prompt words selected by the server from the preset candidate prompt word set based on the first learning model; the first learning model may be a learning model based on a logistic regression algorithm, etc., and the first learning model may be used to classify the data uploaded by the client, and then sort the preset candidate prompt words based on the classified data. For example, the server classifies the received non-personalized data (i.e., anonymized data, such as the click rate and popularity of the candidate words), and then sorts the preset candidate prompt words based on different types of data, and takes the top N candidate prompt words as the selected candidate prompt words to form the first candidate prompt word set.

[0130] Step S72, the client device determines a target prompt word based on the ranking result of the candidate prompt words in the first candidate prompt word set;

[0131] Optionally, when determining the target prompt word, the client device may directly use the sorting result in the first candidate prompt word set, for example, taking the first N candidate prompt words as the target prompt word.

[0132] In other embodiments of the present application, a candidate target prompt word can be determined based on the sorting result, and then the final target prompt word can be determined from the candidate target prompt words based on preset rules, wherein the preset rules can refer to the relevant description in Example 1, for example: the client device updates the local learning model according to the model parameters received from the server, and re-sorts the candidate target prompt words based on the updated local learning model to obtain a sorting result; the client device determines the final target prompt word from the first candidate prompt word set based on the sorting result obtained after the secondary sorting.

[0133] Step S73, display the target prompt word.

[0134] The preferred implementation methods of the embodiments of the present application can be found in the relevant descriptions in Examples 1-5, which will not be repeated here.

[0135] Example 7

[0136] The present application also provides a client device, such as Figure 8 As shown, the device includes:

[0137] The communication module is used to receive first data from a server, wherein the first data is determined by the server after analyzing data of a specified type based on a first learning model.

[0138] Specifically, the communication module is a module that supports data communication between a server and a client device, and the specified type of data may be data obtained by collecting personalized data of the user and performing privacy protection processing on the personalized data of the user.

[0139] The first learning model may be a machine learning model trained using user feature data, display data, and user feedback data. For example, in a search scenario, the first learning model may be data trained using user feature data, related search term exposure, and click logs. In a web page reordering scenario, the first learning model may be data trained using user feature data, related keywords, and web page reordering results.

[0140] The processor is used to process the first data based on a local second learning model to obtain second data.

[0141] Specifically, the second learning model may be a sorting model, and the client device selects the first data based on the sorting model, sorts the selected results according to a preset rule, and obtains the second data from the selected sorted results. For example, the preset rule may be a usage frequency, and the selected results are sorted according to the usage frequency, and the top N candidate prompt words are selected as the second data.

[0142] The data output module is used to output the second data.

[0143] Specifically, the data output module can be a device such as a printer and a display. When the data output module is a display, it can be a display of a client device. Specifically, the second data can be displayed in a display interface of the client device, and the second data can be displayed in a drop-down list for user selection, or the second data can be displayed in a scrolling display for user selection.

[0144] Example 8

[0145] The present application also provides a client device, such as Fig. 9 As shown, the device includes:

[0146] The communication module is used to receive the model parameters and the first candidate prompt word set sent by the server, wherein the first candidate prompt word set includes candidate prompt words selected by the server from the candidate prompt word set based on the designated type data uploaded by the client.

[0147] It should be noted that the server is used to provide a search engine to the client device, and the user searches for information based on the search engine on the client, and the communication module supports data transmission between the two.

[0148] The model parameters sent by the server are obtained by aggregating the model parameters reported by each client under the same search engine.

[0149] The candidate prompt words in the first candidate prompt word set are selected from the candidate prompt word set. Specifically, a selection condition is obtained according to the specified type data uploaded by the client, and the candidate prompt words are selected from the candidate prompt word set according to the selection condition, wherein the candidate words in the candidate prompt word set can be related recommended candidate words stored locally on the client after the user inputs the search word, and the specified type data is data obtained after the collected personalized data of the user is processed by using a privacy protection algorithm.

[0150] The processor is used to update the local learning model according to the model parameters, and determine the ranking result of the candidate prompt words in the first candidate prompt word set based on the updated local learning model; and determine the target prompt word from the first candidate prompt word set based on the ranking result.

[0151] It should be noted that the local learning model is used to obtain the scoring index of the candidate prompt word, and the scoring index is used to score the candidate word and recommend the candidate word to the user based on the scoring result. The local learning model needs to be continuously updated to accurately train the scoring index, and the local learning model can be updated according to the model parameters.

[0152] A display, used for displaying target prompt words.

[0153] It should be noted that the target prompt word is the search prompt word recommended to the user, and the user can search for information based on the target prompt word.

[0154] Specifically, the display may be a display of a client device, and the target prompt word may be displayed in a display interface of the client device, for example, the target prompt word may be displayed in a drop-down list of a search box for selection by the user.

[0155] Example 9

[0156] This application provides Fig.10 The system for determining the search prompt words shown. Fig.10 It is a schematic diagram of a system for determining search prompt words according to Embodiment 5 of the present application.

[0157] The client device and the server, wherein the server is used to obtain a first candidate prompt word set based on the prompt word selected from the candidate prompt word set based on the specified type of data uploaded by the client device, and send the model parameters to the client device.

[0158] It should be noted that the server is used to provide a search engine to the client, and the user searches for information based on the search engine on the client.

