Question recommendation method, question answering method and related device

Abstract

The invention relates to a question recommendation method, a question answering method and a related device. The method comprises the steps of displaying a question-answering interface, obtaining current input content on the question-answering interface, and displaying a recommended question on the question-answering interface based on the input content; wherein the recommendation question is generated according to user data; the user data at least comprises local content of the user in the electronic equipment. According to the method, various local information of the user is understood, potential recommendation problems are prepared, and user behavior preferences are understood at the same time. When the user carries out search questions and answers, the recommendation questions conforming to user preferences can be given in real time, the user is assisted to carry out query better, and the end-to-end efficiency and experience are improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a problem recommendation method, a question-answering method, and related apparatus. Background Technology

[0002] With the development of information technology, people's lives are closely intertwined with electronic devices (such as mobile phones and tablets). These devices allow for communication and the recording of information in various forms, fields, and scenarios. However, over time, a massive amount of fragmented information accumulates, often causing users' memories to become blurred. In such cases, traditional solutions cannot provide effective search suggestions. Summary of the Invention

[0003] Therefore, it is necessary to provide a question recommendation method, question answering method, and related apparatus that can provide effective search suggestions to address the aforementioned technical problems.

[0004] In a first aspect, this application provides a question recommendation method applied to an electronic device. The method includes: displaying a question-and-answer interface and obtaining the current input content on the question-and-answer interface; displaying recommended questions on the question-and-answer interface based on the input content; wherein the recommended questions are generated based on user data; and the user data includes at least the user's local content on the electronic device.

[0005] In one embodiment, displaying a question-and-answer interface includes: displaying the question-and-answer interface in response to a triggering operation of the question-and-answer entry; the question-and-answer entry is used to provide question-and-answer functionality after being triggered.

[0006] In one embodiment, the question-and-answer entry point includes a button control; the triggering operation includes one or more of the following: a touch operation on the button control, a voice input operation, a touch operation on the fingerprint sensor, and a key operation.

[0007] In one embodiment, the user data also includes online favorites, and the method further includes: generating questions based on online favorites to obtain first question material; generating questions based on local content to obtain second question material; and generating recommended questions based on at least one of the first question material and the second question material.

[0008] In one embodiment, the first question material includes multiple first pairs, each pair including a first question statement and corresponding online collection content; generating the first question material based on the online collection content includes: analyzing and processing the online collection content through a large language model to obtain the first question statement.

[0009] In one embodiment, the first question statement includes a statement in sentence form.

[0010] In one embodiment, the second question material includes multiple second pairs, each pair including a second question statement and corresponding local content. Generating the second question material based on the local content includes: obtaining permission information corresponding to the local content; the permission information indicates the cloud's access permissions for the local content; if the permission information determines that the cloud has permission to access the local content, then the local content is processed through a large language model to obtain the second question statement; if the permission information determines that the cloud does not have permission to access the local content, then the local content is processed through an edge-side model to obtain the second question statement.

[0011] In one embodiment, the second question statement includes key information in the form of a phrase or a short sentence.

[0012] In one embodiment, displaying recommended questions on the question-and-answer interface based on the input content includes: if the input content does not meet the question-and-answer conditions, displaying randomly selected recommended questions; if the input content meets the question-and-answer conditions, determining the recommended questions to be displayed on the question-and-answer interface based on the input content.

[0013] In one embodiment, the randomly selected recommendation question is related to the time when the recommendation question was generated.

[0014] In one embodiment, the randomly selected recommended questions include those generated at times closest to the display time of the question-and-answer interface; and / or, the randomly selected recommended questions include those generated at times furthest from the display time of the question-and-answer interface.

[0015] In one embodiment, the question-answering conditions include determining that the input content is meaningful and determining that the input content has a question-answering intent.

[0016] In one embodiment, the user data also includes online favorites; determining recommended questions to be displayed on the Q&A interface based on the input content includes: performing phrase clustering on the input content to obtain clustered input content; using multiple retrieval methods to search the clustered input content based on a first question material and a second question material to obtain multiple retrieval results; the first question material is generated from questions based on online favorites; the second question material is generated from questions based on local content; the multiple retrieval results are merged and sorted to obtain a question material list; and recommended questions are generated based on the question materials in the question material list.

[0017] In one embodiment, the input content is clustered to obtain clustered input content, including: in response to detecting terms in the input content, clustering the terms to obtain clustered input content; wherein, the terms include one or more of the following: prepositional terms, time terms, name terms, alphabetic terms, and colloquial terms; the clustering process includes filtering prepositional terms, a question material matching strategy for time terms, and standardization processing for name terms, alphabetic terms, and colloquial terms.

[0018] In one embodiment, the method further includes: when any question material in the question material list belongs to the second question material, if the length of the question statement in any question material is less than the length threshold, then the question statements in any question material are assembled to obtain question statements that meet the semantic fluency requirements.

[0019] In one embodiment, assembling the question statements in any question material to obtain question statements that meet the semantic fluency requirements includes: obtaining the data source type name corresponding to any question material; comparing the last text of the question statement in any question material with the data source type name to obtain a comparison result; and selecting a target template to expand and rewrite the question statement in any question material based on the comparison result.

[0020] In one embodiment, the method further includes: determining the order of multiple recommended questions based on user personalization information; wherein the user personalization information includes the distance between the generation time of the recommended questions and the display time of the question-and-answer interface, the semantic similarity between the recommended questions and the input content, and one or more of the following: question type preference and full user data.

[0021] In one embodiment, the question-and-answer interface includes a search bar; the input content includes an input string determined based on the characters currently entered in the search bar.

[0022] In one embodiment, the number of characters in the input string is greater than a number threshold.

[0023] Secondly, this application also provides a question-answering method applied to electronic devices. The method includes: obtaining user input content on a question-answering interface; when the input content is a recommended question, generating a corresponding answer based on the content corresponding to the recommended question using a large language model; wherein the recommended question is the recommended question in the question recommendation method described above.

[0024] Thirdly, this application also provides a question recommendation device applied to an electronic device. The device includes: an input acquisition unit for displaying a question-and-answer interface and acquiring the input content currently displayed on the question-and-answer interface; and a question display unit for displaying recommended questions on the question-and-answer interface based on the input content. The recommended questions are generated based on user data, and the user data includes at least the user's local content on the electronic device.

[0025] Fourthly, this application also provides a question-answering device applied to an electronic device, the device comprising: a content acquisition unit for acquiring user input content on a question-answering interface; and an answering unit for generating a corresponding answer based on the content corresponding to the recommended question using a large language model when the input content is a recommended question; wherein the recommended question is the recommended question in the above-mentioned question recommendation method.

[0026] Fifthly, this application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method.

[0027] Sixthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0028] In a seventh aspect, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described method.