[0159] The candidate words in the candidate prompt word set may be related recommended candidate words stored locally by the client after the user inputs the search word. A selection condition is obtained according to the specified type data uploaded by the client, and a candidate prompt word is selected from the candidate prompt word set according to the selection condition to obtain a first candidate prompt word set.

[0160] The model parameters sent to the client device are model parameters obtained by aggregating model parameters of learning models uploaded by multiple clients of a type of search engine by the server.

[0161] The client device is used to receive model parameters and a first candidate prompt word set; update the local learning model according to the model parameters, and input the candidate prompt words in the first candidate prompt word set into the local learning model for analysis to obtain a scoring index of the candidate prompt words; determine a target prompt word from the first candidate prompt word set based on the scoring index; and display the target prompt word.

[0162] It should be noted that the local learning model of the client device is used to obtain the scoring index of the candidate prompt word, and the scoring index is used to score the candidate word and recommend the candidate word to the user based on the scoring result. The local learning model needs to be continuously updated to accurately train the scoring index, and the local learning model can be updated according to the model parameters.

[0163] There may be multiple scoring indicators, and different target prompt words may be determined from the first candidate prompt words according to different scoring indicators, or the scores of various scoring indicators may be weighted averaged to determine the target prompt word.

[0164] Specifically, the target prompt word may be displayed in a display interface of the client device, for example, the target prompt word may be displayed in a drop-down list of a search box for selection by the user.

[0165] Example 10

[0166] According to an embodiment of the present application, a device for implementing the above-mentioned method for determining a search prompt word is also provided. Fig.11 As shown, the device comprises:

[0167] The first receiving unit 111 is configured to use a client device to receive model parameters and a first candidate prompt word set sent by a server, wherein the first candidate prompt word set includes candidate prompt words selected by the server from the candidate prompt word set based on designated type data uploaded by the client.

[0168] It should be noted that the server is used to provide a search engine to the client, and the user searches for information based on the search engine on the client.

[0169] The model parameters sent by the server are obtained by aggregating the model parameters reported by each client under the same search engine.

[0170] The candidate prompt words in the first candidate prompt word set are selected from the candidate prompt word set. Specifically, a selection condition is obtained according to the specified type data uploaded by the client, and the candidate prompt words are selected from the candidate prompt word set according to the selection condition, wherein the candidate words in the candidate prompt word set can be related recommended candidate words stored locally on the client after the user inputs the search word, and the specified type data is data obtained after the collected personalized data of the user is processed by using a privacy protection algorithm.

[0171] The updating unit 112 is used to update the local learning model according to the model parameters using the client device, and input the candidate prompt words in the first candidate prompt word set into the local learning model for analysis to obtain the scoring index of the candidate prompt words.

[0172] It should be noted that the local learning model of the client device is used to obtain the scoring index of the candidate prompt word, and the scoring index is used to score the candidate word and recommend the candidate word to the user based on the scoring result. The local learning model needs to be continuously updated to accurately train the scoring index, and the local learning model can be updated according to the model parameters.

[0173] The first determining unit 113 is configured to determine a target prompt word from the first candidate prompt word set based on a scoring index through a client device.

[0174] It should be noted that there may be multiple scoring indicators, and different target prompt words may be determined from the first candidate prompt words according to different scoring indicators, or the scores of various scoring indicators may be weighted averaged to determine the target prompt word.

[0175] The first display unit 114 is used to display the target prompt word.

[0176] It should be noted that the target prompt word is the search prompt word recommended to the user, and the user can search for information based on the target prompt word.

[0177] Specifically, the target prompt word may be displayed in a display interface of the client device, for example, the target prompt word may be displayed in a drop-down list of a search box for selection by the user.

[0178] According to the above-mentioned embodiment of the present application, the device also includes: a first acquisition unit, used to obtain personalized data of the target object through a client device, wherein the personalized data includes at least search behavior information of the target object and identity information used to uniquely identify the target object; a processing unit, used to process the personalized data through the client device using a privacy protection algorithm to obtain non-personalized data, and use the non-personalized data as designated type data.

[0179] In one embodiment, the target object is a user, and the personalized data of the target object is the fine-grained feature data of the user. The identity information used to uniquely identify the target object in the personalized data may include user ID, device ID, IP, user's word embedding vector, user's current location, user's age, user's gender and other feature information. The search behavior information in the personalized data may include currently input keywords and historical search behavior log information, specifically, it may include purchasing power obtained based on purchase data, user search term records, search term prefixes entered by the user, a list of search terms exposed by the system, search terms clicked by the user and other feature information.

[0180] It should be noted that in order to prevent excessive memory usage, the client device only stores part of the data within a preset period, for example, data within the past three months.

[0181] In addition, in order to protect the privacy of the user, after the client device obtains the personalized data of the target object, it uses a privacy protection algorithm to process the personalized data. Specifically, a differential privacy algorithm can be used, or other privacy protection noise algorithms can be used to remove the data that can uniquely identify the user after being reported to the server in the personalized data, and obtain non-personalized data. The obtained non-personalized data is reported to the server as a specified type of data. Specifically, the data that can uniquely identify the user after being reported to the server includes but is not limited to user ID, device ID, IP, user's word embedding vector and other data.