[0029] The aforementioned question recommendation method, question answering method, and related apparatus involve an electronic device displaying a question-and-answer interface, acquiring the current input content on the interface, and displaying recommended questions based on the input content. The recommended questions are generated based on user data, which includes at least the user's local content on the electronic device. This application prepares potential recommended questions by understanding various types of local user information and simultaneously understanding user behavioral preferences. When a user searches for questions and answers, embodiments of this application can provide recommended questions that match the user's preferences in real time, assisting the user in better searching and improving end-to-end efficiency and experience. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1A block diagram of a computing system that performs a visual or multimodal search service in one embodiment;

[0032] Figure 2 This is a structural block diagram of an electronic device in one embodiment;

[0033] Figure 3 This is a schematic diagram of the internal structure of a data processing system of an electronic device in one embodiment;

[0034] Figure 4 This is a diagram illustrating traditional dropdown suggestions.

[0035] Figure 5 This is a flowchart illustrating the traditional question-and-answer search process.

[0036] Figure 6 This is a flowchart illustrating a problem recommendation method in one embodiment;

[0037] Figure 7 This is a schematic diagram illustrating the triggering of the display of the question-and-answer interface in one embodiment;

[0038] Figure 8 This is a schematic diagram illustrating the display of recommended questions on a question-and-answer interface in one embodiment.

[0039] Figure 9 This is a flowchart illustrating the process of generating recommendation questions in one embodiment;

[0040] Figure 10 This is a flowchart illustrating the process of obtaining a recommendation question in one embodiment;

[0041] Figure 11 This is a flowchart illustrating the process of displaying recommended questions in a question recommendation method within one embodiment.

[0042] Figure 12 This is a flowchart illustrating the process of obtaining a list of problem materials in one embodiment;

[0043] Figure 13 This is a flowchart illustrating the problem assembly process in one embodiment;

[0044] Figure 14 This is a schematic diagram of the problem recommendation ranking results in one embodiment;

[0045] Figure 15 This is an example illustration of the effect of a problem recommendation method in one embodiment;

[0046] Figure 16 A schematic diagram of the overall framework of the problem is recommended for one embodiment;

[0047] Figure 17 This is a flowchart illustrating a question-and-answer method in one embodiment;

[0048] Figure 18 Here is a structural block diagram of a problem recommendation device in one embodiment;

[0049] Figure 19 This is a structural block diagram of a question-and-answer device in one embodiment. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0051] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the solutions, or any combination of multiple solutions. In addition, the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0052] Figure 1 A block diagram of an example computing system 100 performing a visual or multimodal search service according to an example embodiment of this application is depicted. The computing system 100 includes a user computing system 110, a server computing system 130, and / or a third computing system 150 communicatively coupled via a network 160.

[0053] User computing system 110 may include any type of computing device, such as, for example, a personal computing device (e.g., a laptop computer or desktop computer), a mobile computing device (e.g., a smartphone or tablet computer), a game console or controller, a wearable computing device (such as a smartwatch or smart glasses), an embedded computing device, or any other type of computing device.

[0054] User computing system 110 includes one or more processors 112 and memory 114. The one or more processors 112 can be any suitable processing device (e.g., processor core, microprocessor, ASIC (Application Specific Integrated Circuit), FPGA (Field Programmable Gate Array), controller, microcontroller, etc.) and can be a single processor or multiple processors operatively connected. Memory 114 may include one or more non-transitory computer-readable storage media, such as RAM (Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), EPROM (Erasable Programmable Read-Only Memory), flash memory devices, disks, etc., and combinations thereof. Memory 114 may store data 116 and instructions 118 executed by processor 112 to cause user computing system 110 to perform operations.

[0055] In some implementations, the user computing system 110 may store or include one or more machine learning models 120. For example, the machine learning model 120 may be, or may otherwise include, various machine learning models, such as neural networks (e.g., deep neural networks) or other types of machine learning models, including nonlinear and / or linear models. Neural networks may include feedforward neural networks, recurrent neural networks (e.g., long short-term memory recurrent neural networks), convolutional neural networks, or other forms of neural networks.

[0056] In some implementations, one or more machine learning models 120 may be received from server computing system 130 via network 160, stored in memory 114 of user computing system 110, and then used or otherwise implemented by one or more processors 112. In some implementations, user computing system 110 may implement multiple parallel instances of a single machine learning model 120 (e.g., performing parallel machine learning model processing across multiple instances of input data and / or detected features).

[0057] Alternatively or concurrently, one or more machine learning models 140 may be included in or otherwise stored and implemented by server computing system 130, which communicates with user computing system 110 according to a client / server relationship. For example, machine learning model 140 may be implemented by server computing system 130 as part of a network service (e.g., viewfinder service, visual search service, image processing service, ambient computing service, and / or overlay application service). Thus, one or more machine learning models 120 may be stored and implemented at user computing system 110, and / or one or more machine learning models 140 may be stored and implemented at server computing system 130.

[0058] Machine learning models 120 or 140 may include one or more generative models, one or more object detection models, one or more segmentation models, one or more classification models, one or more embedding models, one or more semantic analysis models, and / or one or more search engines, etc.

[0059] One or more generative models can be used to process displayed data and / or one or more processing outputs to generate natural language output (e.g., natural language output including additional information about the displayed data and / or entities associated with the data depicted in the displayed content), generate images, and / or other model-generated media content items. For example, one or more web resources can be accessed and processed to generate summaries on a specific topic. One or more object detection models can be used to perform object detection in the displayed data. One or more segmentation models can be used to segment objects and / or text fragments from the displayed content. One or more classification models can be used to perform object classification, image classification, entity classification, format classification, sentiment classification, and / or other classification tasks. One or more embedding models can be used to embed parts and / or all of the displayed data. The embeddings can then be used to search for similar objects and / or text, classify, group, and / or compress. Semantic analysis models can be used to process the displayed data to generate semantic outputs describing the understanding of the displayed data regarding topic understanding, scene understanding, focus, pattern recognition, application understanding, and / or one or more other semantic outputs.

[0060] One or more search engines may process displayed data, portions of the displayed data, and / or the output of one or more machine learning models to determine one or more search results. The one or more search results may include web pages, images, text, videos, and / or other data. Search results may be determined based on feature mapping, feature matching, embedding search, metadata search, tag search, clustering, and / or other search techniques. Search results may be determined based on query intent classification, search result classification, and / or entity classification. The model's output and / or search results may be sent back to the user's computing device to be provided to the user through one or more user interface elements generated and provided by a visual search interface.

[0061] User computing system 110 may also include one or more user input components 122 for receiving user input. For example, user input component 122 may be a touch-sensitive component (e.g., a touch-sensitive display or touchpad) that is sensitive to the touch of a user input object (e.g., a finger or stylus). Touch-sensitive components can be used to implement a virtual keyboard. Other example user input components include a microphone, a traditional keyboard, or other devices that a user can use to provide user input.

[0062] In some implementations, the user computing system 110 may store and / or provide one or more user interfaces 124, which may be associated with one or more applications. The one or more user interfaces 124 may be configured to receive input and / or provide data for display (e.g., image data, text data, audio data, one or more user interface elements, augmented reality experiences, virtual reality experiences, and / or other data for display). The user interface 124 may be associated with one or more other computing systems (e.g., server computing system 130 and / or third-party computing system 150). The user interface 124 may include a viewfinder interface, a search interface, a generative model interface, a social media interface, a media content gallery interface, etc.

[0063] User computing system 110 may include one or more sensors 126 and / or receive data from one or more sensors 126. The one or more sensors 126 may be housed in a housing assembly that houses one or more processors 112, memory 114, and / or one or more hardware components that may store one or more software packages and / or cause one or more software packages to be executed. The one or more sensors 126 may include one or more image sensors (e.g., cameras), one or more radar sensors, one or more audio sensors (e.g., microphones), one or more inertial sensors (e.g., inertial measurement units), one or more biosensors (e.g., heart rate sensors, pulse sensors, retinal sensors, and / or fingerprint sensors), one or more infrared sensors, one or more location sensors (e.g., Global Positioning System (GPS),) one or more touch sensors (e.g., conductive touch sensors and / or mechanical touch sensors), and / or one or more other sensors. One or more sensors may be used to obtain data associated with the user environment (e.g., images of the user environment, records of the environment, and / or the user's location).

[0064] User computing system 110 may include user computing device 111 and / or portions thereof. User computing device 111 may include mobile computing devices (e.g., smartphones or tablets), desktop computers, laptops, smart wearable devices and / or smart home appliances, or aircraft or vehicle-mounted devices, etc. Additionally and / or alternatively, user computing system 110 may acquire data from one or more user computing devices 111 and / or generate data using one or more user computing devices 111. For example, a smartphone camera may be used to capture image data depicting the environment, and / or an overlay application on user computing device 111 may be used to track and / or process data provided to the user. Similarly, one or more sensors associated with a smart wearable device may be used to acquire data about the user and / or about the user's environment (e.g., a camera housed in the user's smart glasses may be used to acquire image data). Additionally and / or alternatively, data may be acquired and uploaded from other user devices that may be specifically used for data acquisition or generation.

[0065] Server computing system 130 includes one or more processors 132 and memory 134. The one or more processors 132 can be any suitable processing device (e.g., processor core, microprocessor, ASIC, FPGA, controller, microcontroller, etc.) and can be a single processor or multiple processors operatively connected. Memory 134 may include one or more non-transitory computer-readable storage media, such as RAM, ROM, EEPROM, EPROM, flash memory devices, disks, etc., and combinations thereof. Memory 134 may store data 136 and instructions 138 that are executed by processor 132 to cause server computing system 130 to perform operations.

[0066] In some implementations, the server computing system 130 includes one or more server computing devices or is otherwise implemented by one or more server computing devices. Where the server computing system 130 includes multiple server computing devices, such server computing devices may operate according to a sequential computing architecture, a parallel computing architecture, or some combination thereof.

[0067] As described above, the server computing system 130 may store or otherwise include one or more machine learning models 140. For example, the machine learning model 140 may be, or may otherwise include, various machine learning models. Example machine learning models include neural networks or other multi-layer nonlinear models. Example neural networks include feedforward neural networks, deep neural networks, recurrent neural networks, and convolutional neural networks.

[0068] Additionally and / or alternatively, server computing system 130 may include search engine 142 and / or be communicatively connected to search engine 142, which may be used to crawl one or more databases (and / or resources). Search engine 142 may process data from user computing system 110, server computing system 130, and / or third-party computing system 150 to determine one or more search results associated with input data. Search engine 142 may perform term-based searches, tag-based searches, Boolean-based searches, image searches, embedding-based searches (e.g., nearest neighbor searches), multimodal searches, and / or one or more other search techniques.

[0069] Server computing system 130 may store and / or provide one or more user interfaces 144 for obtaining input data and / or providing output data to one or more users. The one or more user interfaces 144 may include one or more user interface elements, such as input fields, navigation tools, content tiles, optional tiles, widgets, data display carousels, dynamic animations, information pop-ups, image enhancement, text-to-speech, speech-to-text, augmented reality, virtual reality, feedback loops, and / or other interface elements.

[0070] User computing system 110 and / or server computing system 130 may train machine learning models 120 and / or 140 via interaction with a third-party computing system 150 communicatively coupled through network 160. The third-party computing system 150 may be separate from or part of the server computing system 130. Alternatively and / or additionally, the third-party computing system 150 may be associated with one or more network resources, one or more network platforms, one or more other users, and / or one or more contexts.

[0071] The third-party computing system 150 may include one or more processors 152 and memory 154. The one or more processors 152 may be any suitable processing device (e.g., processor core, microprocessor, ASIC, FPGA, controller, microcontroller, etc.) and may be a single processor or multiple processors operatively connected. The memory 154 may include one or more non-transitory computer-readable storage media, such as RAM, ROM, EEPROM, EPROM, flash memory devices, disks, etc., and combinations thereof. The memory 154 may store data 156 and instructions 158 executed by the processor 152 to cause the third-party computing system 150 to perform operations. In some implementations, the third-party computing system 150 includes one or more server computing devices or is otherwise implemented by one or more server computing devices.

[0072] Network 160 can be any type of communication network, such as a local area network (e.g., intranet), a wide area network (e.g., the Internet), or a combination thereof, and may include any number of wired or wireless links. Generally, communication over network 180 may be performed using various communication protocols (e.g., TCP / IP, HTTP, SMTP, FTP), encodings or formats (e.g., HTML, XML), and / or protection schemes (e.g., VPN, Secure HTTP, SSL) via any type of wired and / or wireless connection.

[0073] The machine learning models described in this application can be used for a variety of tasks, applications, and / or use cases.

[0074] In some implementations, the input to the machine learning model of this application can be image data. The machine learning model can process the image data to generate output. As an example, the machine learning model can process image data to generate image recognition output (e.g., image data recognition, latent embedding of image data, encoded representation of image data, hashing of image data, etc.). As another example, the machine learning model can process image data to generate image segmentation output. As another example, the machine learning model can process image data to generate image classification output. As another example, the machine learning model can process image data to generate image data modification output (e.g., image data alteration, etc.). As another example, the machine learning model can process image data to generate encoded image data output (e.g., encoded and / or compressed representation of image data, etc.). As another example, the machine learning model can process image data to generate upgraded image data output. As yet another example, the machine learning model can process image data to generate predictive output.

[0075] In some implementations, the input to the machine learning model of this disclosure can be text or natural language data. The machine learning model can process the text or natural language data to generate output. As an example, the machine learning model can process natural language data to generate language-encoded output. As another example, the machine learning model can process text or natural language data to generate latent text embedding output. As another example, the machine learning model can process text or natural language data to generate translation output. As another example, the machine learning model can process text or natural language data to generate classification output. As another example, the machine learning model can process text or natural language data to generate text segmentation output. As another example, the machine learning model can process text or natural language data to generate semantic intent output. As another example, the machine learning model can process text or natural language data to generate upgraded text or natural language output (e.g., text or natural language data of higher quality than the input text or natural language, etc.). As another example, the machine learning model can process text or natural language data to generate predictive output.