[0182] According to the above-mentioned embodiment of the present application, the device also includes: a training unit, which is used to train the local learning model based on personalized data through the client device before the client device sends the specified type of data to the server, and count the number of training times of the local learning model in the current statistical period; a comparison unit, which is used to use the client device to compare the number of training times and a preset threshold; a second determination unit, which is used to determine through the client device that it is allowed to send the specified type of data to the server when the number of training times is greater than the preset threshold.

[0183] It should be noted that the local learning model is used to obtain the scoring indicators of the candidate prompt words. The basic data for training the local learning model is non-personalized data. The model is continuously trained and updated with the obtained non-personalized data to obtain a model that can accurately obtain the scoring indicators. When the number of training times reaches the preset threshold, it means that the local model meets the requirements, the model parameters are uploaded to the server, and the specified type of data is allowed to be sent to the server.

[0184] According to the above-mentioned embodiment of the present application, the search behavior information includes: the historical selection record of the target object for the prompt word, the search word prefix input by the target object, and the prompt word list exposed by the server.

[0185] It should be noted that when the target object selects a prompt word, a selection record will be generated, and the search behavior information includes the user's historical selection data; when the user searches, he will first enter the prefix of the search word, and the search behavior information includes the search word prefix entered by the user; before the user enters the search word, the server can provide a prompt word list for the user to choose, and the search behavior information includes the prompt word list exposed by the server. In addition, the search behavior information also includes other data related to the user's search behavior.

[0186] According to the above embodiment of the present application, the first determining unit 113 includes: a first sorting module, configured to sort the candidate prompt words in the first candidate prompt word set based on the scoring index through the client device, and use the top N candidate prompt words as target prompt words.

[0187] It should be noted that the candidate prompt words in the first candidate prompt word set are sorted according to the determined evaluation index and the selected sorting algorithm. Specifically, the sorting method can be a point-to-point method, a pair method and a list method. After sorting, the top candidate prompt words are used as target prompt words and displayed for user selection.

[0188] In addition, after displaying the target keyword, the displayed target prompt word is saved, and information such as which target prompt words the user clicks is stored, providing a data basis for subsequent collection of personalized data.

[0189] According to the above-mentioned embodiment of the present application, the model parameters received by the client device are obtained by aggregating the model parameters of the local learning model of the client device and the model parameters of the learning models of other client devices. The other client devices share the same type of search engine provided by the server with the client device.

[0190] It should be noted that the server provides a search engine to multiple clients. The server receives model parameters of the learning model uploaded by multiple clients of the same type of search engine, and aggregates the model parameters. Specifically, the federated average update algorithm can be used for aggregation, or other distributed algorithms can be used for aggregation. The aggregated model parameters are used as the model parameters sent to the client device to update the local learning model of the client device.

[0191] It should be noted that the above units and modules correspond to the steps in Example 1, and the examples and application scenarios implemented by the above units and modules are the same as those of the corresponding steps, but are not limited to the contents disclosed in the above Example 1. It should be noted that the above units and modules, as part of the device, can be run in the computer terminal 10 provided in Example 1.

[0192] Embodiment 11

[0193] According to an embodiment of the present application, a device for implementing the above-mentioned method for determining a search prompt word is also provided. Fig.12 As shown, the device comprises:

[0194] The second acquisition unit 121 is used to acquire model parameters of learning models in different client devices through the server, wherein the different client devices run the same type of search engine provided by the server.

[0195] It should be noted that the server is used to provide a search engine to the client, and the user searches for information based on the search engine on the client. The server receives model parameters of the learning model uploaded by multiple clients of the same type of search engine.

[0196] The aggregation unit 122 is used to aggregate the model parameters corresponding to different client devices through the server to obtain target model parameters.

[0197] Specifically, a federated average update algorithm may be used for aggregation, or other distributed algorithms may be used for aggregation, and the model parameters obtained by aggregation are used as model parameters sent to the client device to update the local learning model of the client device.

[0198] The selection unit 123 is configured to receive the designated type data reported by the client device through the server, and select a candidate prompt word from the candidate prompt word set based on the designated type data to obtain a first candidate prompt word set.

[0199] It should be noted that the designated type of data is data obtained by processing the collected personalized data of the user using a privacy protection algorithm, wherein the personalized data at least includes the search behavior information of the target object and the identity information used to uniquely identify the target object.

[0200] The candidate words in the candidate prompt word set may be related recommended candidate words stored locally on the client after the user inputs a search word. The candidate prompt words are selected from the candidate prompt word set according to the designated type data uploaded by the client to obtain the first candidate prompt word set.

[0201] The sending unit 124 is configured to send the first candidate prompt word set and the target model parameters to the client device through the server.

[0202] Specifically, the first candidate prompt word set and the target model parameters are sent to the client device to update the local learning model according to the model parameters, and the candidate prompt words in the first candidate prompt word set are input into the local learning model for analysis to obtain scoring indicators of the candidate prompt words, so as to score the candidate words and recommend the candidate words to the user according to the scoring results.

[0203] According to the above embodiment of the present application, the selection unit 123 includes: a second sorting module, used to sort the candidate prompt words in the candidate prompt word set based on the specified type of data; and a determination module, used to select the top N candidate prompt words as the selected candidate prompt words.