[0076] In some implementations, the input to the machine learning model of this disclosure can be speech data. The machine learning model can process the speech data to generate an output. As an example, the machine learning model can process speech data to generate a speech recognition output. As another example, the machine learning model can process speech data to generate a speech translation output. As another example, the machine learning model can process speech data to generate a latent embedding output. As another example, the machine learning model can process speech data to generate an encoded speech output (e.g., an encoded and / or compressed representation of the speech data). As another example, the machine learning model can process speech data to generate an upgraded speech output (e.g., speech data of higher quality than the input speech data). As another example, the machine learning model can process speech data to generate a text representation output (e.g., a text representation of the input speech data). As yet another example, the machine learning model can process speech data to generate a predicted output.

[0077] In some implementations, the input to the machine learning model of this disclosure can be sensor data. The machine learning model can process the sensor data to generate output. As an example, the machine learning model can process sensor data to generate a recognition output. As another example, the machine learning model can process sensor data to generate a prediction output. As another example, the machine learning model can process sensor data to generate a classification output. As another example, the machine learning model can process sensor data to generate a segmentation output. As another example, the machine learning model can process sensor data to generate a segmentation output. As another example, the machine learning model can process sensor data to generate a visualization output. As another example, the machine learning model can process sensor data to generate a diagnostic output. As another example, the machine learning model can process sensor data to generate a detection output.

[0078] In some cases, the input includes visual data, and the task is a computer vision task. In other cases, the input includes pixel data from one or more images, and the task is an image processing task. For example, an image processing task could be image classification, where the output is a set of scores, each corresponding to a different object class and representing the probability that one or more images depict an object belonging to said object class.

[0079] User computing system 110 may include multiple applications (e.g., applications 1 to N). Each application may include its own corresponding machine learning library and machine learning model. For example, each application may include a machine learning model. Example applications include text messaging applications, email applications, dictation applications, virtual keyboard applications, browser applications, etc. In some implementations, each application may use an API (e.g., a public API across all applications) to communicate with a central intelligence layer (and the models stored therein).

[0080] The central intelligence layer may include multiple machine learning models. For example, a corresponding machine learning model (e.g., a model) may be provided for each application, and the machine learning models are managed by the central intelligence layer. In other implementations, two or more applications may share a single machine learning model. For example, in some implementations, the central intelligence layer may provide a single model (e.g., a single model) for all applications. In some implementations, the central intelligence layer is included within the operating system of the computing system 1100 or otherwise implemented by the operating system.

[0081] The central intelligence layer can communicate with the central device data layer. The central device data layer can be a centralized data repository for the computing system 100. The central device data layer can communicate with many other components of the computing device, such as, for example, one or more sensors, a context manager, a device state component, and / or additional components. In some implementations, the central device data layer can communicate with each device component using an API (e.g., a private API).

[0082] User computing device 111 can be an electronic device. Figure 2 This is a schematic diagram of the data processing system framework of electronic device 200 in some embodiments. Electronic device 200 includes a data acquisition module 210, a large model 220, and a database 230. Electronic device 200 can realize a series of functions such as data acquisition, preprocessing, storage, mining, retrieval, recommendation, and question answering for various applications.

[0083] In the process of displaying application page content or web page content, the electronic device 200 can detect that the user has started the data recording function through a preset triggering method. The preset triggering method can be the triggering of the target function key, the gesture triggering (such as three-finger swiping up), or the triggering of the data recording application, etc.

[0084] The data acquisition module 210 is used to record the screen display content of the electronic device 200 or the page content of the application running in the background, based on the enabled data recording function. The screen display content or the page content of the background application can include data such as text, images, videos, and audio. If it is text, the text data can be directly acquired. If it is an image, the image can be downloaded based on its URL. If it is a video, the URL address of the video can be obtained, and all or part of the video content can be downloaded based on the URL address. If it is audio, the download address of the audio can be obtained, and all or part of the audio data can be downloaded. The data acquisition module 210 can also record the screen display content by taking screenshots or by collecting user-inputted voice data through a microphone for preprocessing.

[0085] The data acquisition module 210 is also used to call the large model 220 or some algorithms to preprocess the recorded screen display content to obtain data in the target format. For example, if the screen display content is text, entity extraction and summarization can be performed; if the screen display content is an image, image recognition can be performed to obtain image content and classify the image; if the screen display content is video, keyframe extraction and keyframe content recognition can be performed, as well as the generation of video summary information. The processed data can be stored in the database 230. The large model 220 can be... Figure 1 Machine learning models in [the context of this].

[0086] Database 230 is a carrier or system used for storing, managing, and retrieving information. Its form varies depending on the storage object, usage scenario, and technical architecture. Database 230 can store data and possess at least one management capability, such as classification, retrieval, synchronization, and access control. Database 230 can be an application, mini-program, API interface, personal dedicated storage hardware, embedded hardware recording module, cloud service database, browser plugin, bookmark management tool, enterprise knowledge base system, cached database, personal information database, etc. Personal dedicated storage hardware can be a portable hard drive with a dedicated management system, encrypted USB flash drive, personal cloud storage hard drive, etc. Embedded hardware recording module can be a dedicated storage module in a smart device used to record various types of data. Cloud service database can be a cloud database providing interface call service functions or a database for personal cloud data management services.

[0087] Database 230 includes a data management module 240 and a data interaction module 250. The data management module 240 can call the large model 220 to mine the stored data, and store the mined data in the database 230, providing data retrieval functions, etc. Mining methods may include data anonymization, tag classification, entity extraction, schedule extraction, to-do item extraction, summary generation, collection generation, recommendation collection, data association, etc. The data management module 240 can also summarize and organize various types of user-recorded data to obtain user profile information. User profile information may include user personal information, personal preference information, etc. User personal information may include user identity information, occupation, education, family information, etc., while personal preference information may include favorite people, things, places, scenery, games, etc., and is not limited to these.

[0088] The data interaction module 250 can provide interaction functions with the application and can obtain the data required for interaction through the data management module 240. Interaction functions may include, but are not limited to, recommendation collections, data search, data detail viewing, data sharing, AI Q&A, and data association. For example, in the AI ​​Q&A function, the data interaction module 250 obtains the question provided by the user in the AI ​​Q&A application and sends the question to the data management module 240. The data management module 240 searches for the corresponding answer and provides it to the data interaction module 250 for feedback to the relevant AI Q&A application.

[0089] Furthermore, the data collected by the data acquisition module 210 in the electronic device 200 and the user profile information mined can be synchronized to the cloud as needed. The collected data can also be processed with the help of the large model in the cloud. The data synchronized to the cloud can also be synchronized to other electronic devices through the cloud to realize data sharing between different devices.

[0090] Figure 3 This is a schematic diagram of the internal structure of the data processing system of an electronic device in some embodiments. The data acquisition module 210 may include an acquisition entry submodule 211, an extraction submodule 212, a screenshot acquisition submodule 213, a preprocessing submodule 214, and a writing submodule 215, etc.