[0204] It should be noted that the selection conditions are obtained according to the specified type of data uploaded by the client, and candidate prompt words are selected from the candidate prompt word set according to the selection conditions to obtain the first candidate prompt word set. For example, the click rate, popularity and other selection conditions of the candidate words can be obtained according to the specified type of data, the candidate words in the candidate prompt word set are sorted, and the TopN candidate prompt words are taken as the selected candidate prompt words to form the first candidate prompt word set.

[0205] It should be noted that the above units and modules correspond to the steps in Example 2, and the examples and application scenarios implemented by the above units and modules are the same as those of the corresponding steps, but are not limited to the contents disclosed in the above Example 1. It should be noted that the above units and modules, as part of the device, can be run in the computer terminal 10 provided in Example 1.

[0206] Example 12

[0207] According to an embodiment of the present application, a device for implementing the above-mentioned method for determining a search prompt word is also provided. Fig.13 As shown, the device comprises:

[0208] The second receiving unit 131 is configured to use a client device to receive a candidate prompt word set sent by a server.

[0209] It should be noted that the server is used to provide a search engine to the client, and the user searches for information on the client based on the search engine. The candidate words in the candidate prompt word set can be related recommended candidate words stored locally on the client after the user enters the search word.

[0210] The first analysis unit 132 is used to input the candidate prompt word set into the learning model for analysis through the client device to obtain the scoring index of the candidate prompt word.

[0211] It should be noted that the learning model of the client device is used to obtain the scoring index of the candidate prompt word, and the scoring index of the candidate prompt word is used to score the candidate word and recommend the candidate word to the user according to the scoring result.

[0212] The third determining unit 133 is configured to determine a target prompt word from the candidate prompt word set according to the scoring index through the client device.

[0213] It should be noted that there may be multiple scoring indicators, and different target prompt words may be determined from the first candidate prompt words according to different scoring indicators, or the scores of various scoring indicators may be weighted averaged to determine the target prompt word.

[0214] The second display unit 134 is used to display the target prompt word.

[0215] It should be noted that the target prompt word is the search prompt word recommended to the user, and the user can search for information based on the target prompt word.

[0216] Specifically, the target prompt word may be displayed in a display interface of the client device, for example, the target prompt word may be displayed in a drop-down list of a search box for selection by the user.

[0217] It should be noted that the above-mentioned determination module and sending module correspond to the steps in Example 3, and the examples and application scenarios implemented by the above-mentioned modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned Example 1. It should be noted that the above-mentioned modules, as part of the device, can be run in the computer terminal 10 provided in Example 1.

[0218] Example 13

[0219] According to an embodiment of the present application, a device for implementing the above-mentioned method for determining a search prompt word is also provided. Fig.14 As shown, the device comprises:

[0220] The third receiving unit 141 is configured to receive a first candidate prompt word set from a server through a client device, wherein the first candidate prompt word set is determined by candidate prompt words selected by the server from a preset candidate prompt word set based on a first learning model.

[0221] The first learning model may be a learning model based on a logistic regression algorithm, etc. The first learning model may be used to classify the data uploaded by the client, and then sort the preset candidate prompt words based on the classified data. For example, the server classifies the received non-personalized data (i.e., anonymized data, such as the click rate and popularity of the candidate words), and then sorts the preset candidate prompt words based on different types of data, and takes the top N candidate prompt words as the selected candidate prompt words to form the first candidate prompt word set.

[0222] The fourth determining unit 142 is configured to determine a target prompt word from the first candidate prompt word set based on a local second learning model through the client device.

[0223] The third display unit 143 is used to display the target prompt word.

[0224] Specifically, the target prompt word may be displayed in a display interface of the client device, for example, the target prompt word may be displayed in a drop-down list of a search box for selection by the user.

[0225] It should be noted that the above-mentioned determination module and sending module correspond to the steps in Example 5, and the examples and application scenarios implemented by the above-mentioned modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned Example 1. It should be noted that the above-mentioned modules, as part of the device, can be run in the computer terminal 10 provided in Example 1.

[0226] Embodiment 14

[0227] According to an embodiment of the present application, a device for implementing the above-mentioned method for determining a search prompt word is also provided. Fig.15 As shown, the device comprises:

[0228] The fourth receiving unit 151 is configured to use a client device to receive a first candidate prompt word set from a server, wherein the candidate prompt words in the first candidate prompt word set are arranged in a predetermined order.

[0229] Optionally, the first candidate prompt word set is determined by the candidate prompt words selected by the server from the preset candidate prompt word set based on the first learning model; the first learning model may be a learning model based on a logistic regression algorithm, etc., and the first learning model may be used to classify the data uploaded by the client, and then sort the preset candidate prompt words based on the classified data. For example, the server classifies the received non-personalized data (i.e., anonymized data, such as the click rate and popularity of the candidate words), and then sorts the preset candidate prompt words based on different types of data, and takes the top N candidate prompt words as the selected candidate prompt words to form the first candidate prompt word set.

[0230] The fifth determining unit 152 is configured to determine a target prompt word based on the ranking result of the candidate prompt words in the first candidate prompt word set through the client device.

[0231] Optionally, when determining the target prompt word, the client device may directly use the sorting result in the first candidate prompt word set, for example, taking the first N candidate prompt words as the target prompt word.

[0232] In other embodiments of the present application, a candidate target prompt word can be determined based on the sorting result, and then the final target prompt word can be determined from the candidate target prompt words based on preset rules, wherein the preset rules can refer to the relevant description in Example 1, for example: the client device updates the local learning model according to the model parameters received from the server, and re-sorts the candidate target prompt words based on the updated local learning model to obtain a sorting result; the client device determines the final target prompt word from the first candidate prompt word set based on the sorting result obtained after the secondary sorting.

[0233] The third display unit 153 is used to display the target prompt word.

[0234] Specifically, the target prompt word may be displayed in a display interface of the client device, for example, the target prompt word may be displayed in a drop-down list of a search box for selection by the user.

[0235] It should be noted that the above-mentioned determination module and sending module correspond to the steps in Example 6, and the examples and application scenarios implemented by the above-mentioned modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned Example 1. It should be noted that the above-mentioned modules, as part of the device, can be run in the computer terminal 10 provided in Example 1.

[0236] Embodiment 15

[0237] According to an embodiment of the present application, a device for implementing the above information processing method is also provided, such as Fig.16 As shown, the device comprises:

[0238] The fifth receiving unit 161 is used for the client device to receive first data from the server, wherein the first data is determined by the server after analyzing the specified type of data based on the first learning model.

[0239] Specifically, the designated type of data may be data obtained by collecting the personalized data of the user and performing privacy protection processing on the personalized data of the user.

[0240] Among them, the user's personalized data includes characteristics such as user ID, city, current location, age, gender, and information entered by the user. Specifically, in the search scenario, the information entered by the user may include data such as the search terms entered by the user and the search terms clicked by the user. In the web page re-arrangement scenario, the information entered by the user may also include keywords entered by the user for web page re-arrangement.

[0241] It should be noted that after the user inputs information, the system will display the corresponding data based on the input information, and the user will also provide further feedback on the displayed data. The first learning model can be a machine learning model trained using user feature data, display data, and user feedback data. For example, in a search scenario, the first learning model can be data trained using user feature data, related search term exposure, and click logs. In a web page re-arrangement scenario, the first learning model can be data trained using user feature data, related keywords, and web page re-arrangement results.

[0242] The specified type of data is input into the first learning model in the server, and the first learning model performs data analysis based on the specified type of data to obtain an analysis result. For example, in a search scenario, after the user's personalized data is input, the first learning model can output the result of whether the prompt word corresponding to the search word is clicked, that is, the result of whether the prompt word is selected. After obtaining the first data, the server sends the first data to the client device, and the client device receives the first data to further process the first data.

[0243] The processing unit 162 is used for the client device to process the first data based on the local second learning model to obtain the second data.

[0244] Specifically, the second learning model may be a sorting model, and the client device selects the first data based on the sorting model, sorts the selected results according to a preset rule, and obtains the second data from the selected sorted results. For example, the preset rule may be a usage frequency, and the selected results are sorted according to the usage frequency, and the top N candidate prompt words are selected as the second data.

[0245] The output unit 163 is used to output the second data.

[0246] Optionally, the second data may be output in a display manner. Specifically, the second data may be displayed in a display interface of the client device, the second data may be displayed in a drop-down list for user selection, or the second data may be displayed in a scrolling manner for user selection.

[0247] Example 16

[0248] According to an embodiment of the present application, a system is further provided, the system comprising:

[0249] Processor; and

[0250] The memory is connected to the processor and is used to provide the processor with instructions for performing the following processing steps:

[0251] The model parameters and a first candidate prompt word set sent by the server are received, where the first candidate prompt word set includes candidate prompt words selected by the server from the candidate prompt word set based on the designated type of data uploaded by the client.

[0252] It should be noted that the model parameters sent by the server are model parameters obtained by aggregating the model parameters reported by various clients under the same search engine.

[0253] The candidate prompt words in the first candidate prompt word set are selected from the candidate prompt word set. Specifically, a selection condition is obtained according to the specified type data uploaded by the client, and the candidate prompt words are selected from the candidate prompt word set according to the selection condition, wherein the candidate words in the candidate prompt word set can be related recommended candidate words stored locally on the client after the user inputs the search word, and the specified type data is data obtained after the collected personalized data of the user is processed by using a privacy protection algorithm.

[0254] The local learning model is updated according to the model parameters, and the candidate prompt words in the first candidate prompt word set are input into the local learning model for analysis to obtain a scoring index for the candidate prompt words.

[0255] It should be noted that the local learning model of the client device is used to obtain the scoring index of the candidate prompt word, and the scoring index is used to score the candidate word and recommend the candidate word to the user based on the scoring result. The local learning model needs to be continuously updated to accurately train the scoring index, and the local learning model can be updated according to the model parameters.

[0256] A target prompt word is determined from the first candidate prompt word set based on the scoring index.

[0257] It should be noted that there may be multiple scoring indicators, and different target prompt words may be determined from the first candidate prompt words according to different scoring indicators, or the scores of various scoring indicators may be weighted averaged to determine the target prompt word.

[0258] Embodiment 17

[0259] The embodiment of the present application may provide a computer terminal, which may be any computer terminal device in a computer terminal group. Optionally, in this embodiment, the computer terminal may also be replaced by a terminal device such as a mobile terminal.