[0091] The data collection entry submodule 211 provides an entry point for data collection. This could be a voice assistant entry point, receiving user voice input commands for data recording; a gesture trigger entry point, such as a long entry point for a hardware button or a three-finger upward swipe entry point; or a user-triggered entry point for specific controls, such as a like control, a favorite control, or a share control. The data collection entry submodule 21 can automatically collect data according to set rules. For example, if the collection trigger condition is preset page content, and the electronic device displays page information containing preset page content, then that page information will be automatically collected. The data collection entry submodule 21 can also receive donated data from applications (such as weather apps, clock apps, pedometer apps, etc.), including weather data, time data, walking data, running mileage, etc., but is not limited to these.

[0092] The extraction submodule 212 can extract images and text from the page content of the application that needs to be recorded.

[0093] The screenshot acquisition submodule 213 can obtain text content by taking a screenshot of the screen content and then calling the OCR recognition algorithm to perform text recognition on the screenshot content.

[0094] The preprocessing submodule 214 can perform word segmentation, noise reduction, or process the collected data into a preset format.

[0095] The write submodule 215 is used to write the collected data or preprocessed data into the database 230.

[0096] The data management module 240 may include a data mining submodule 241, a question-and-answer submodule 242, a retrieval submodule 243, a source data processing submodule 244, etc.

[0097] The data mining submodule 241 is used to mine data. Mining methods may include data anonymization, tag classification, entity extraction, schedule extraction, to-do item extraction, summary generation, collection generation, recommendation collection, data association, etc.

[0098] The question-and-answer submodule 242 is used to search for relevant data based on user questions and generate corresponding answers based on the searched data.

[0099] The retrieval submodule 243 is used to retrieve relevant data from the database 230 according to the search request, and to organize the retrieved data or directly feed it back to the user.

[0100] The source data processing submodule 244 provides operations for adding, modifying, deleting, and updating source data. Source data can refer to data collected and recorded in database 230, or data collected after preprocessing.

[0101] The data interaction module 250 includes a data addition / editing / viewing submodule 251, a data recommendation submodule 252, a data search submodule 254, and a related data submodule 254.

[0102] The data addition / editing / viewing submodule 251 provides functions for adding, editing, and viewing data.

[0103] The data recommendation submodule 252 is used to recommend aggregated data to users or recommend stored data according to user profiles.

[0104] The data search submodule 253 is used to obtain the user's search requirements and transmit them to the retrieval submodule 243 in the data management module 240. The retrieval submodule 243 retrieves the corresponding data from the database 230 according to the search requirements and feeds the retrieved data back to the data search submodule 253.

[0105] The associated data submodule 254 is used to associate or recommend recorded data to other applications.

[0106] In traditional technologies, taking electronic devices (such as mobile phones) as examples, mobile phones can be used to obtain, communicate, and record information in different forms, fields, and scenarios. Examples include text messages, chat logs (screenshots), emails, various documents, notes, schedules, to-do lists, and browsed / bookmarked web pages. However, after a period of time, or when these fragmented pieces of information accumulate to a certain quantity, users' memories often become blurred, leading to situations where they "remember seeing it, only remember some keywords or the general meaning, but can't recall where it was or how to accurately find it."

[0107] Traditional technologies related to search and question answering mainly include: ① In internet search engines, the use of drop-down suggestion keywords, such as... Figure 4 As shown, statistical analysis of user search behavior is used to store popular search terms in a database, and related recommendations are made during the user's search input process (search process). Similarly, some vertical industries, such as medical and legal, have also adopted this information retrieval technology. ② A knowledge base and retrieval question-and-answer system are implemented based on the Retrieval Augmentation Generation (RAG) system. The content is semantically vectorized (Embedded) and stored in the database. When a user queries, the semantic vector similarity between the question and the content is calculated to find the corresponding content. Then, the Large Language Model (LLM) combines the question and the content to generate a generative answer. The process is as follows. Figure 5 As shown in the diagram, Reranker represents the fine-tuning of database search results, prioritizing more relevant original text results. ③ In some question-answering systems, online questions are stored, and after an answer is completed, similar pre-set questions are matched from the database according to a certain strategy for post-search recommendations.

[0108] Furthermore, traditional database / knowledge base-based question-answering methods mainly include pure cloud-based large-model processing solutions and pure local retrieval and response solutions. Among these, the pure cloud-based large-model processing solution: Most mainstream intelligent question-answering applications currently employ a pure cloud-based processing architecture. After the user inputs a query request in the application interface, the terminal device encapsulates the query request into a data packet and transmits it to the cloud server cluster via network protocols. The large-scale language model deployed in the cloud, relying on a distributed computing framework, performs semantic parsing, knowledge extraction, and logical reasoning on the query request, generates a structured response result, and then transmits it back to the user terminal via HTTP / HTTPS protocols for rendering and display. Throughout the entire interaction process, the terminal device only undertakes the human-computer interaction interface function; the core natural language processing and knowledge generation tasks are supported by cloud computing power. The pure local retrieval and response solution: Some lightweight question-answering applications adopt a pure terminal retrieval architecture. These applications store a structured knowledge base on the terminal through an embedded database and integrate a retrieval engine based on the BM25 algorithm. When the user inputs a query request, the terminal calls the retrieval module to perform keyword segmentation, inverted index matching, and relevance ranking, directly returning the Top-N retrieval results to the application layer. If the retrieval recall rate is zero, the preset empty result response mechanism is triggered.

[0109] The disadvantages of traditional pure cloud-based large-scale model processing solutions include at least the following: 1. Inability to classify query requests by scenario and plan processing paths. All query requests adopt a unified cloud-based large-scale model inference process. For simple queries that can be covered by the local knowledge base (such as personal calendar queries and local file retrieval), cloud computing resources are still consumed, resulting in wasted computing power. 2. Lack of semantic representation capabilities. Relying solely on the word embedding layer of the large model itself for semantic understanding leads to low semantic mapping accuracy in scenarios such as professional domain terminology and long-tail expressions, affecting the accuracy of responses. 3. Inefficient resource scheduling. Because it is impossible to classify query requests by scenario and plan processing paths, it is impossible to dynamically allocate cloud / terminal resources according to the complexity of query requests. This results in a high proportion of high-end GPU (Graphics Processing Unit) computing power being consumed by simple queries, leading to significant resource mismatch. The disadvantages of traditional pure local retrieval and response solutions include at least the following: 1. Incomplete retrieval chain, resulting in low relevance accuracy of Top-N retrieval results. In multi-turn dialogue scenarios, the drift rate of retrieval results increases. 2. Low semantic understanding level leads to low retrieval recall rates for synonyms and cross-domain expressions. 3. Discontinuous generation capabilities prevent the completion of complex tasks such as information integration and logical reasoning, resulting in low response efficiency in query scenarios requiring multi-source information fusion.

[0110] Furthermore, from the perspective of practical application scenarios, users need to query personal content on their phones and receive real-time search suggestions when they only remember a small amount of information, such as providing search suggestions based on a small amount of input. Traditional solutions have at least the following shortcomings: (1) Recommended questions are given after the answer is completed, but users actually expect to receive complete search questions when they input a small amount of key information. (2) The recommended question library is built based on a large number of users' historical questions, representing the statistical popularity of questions, but these questions are not necessarily what users want to ask. Users expect to be recommended questions related to their own information, rather than network questions unrelated to local information. (3) The same search may correspond to multiple different information types. The list of recommended questions is based on statistical information, and the order of multiple questions is fixed, which is inconsistent with the user's own intention and preference.