[0260] Optionally, in this embodiment, the computer terminal may be located in at least one network device among a plurality of network devices of the computer network.

[0261] In this embodiment, the above-mentioned computer terminal can execute the program code of the following steps in the method for determining the search prompt word of the application: the client device receives the model parameters and the first candidate prompt word set sent by the server, and the first candidate prompt word set includes the candidate prompt words selected by the server from the candidate prompt word set based on the specified type of data uploaded by the client; the client device updates the local learning model according to the model parameters, and inputs the candidate prompt words in the first candidate prompt word set into the local learning model for analysis to obtain the scoring index of the candidate prompt words; the client device determines the target prompt word from the first candidate prompt word set based on the scoring index; and displays the target prompt word.

[0262] Optionally, Fig.17 is a structural block diagram of a computer terminal according to an embodiment of the present application. Fig.17 As shown, the computer terminal A may include: one or more (only one is shown in the figure) processors and a memory.

[0263] Among them, the memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the method and device for determining the search prompt word in the embodiment of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, the above-mentioned method for determining the search prompt word is realized. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories can be connected to the terminal A via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0264] The processor can call the information and application stored in the memory through the transmission device to execute the following steps: the client device receives the model parameters and the first candidate prompt word set sent by the server, and the first candidate prompt word set includes candidate prompt words selected by the server from the candidate prompt word set based on the specified type of data uploaded by the client; the client device updates the local learning model according to the model parameters, and inputs the candidate prompt words in the first candidate prompt word set into the local learning model for analysis to obtain the scoring index of the candidate prompt words; the client device determines the target prompt word from the first candidate prompt word set based on the scoring index; and displays the target prompt word.

[0265] Optionally, the processor may also execute program code of the following steps: the client device obtains personalized data of the target object, wherein the personalized data includes at least search behavior information of the target object and identity information for uniquely identifying the target object; the client device processes the personalized data using a privacy protection algorithm to obtain non-personalized data, and uses the non-personalized data as data of a specified type.

[0266] Optionally, the processor may also execute program code of the following steps: before the client device sends data of a specified type to the server, the client device trains a local learning model based on personalized data, and counts the number of trainings of the local learning model within the current statistical period; the client device compares the number of trainings with a preset threshold; when the number of trainings is greater than the preset threshold, the client device determines that sending data of a specified type to the server is allowed.

[0267] Optionally, the processor may further execute program code of the following steps: the client device sorts the candidate prompt words in the first candidate prompt word set based on the scoring index, and uses the top N candidate prompt words as target prompt words.

[0268] Optionally, the processor may also execute the program code of the following steps: the model parameters received by the client device are obtained by aggregating the model parameters of the local learning model of the client device and the model parameters of the learning models of other client devices, and the other client devices share the same type of search engine provided by the server with the client device.

[0269] In an embodiment of the present application, a client device is used to receive a model parameter and a first candidate prompt word set sent by a server, wherein the first candidate prompt word set includes candidate prompt words selected by the server from the candidate prompt word set based on data of a specified type uploaded by the client; the client device updates a local learning model according to the model parameters, and inputs the candidate prompt words in the first candidate prompt word set into the local learning model for analysis to obtain a scoring index for the candidate prompt words; the client device determines a target prompt word from the first candidate prompt word set based on the scoring index; and the target prompt word is displayed in a manner that selects a candidate prompt word from the candidate prompt word set based on data of a specified type uploaded by the client, and determines the target prompt word from the first candidate prompt word set based on the scoring index, thereby achieving the purpose of providing search suggestions to users without directly using user personalized data, thereby solving the technical problem of not being able to simultaneously take into account the search effect of personalized search words and the security of non-personalized search word suggestion schemes.

[0270] It can be understood by those skilled in the art that Fig.17 The structure shown is for illustration only, and the computer terminal may also be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a handheld computer, a mobile Internet device (Mobile Internet Devices, MID), a PAD, or other terminal devices. Fig.17 The structure of the electronic device is not limited. For example, the computer terminal 10 may also include Fig.17 More or fewer components (such as network interfaces, display devices, etc.) shown in, or having Fig.17 Different configurations shown.

[0271] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, and the storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0272] Embodiment 18

[0273] The embodiment of the present application further provides a storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the method for determining the search prompt word provided in the first embodiment.

[0274] Optionally, in this embodiment, the above storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group.

[0275] Optionally, in this embodiment, the storage medium is configured to store program codes for executing the following steps: the client device receives the model parameters and the first candidate prompt word set sent by the server, the first candidate prompt word set including candidate prompt words selected by the server from the candidate prompt word set based on the specified type of data uploaded by the client; the client device updates the local learning model according to the model parameters, and inputs the candidate prompt words in the first candidate prompt word set into the local learning model for analysis to obtain scoring indicators for the candidate prompt words; the client device determines the target prompt word from the first candidate prompt word set based on the scoring indicator; and displays the target prompt word.

[0276] Optionally, the processor may also execute program code of the following steps: the client device obtains personalized data of the target object, wherein the personalized data includes at least search behavior information of the target object and identity information for uniquely identifying the target object; the client device processes the personalized data using a privacy protection algorithm to obtain non-personalized data, and uses the non-personalized data as data of a specified type.