[0111] In summary, traditional methods lack effective solutions for information retrieval and question answering when users have fuzzy memories. This application provides a personalized search recommendation scheme for client-side content question answering. It understands various types of local user information, prepares potential recommended questions, and comprehends user behavioral preferences. When a user performs a local search for questions and answers, it can provide recommended questions that match the user's preferences in real time during the input process. For example, it can provide personalized search question completion recommendations while the user is inputting, assisting the user in making better queries, reducing user input time, and improving end-to-end efficiency and experience. Furthermore, this application can filter matching memory data from a large amount of memory data to provide answers, ensuring the accuracy and extensibility of the question and answer, and reducing the latency of displaying the first character of the question and answer, greatly improving the user experience.

[0112] It is understood that this application can be applied to the fields of information retrieval and artificial intelligence. This application provides a personalized search question recommendation scheme for client-side content question answering. It should be noted that the beneficial effects or technical problems solved by this application are not limited to this one; they may also be other implicit or related problems, as detailed in the following description of the embodiments.

[0113] Before introducing the specific embodiments of this application, the technical terms involved in this application will be explained:

[0114] RAG: Retrieval-Augmented Generation.

[0115] LLM: Large Language Model.

[0116] RRF: Reciprocal Rank Fusion.

[0117] AIGC: Artificial Intelligence Generated Content.

[0118] Embedding: An approach for mapping high-dimensional inputs (e.g., textual or visual content) onto compact latent vector representations which preserve inter-sample semantic correlations.

[0119] DMP: Search Service, a converged search service.

[0120] Reranker: Reranking Model, a model or algorithm that re-ranks the retrieved results.

[0121] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0122] In one exemplary embodiment, such as Figure 6 As shown, a problem recommendation method is provided. Taking the application of this method to an electronic device as an example, the method includes the following steps 302 to 304.

[0123] Step 302: Display the Q&A interface and obtain the current input content on the Q&A interface.

[0124] Specifically, the electronic device can display a question-and-answer interface. Optionally, the question-and-answer interface can refer to a user interface used for inputting a question (Query) and displaying the answer content; for example, the question-and-answer interface can be a pop-up, a floating layer, or a standalone interface. Further, the question-and-answer interface can be an interactive interface between the electronic device and the user, and the question-and-answer interface can include a display area and a question input area, etc.; wherein, the input area can be used to obtain the user's input content (e.g., inputting a question Query), and the display area can be used to display recommended questions.

[0125] Step 304: Based on the input content, display recommended questions on the question-and-answer interface. The recommended questions are generated based on user data; the user data includes at least the user's local content on the electronic device.

[0126] Specifically, the electronic device can display a question-and-answer interface and, based on the input content, display recommended questions on the interface. The electronic device can match potential recommended questions (i.e., recommended questions generated based on user data) according to the input content, thereby determining the recommended questions displayed on the question-and-answer interface. For example, embodiments of this application can provide recommended questions that match user preferences in real time during user input, assisting users in making better queries, while reducing user input time and improving end-to-end efficiency and experience.

[0127] Taking a question-and-answer interface that includes a display area and an input area as an example, when there is input content in the input area, based on the embodiments of this application, the display area can display recommended questions after the input content has been completed, and the completed recommended questions match the user's preferences, thereby reducing the user's input time. Furthermore, even when there is no input content in the input area, based on the embodiments of this application, the electronic device can still display recommended questions that match the user's preferences in the display area of ​​the question-and-answer interface to assist the user in making better searches.

[0128] The recommendation problem in this embodiment is generated based on user data, which can refer to a user's personal content. For example, user data can refer to various types of local user data (also known as various types of local user information). In this embodiment, user data at least includes the user's local content on the electronic device; optionally, user data can include online bookmarks. Further, user data includes both online bookmarks and the user's local content on the electronic device; that is, user data is divided into two categories: online bookmarks and local content. It is understood that online bookmarks do not involve user privacy, while local content may.

[0129] Electronic devices analyze and process various types of local user data to identify potential recommended questions. These questions can then be displayed on a question-and-answer interface, enabling real-time question recommendations. Furthermore, when a user enters search input, the electronic device can also match prepared questions (i.e., the aforementioned potential recommended questions) in real-time based on the user's input, thus providing real-time question recommendations. This application addresses the pain points of users who "can't remember, don't know how to ask, or find typing cumbersome." For example, once potential recommended questions are obtained, they can be stored in a database. When the question-and-answer interface is displayed, the recommended questions in the database can be shown; and when a user enters search input, the electronic device can also match recommended questions from the database in real-time.

[0130] Optionally, the electronic device can analyze and process various types of local user data to obtain recommended question materials, which are then stored in a database. When the user enters a search query, the electronic device can match the question materials in the database in real time based on the user's input and combine it with a series of recommendation logics to provide real-time question recommendations to the user. Furthermore, even if the user does not enter a search query, the electronic device can still display recommended questions based on the question materials on the question-and-answer interface.

[0131] Furthermore, regarding the local content of the user on the electronic device in this application, the local content may include, but is not limited to, documents, notes, schedules, to-do lists, text messages, and emails. In some embodiments, this application can expand to include more modalities of local data, and the local content may also include audio, images, and videos. It is understood that audio can be converted to text using ASR (Automatic Speech Recognition) technology, images can be described using OCR (Optical Character Recognition) technology combined with captioning technology to obtain image text descriptions, and videos can be described using a multimodal large model to obtain video text descriptions. It should be noted that the above-described methods for processing local content are merely illustrative descriptions, and other forms of processing local content may also be used, not limited to those mentioned in the above embodiments, as long as they can achieve the function of processing the relevant data into corresponding text.

[0132] Traditional solutions rely on hot keyword statistics for internet search questions, failing to consider most of the user's local content and thus unable to offer relevant questions. This application's question recommendation method, however, achieves personalized search recommendations based on device-side content. The questions recommended by the electronic device are generated from the user's personal content, and are prepared based on that content, ensuring all recommended questions are related to the user's personalized information. This application addresses the problem of users having vague memories of their personal knowledge, content, and information, and not knowing how to search. Simultaneously, it reduces user input time, improves efficiency, and enhances the search experience.

[0133] In one embodiment, displaying a Q&A interface includes: displaying the Q&A interface in response to a triggering operation on the Q&A entry point; the Q&A entry point is used to provide Q&A functionality after being triggered.

[0134] Specifically, the electronic device can display a question-and-answer interface in response to a triggering operation on the question-and-answer entry point, wherein the question-and-answer entry point is used to provide question-and-answer functionality after being triggered. For example, triggering operations on the question-and-answer entry point include, but are not limited to: click operations, long press operations, swipe operations, touch operations, button operations, gesture operations, and free combinations of at least two of the above operations. It is worth noting that the triggering operations on the question-and-answer entry point include, but are not limited to, the aforementioned types. Those skilled in the art should understand that any operation capable of achieving the above functions falls within the protection scope of the embodiments of this application.

[0135] In this embodiment of the application, when the electronic device triggers the question and answer function at the question and answer entry point, it displays a question and answer interface, and then displays recommended questions obtained based on user data on the question and answer interface. This enables the device to provide recommended questions that match user preferences in real time, helping users to make better queries, while reducing user input time and improving end-to-end efficiency and experience.

[0136] In one embodiment, the question-and-answer entry point may include a button control; the triggering operation includes one or more of the following: a touch operation on the button control, a voice input operation, a touch operation on the fingerprint sensor, and a key operation. Specifically, the question-and-answer entry point in this application embodiment may be a button control. This application does not limit the specific implementation form of the question-and-answer entry point. For example, the question-and-answer entry point may include at least one of a button control, an image button control, a combo box control, and a user-created control.

[0137] Furthermore, the triggering operation can include one or more of the following: touch operation on a button control, voice input operation, touch operation on a fingerprint sensor, and button operation. Taking a user's question-and-answer operation in the app as an example, such as... Figure 7 As shown, when a user opens the app, interface 700 is displayed. The user can click the "Ask a Question" button 702 on interface 700 to enter... Figure 8 The question-and-answer interface 800 shown exemplifies this feature. The electronic device displays three cached recommended questions 82 on the interface. These recommended questions are generated based on user data, reflect user preferences, and assist users in making better searches. It is understandable that... Figure 8 The question-and-answer interface 800 may also include an input area 84, which can be used to obtain user input. For example, the input area 84 may be a search bar.

[0138] Regarding the generation of recommendation questions, in some embodiments, user data may also include online favorites; such as... Figure 9 As shown, the method may also include steps 402 to 406.

[0139] Step 402: Generate questions based on online collections to obtain the first question material.

[0140] Specifically, in this application embodiment, "online favorites content" can refer to internet favorites content, such as web pages that a user has viewed, saved, or liked. The internet favorites content includes the user's personalized characteristics. The electronic device can generate questions based on the online favorites content to obtain the first question material. Taking internet favorites content as an example, the electronic device can generate search questions based on the internet favorites content to obtain the first question material.

[0141] Step 404: Generate questions based on local content to obtain the second question material;

[0142] Specifically, the electronic device can generate questions based on local content to obtain second question materials. For example, the specific implementation of generating questions based on local content can be the same as, or different from, generating questions based on online collections; this application does not limit this approach.

[0143] Step 406: Generate a recommendation question based on at least one of the first question material and the second question material.

[0144] Specifically, upon receiving first and second question materials, the electronic device can generate a recommended question based on at least one of them. For example, based on user input, if the first question material is matched, a recommended question is generated based on it; if the second question material is matched, a recommended question is generated based on both. Furthermore, this application can also integrate local content and internet content search recommendation questions. For instance, it can jointly recommend based on the first and second question materials, and the mixed sorting constitutes fusion. This process requires consideration of the sorting method to better assist users in their searches, while reducing user input time and improving end-to-end efficiency and experience.

[0145] In one embodiment, the first question material includes multiple first pairs, each pair including a first question statement and corresponding online collection content; generating the first question material based on the online collection content includes: analyzing and processing the online collection content through a large language model to obtain the first question statement.

[0146] Specifically, electronic devices can analyze and process online saved content using a large language model to obtain the first question statement. Furthermore, the electronic device can combine the generated recommended questions with corresponding content into pairs and store them in a database. When a user clicks on a recommended question, the corresponding content is directly provided to the large language model for an answer, thus eliminating the need to search the database and improving question-answering efficiency. For example, the first question material may include multiple first pairs, each pair consisting of the first question statement and corresponding online saved content.

[0147] In one embodiment, the first question statement includes a sentence-like statement. Specifically, the question statement generated by the electronic device based on online bookmarks can be a sentence-like statement. It is understood that the question statement in this embodiment can serve as a recommended question.

[0148] In one embodiment, the second question material includes multiple second pairs, each pair comprising a second question statement and corresponding local content; questions are generated based on the local content to obtain the second question material, including:

[0149] Obtain permission information corresponding to local content; permission information is used to indicate the cloud's access permissions for local content;

[0150] If the permission information determines that the cloud has permission to access local content, then the local content is processed through the large language model to obtain the second question statement;

[0151] If the permission information determines that the cloud does not have permission to access local content, then the local content is processed through the client-side model to obtain the second question statement.

[0152] Specifically, given that online bookmarks do not involve user privacy, while local content may, users can choose whether to accept data upload to the cloud when generating search questions for local content. If they accept, local content can also be uploaded to the cloud and used to generate search questions using the same technical methods as online bookmarks. If they do not accept, local content data can be processed only on the device side, for example, by using a device-side model to filter spam and extract key information to obtain a second search question, which can then be used as a device-side recommendation question. Optionally, the device-side model can include, but is not limited to, deep learning models such as LSTM (Long Short-Term Memory) and TinyBERT, where TinyBERT is a simplified model based on BERT (Bidirectional Encoder Representations from Transformers).

[0153] In some embodiments, the second question statement includes key information in the form of a phrase or a short sentence. Specifically, the question statement obtained by the electronic device from question generation based on local content can be key information in the form of a phrase or a short sentence.

[0154] To further illustrate this application, a specific example is provided below. Taking internet-collected content as an example, electronic devices can analyze and process various types of local user data through the question preparation module to obtain potential recommended question materials, which are then stored in a database (also known as a question database). Furthermore, the question preparation module can include the following sub-modules: local content question preparation, internet-collected content question preparation, question-answer mapping, question content security review, and question indexing, as follows: Figure 10 As shown.

[0155] User data can be divided into two categories: internet favorites and local content. Local content includes documents, notes, calendars, to-do lists, SMS messages, and emails. Internet favorites do not involve user privacy, while local content may. When generating search questions for internet content, a large language model can be used on the cloud to upload the content and generate sentence-style question statements (internet content questions). For local content, users can choose whether to accept data upload to the cloud. If they accept, the local content is also uploaded to the cloud using the same technical approach as internet favorites; if they do not accept, the data is processed only on the device side, using a device-side model to filter spam and extract key information, obtaining key information in the form of phrases or short sentences, which serves as the device-side content recommendation question (local content question).

[0156] In practical applications, recommended questions must comply with content security and must not recommend prohibited text. After obtaining potential recommended questions, they can be reviewed by the content security module to remove non-compliant recommended questions.

[0157] Considering that if a user clicks on a recommended question, searching the entire dataset would be as inefficient as the existing logic, this application's embodiment uses content (internet saved content, local content) as the recommended question input. This content can then be used as the input for answering the question, eliminating the need to search the database and improving question-and-answer efficiency. Figure 10 As shown, this application stores the generated recommendation question and its corresponding content (Query, Content) together in the database. When a user clicks on a recommendation question, the corresponding content is directly provided to the large language model for answering.