[0277] Optionally, the processor may also execute program code of the following steps: before the client device sends data of a specified type to the server, the client device trains a local learning model based on personalized data, and counts the number of trainings of the local learning model within the current statistical period; the client device compares the number of trainings with a preset threshold; when the number of trainings is greater than the preset threshold, the client device determines that sending data of a specified type to the server is allowed.

[0278] Optionally, the processor may further execute program code of the following steps: the client device sorts the candidate prompt words in the first candidate prompt word set based on the scoring index, and uses the top N candidate prompt words as target prompt words.

[0279] Optionally, the processor may also execute the program code of the following steps: the model parameters received by the client device are obtained by aggregating the model parameters of the local learning model of the client device and the model parameters of the learning models of other client devices, and the other client devices share the same type of search engine provided by the server with the client device.

[0280] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0281] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0282] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0283] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0284] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0285] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk and other media that can store program codes.

[0286] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for determining a search prompt word, comprising: The client device receives the model parameters and the first candidate prompt word set sent by the server, wherein the first candidate prompt word set includes candidate prompt words selected by the server from the candidate prompt word set based on the designated type of data uploaded by the client; The client device updates the local learning model according to the model parameters, and determines a ranking result of candidate prompt words in the first candidate prompt word set based on the updated local learning model, wherein the model parameters are obtained by aggregating model parameters of the local learning model of the client device and model parameters of learning models of other client devices, and the other client devices share the same type of search engine provided by the server with the client device; the client device determines a target prompt word from the first candidate prompt word set based on the ranking result; displaying the target cue word; The client device acquires personalized data of the target object, wherein the personalized data at least includes search behavior information of the target object and identity information for uniquely identifying the target object; The client device processes the personalized data using a privacy protection algorithm to obtain non-personalized data, and uses the non-personalized data as the designated type of data.

2. The method according to claim 1, wherein: Before the client device sends the specified type of data to the server, the method further includes: The client device trains the local learning model based on the personalized data, and counts the number of training times of the local learning model in a current statistical period; The client device compares the number of training times with a preset threshold; When the number of training times is greater than the preset threshold, the client device determines to allow sending the designated type of data to the server.

3. The method according to claim 1, wherein: The search behavior information includes: the historical selection record of the target object for the prompt word, the search word prefix input by the target object, and the prompt word list exposed by the server.

4. The method according to claim 1, wherein: Determining a ranking result of candidate prompt words in the first candidate prompt word set based on the updated local learning model includes: The client device inputs the candidate prompt words in the first candidate prompt word set into the local learning model for analysis to obtain a scoring index for the candidate prompt words; The client device sorts the candidate prompt words in the first candidate prompt word set based on the scoring index to obtain the sorting result; The client device determines a target prompt word from the first candidate prompt word set based on the sorting result, including: taking the top N candidate prompt words in the sorting result as the target prompt word.

5. A method for determining a search prompt word, comprising: The server obtains model parameters of the learning model in different client devices, wherein the different client devices run the same type of search engine provided by the server; The server aggregates the model parameters corresponding to different client devices to obtain target model parameters, wherein the model parameters are obtained by aggregating model parameters of a learning model of the client device and model parameters of learning models of other client devices, and the other client devices share the same type of search engine provided by the server with the client device; the server receives the designated type data reported by the client device, and selects a candidate prompt word from a candidate prompt word set based on the designated type data to obtain a first candidate prompt word set; The server sends the first candidate prompt word set and the target model parameters to the client device, and obtains personalized data of the target object based on the client device, wherein the personalized data at least includes search behavior information of the target object and identity information for uniquely identifying the target object; The client device processes the personalized data using a privacy protection algorithm to obtain non-personalized data, and uses the non-personalized data as the designated type of data.

6. The method according to claim 5, wherein: Selecting candidate prompt words from the candidate prompt words based on the specified type of data to obtain a first candidate prompt word set includes: sorting the candidate prompt words in the candidate prompt word set based on the specified type of data; The top N candidate prompt words are selected as the candidate prompt words.

7. A method for determining a search prompt word, comprising: The client device receives the candidate prompt word set sent by the server; The client device inputs the candidate prompt word set into the learning model for analysis to obtain a scoring index for the candidate prompt word, wherein the client device receives a model parameter sent by the server, the model parameter being obtained by aggregating a model parameter of the learning model of the client device and a model parameter of a learning model of another client device, the other client device and the client device sharing the same type of search engine provided by the server; The client device determines a target prompt word from the candidate prompt word set according to the scoring index; The target cue word is displayed.

8. A method for determining a search prompt word, comprising: The client device receives a first candidate prompt word set from the server, wherein the first candidate prompt word set is determined by candidate prompt words selected by the server from a preset candidate prompt word set based on a first learning model; The client device determines a target prompt word from the first candidate prompt word set based on a local second learning model, wherein the client device receives a model parameter sent by the server, the model parameter being obtained by aggregating a model parameter of the local second learning model of the client device and a model parameter of a learning model of another client device, the other client devices sharing the same type of search engine provided by the server with the client device; The target cue word is displayed.

9. A method for determining a search prompt word, comprising: The client device receives a first candidate prompt word set from the server, wherein the candidate prompt words in the first candidate prompt word set are arranged in a predetermined order; The client device determines a target prompt word based on a ranking result of candidate prompt words in the first candidate prompt word set, wherein the client device receives a model parameter sent by the server, the model parameter being obtained by aggregating a model parameter of a local learning model of the client device and a model parameter of a learning model of other client devices, the other client devices sharing the same type of search engine provided by the server with the client device; The target cue word is displayed.

10. An information processing method, comprising: The client device receives first data from the server, wherein the first data is determined by the server after analyzing data of a specified type based on a first learning model; The client device processes the first data based on a local second learning model to obtain second data, wherein the client device receives model parameters sent by the server, and the model parameters are obtained by aggregating model parameters of the local second learning model of the client device and model parameters of learning models of other client devices, and the other client devices share the same type of search engine provided by the server with the client device; outputting the second data; The client device obtains personalized data of the target object, which personalized data at least includes search behavior information of the target object and identity information used to uniquely identify the target object; the client device uses a privacy protection algorithm to process the personalized data to obtain non-personalized data, and uses the non-personalized data as the designated type of data.

11. A client device, comprising: A communication module, configured to receive first data from a server, wherein the first data is determined by the server after analyzing data of a specified type based on a first learning model; wherein personalized data of a target object is obtained based on a client device, and the personalized data at least includes search behavior information of the target object and identity information for uniquely identifying the target object; the client device processes the personalized data using a privacy protection algorithm to obtain non-personalized data, and uses the non-personalized data as the data of the specified type; A processor, configured to process the first data based on a local second learning model to obtain second data, wherein the client device receives model parameters sent by the server, the model parameters being obtained by aggregating model parameters of the local second learning model of the client device and model parameters of learning models of other client devices, the other client devices sharing the same type of search engine provided by the server with the client device; A data output module is used to output the second data.

12. A client device, comprising: a communication module, configured to receive a model parameter and a first candidate prompt word set sent by a server, wherein the first candidate prompt word set includes candidate prompt words selected by the server from the candidate prompt word set based on data of a specified type uploaded by a client, and to obtain personalized data of a target object based on the client device, wherein the personalized data includes at least search behavior information of the target object and identity information for uniquely identifying the target object; The client device processes the personalized data using a privacy protection algorithm to obtain non-personalized data, and uses the non-personalized data as the designated type of data; a processor, configured to update the local learning model according to the model parameters, and determine a ranking result of candidate prompt words in the first candidate prompt word set based on the updated local learning model; and determine a target prompt word from the first candidate prompt word set based on the ranking result, wherein the model parameters are obtained by aggregating model parameters of the local learning model of the client device and model parameters of learning models of other client devices, and the other client devices share the same type of search engine provided by the server with the client device; A display is used to display the target prompt word.

13. A system for determining a search prompt word, comprising: A client device and a server, wherein The server is used to select a prompt word from a candidate prompt word set based on the designated type data uploaded by the client device, obtain a first candidate prompt word set, send a model parameter to the client device, and obtain personalized data of a target object based on the client device, wherein the personalized data at least includes search behavior information of the target object and identity information for uniquely identifying the target object; the client device processes the personalized data using a privacy protection algorithm to obtain non-personalized data, and uses the non-personalized data as the designated type data; The client device is configured to receive a model parameter and a first candidate prompt word set sent by a server, wherein the first candidate prompt word set includes candidate prompt words selected by the server from a candidate prompt word set based on data of a specified type uploaded by the client; update a local learning model according to the model parameter, determine a ranking result of the candidate prompt words in the first candidate prompt word set based on the updated local learning model, determine a target prompt word from the first candidate prompt word set based on the ranking result; and display the target prompt word, wherein the model parameter is obtained by aggregating the model parameter of the local learning model of the client device and the model parameter of the learning model of other client devices, and the other client devices share the same type of search engine provided by the server with the client device.

14. A storage medium comprising a stored program, wherein: When the program is running, the device where the storage medium is located is controlled to execute the method for determining the search prompt word according to any one of claims 1 to 4.

15. A system comprising: processor; as well as A memory, connected to the processor, configured to provide the processor with instructions for processing the following processing steps: Receiving the model parameters and the first candidate prompt word set sent by the server, wherein the first candidate prompt word set includes candidate prompt words selected by the server from the candidate prompt word set based on the designated type of data uploaded by the client; updating the local learning model according to the model parameters, determining a ranking result of candidate prompt words in the first candidate prompt word set based on the updated local learning model, and determining a target prompt word from the first candidate prompt word set based on the ranking result; The target prompt word is displayed, wherein the model parameters are obtained by aggregating the model parameters of the local learning model of the client device and the model parameters of the learning models of other client devices, the other client devices share the same type of search engine provided by the server with the client device, and the personalized data of the target object is obtained based on the client device, and the personalized data at least includes the search behavior information of the target object and the identity information used to uniquely identify the target object; The client device processes the personalized data using a privacy protection algorithm to obtain non-personalized data, and uses the non-personalized data as the designated type of data.

Citation Information

Patent Citations

  • Intelligent prompt method, module and system for search

    CN103631929A

  • Personalized search suggestion method and device

    CN104424300A

  • Method and device for complementing search terms and establishing individual interest model

    CN105912669A

  • Method and device for recommending input association based on historical search of user

    CN108227955A

  • Search method, client, server, and storage medium

    CN109213920A