[0158] Understandable. Figure 10The database is located on the client side, thus ensuring user data privacy and security. Furthermore, the generated recommendation question and its corresponding content (Query, Content) are jointly stored in the database. When a user clicks on a recommendation question, the corresponding content can be directly provided to the large language model for an answer. Taking the RAG system as an example, the original content is divided into chunks, and indexed, stored, and retrieved in units of "chunks." In this embodiment, the content corresponding to the recommendation question is also in units of "chunks." If uploaded to the cloud, only the corresponding chunk of content is uploaded, not the entire content. This will be clarified to users in the user manual when the feature is enabled in the actual product. If the user does not accept chunk-based cloud uploads, the entire feature will not be enabled. Optionally, this embodiment also applies to situations where the user accepts partial cloud uploads, i.e., the scope of cloud uploads can be selected.

[0159] Furthermore, the embodiments of this application can utilize the existing RAG system's technical capability for retrieval via semantic vectors (Embedding), while also supporting traditional retrieval methods such as keyword retrieval. The database storage simultaneously establishes inverted indexes and semantic vector indexes. The embedding model required to establish the semantic vector index can be an open-source model or a model finely tuned based on the system's historical data distribution.

[0160] In one embodiment, if the electronic device does not currently receive user input on the question-and-answer interface, displaying recommended questions on the interface may include: if no user input is currently received, then displaying randomly selected recommended questions. Specifically, when no user input is currently received, the electronic device may display randomly selected recommended questions. For example, recommended questions may be randomly selected from a database and then displayed. It is understood that the method for obtaining the randomly selected recommended questions can be found in the descriptions of the various embodiments below, and will not be repeated here.

[0161] In an exemplary embodiment, given that the electronic device currently obtains the user's input content on the question-and-answer interface, recommended questions are displayed on the interface based on the input content, such as... Figure 11 As shown, steps 502 to 504 are included.

[0162] Step 502: If the input content does not meet the question-and-answer criteria, a randomly selected recommended question will be displayed.

[0163] Specifically, the system obtains the user's input on the question-and-answer interface. If the input does not meet the question-and-answer criteria, the electronic device displays randomly selected recommended questions.

[0164] Step 504: If the input content meets the question-and-answer criteria, then determine the recommended questions to be displayed on the question-and-answer interface based on the input content.

[0165] Specifically, once the user's input on the question-and-answer interface is received, and if the input meets the question-and-answer criteria, the electronic device can determine the recommended questions to display on the interface based on the input. These criteria can be used to segment the user's input, resulting in faster output and even faster and more accurate answers.

[0166] In one embodiment, the randomly selected recommended question is related to the time the recommended question was generated. Specifically, the recommended questions in this embodiment conform to user preferences. When no user input is currently available on the question-and-answer interface, or when user input is available but does not meet the question-and-answer criteria, the electronic device can select a recommended question according to a corresponding strategy. In this embodiment, the randomly selected recommended question is related to the time the recommended question was generated, thus aligning with the user's actual usage habits and meeting product design requirements.

[0167] In one embodiment, the randomly selected recommended questions include those whose generation time is closest to the display time of the question-and-answer interface; and / or, the randomly selected recommended questions include those whose generation time is furthest from the display time of the question-and-answer interface. Specifically, given that users may want to search for things they have recently viewed, this application proposes that the randomly selected recommended questions include those whose generation time is closest to the display time of the question-and-answer interface, thereby giving users a sense of familiarity and making them feel that the electronic device displaying the recommended questions understands the user (having viewed the user's saved items). Alternatively, the strategy can be changed to recommend the earliest entries in the database, i.e., the randomly selected recommended questions include those whose generation time is furthest from the display time of the question-and-answer interface; this application is not limited to this.

[0168] In one embodiment, the question-answering conditions include at least one of determining that the input content is meaningful and determining that the input content has a question-answering intent. Specifically, the electronic device can triage user input content, for example, determining whether the input content is meaningful or whether the input content has a question-answering intent. This can be determined through query intent recognition (e.g., semantic understanding), or by using regular expression checks, machine learning models, etc., to quickly check whether the input content is meaningful. For example, if the user inputs a string without semantic meaning, it can be determined that the input content is meaningless.

[0169] The electronic device in this embodiment may further include a search-based question recommendation module. This module can, based on user input, match question materials (also known as query materials) in a database in real time, and combine this with a series of recommendation logics to provide real-time question recommendations to the user, addressing the pain points of users "not remembering, not knowing how to ask, and finding typing troublesome." Optionally, the search-based question recommendation module may include question script management, search matching, question assembly logic management, and question recommendation logic management.

[0170] Regarding question script management, search matching, question assembly logic management, and question recommendation logic management, in some embodiments, determining the recommended questions displayed on the Q&A interface based on the input content includes: clustering the input content into scripts to obtain clustered input content; using multiple retrieval methods to search the clustered input content based on first and second question materials to obtain multiple retrieval results; the first question material is generated from questions based on online collections; the second question material is generated from questions based on local content; the multiple retrieval results are merged and sorted to obtain a question material list; and recommended questions are generated based on the question materials in the question material list.

[0171] Specifically, the electronic device can perform dialogue clustering on the input content to obtain clustered input content. This dialogue clustering can be used for question dialogue management, allowing for better matching and recommendation during the user input matching process by clustering and managing the input content of various user types. Furthermore, the electronic device can use multiple retrieval methods to search the clustered input content based on first and second question materials, obtaining multiple search results. These results are then merged and sorted to obtain a question material list, and recommended questions are generated based on the question materials in the list. This application prepares questions based on user content; assuming the input matches the question material, all recommended questions are related to the user's personalized information.

[0172] In one embodiment, the input content is clustered to obtain clustered input content, including: in response to detecting terms in the input content, clustering the terms to obtain clustered input content; wherein, the terms include one or more of the following: prepositional terms, time terms, name terms, alphabetic terms, and colloquial terms; the clustering process includes filtering prepositional terms, a question material matching strategy for time terms, and standardization processing for name terms, alphabetic terms, and colloquial terms.

[0173] Specifically, regarding question wording management, when users search, depending on their individual language habits, they may directly begin typing key information, or they may first enter pre-selective prompts (i.e., pre-selective terms) such as "help me / find it." These pre-selective prompts will not appear in the prepared recommended questions. For example, a user might want to find schedule-related information and enter relative or absolute time terms (i.e., time terms) such as "last week / tomorrow / this month / Saturday / National Day." The absolute times corresponding to these terms change with the days, requiring specific identification and logical processing.

[0174] This application embodiment can cluster and manage various possible user input terms to facilitate matching and recommendation during the user input process. For example, terms can include the following five categories; in practice, more types can be identified and summarized based on actual circumstances.

[0175] For pre-terms (e.g., pre-class input Term), the word representation is as follows:

[0176] [Please, please, please help, please help me, please find, please check, please search, please give me, please help me, find, view, query, search, find for me, check for me, search for me, help me find, help me check, help me search, search below, find below, check below, search once, find once, check once] etc.

[0177] When the system detects that a user has entered this type of term, it filters the input and starts matching from subsequent inputs. For example, if a user enters "help me find Yunnan", the system will automatically filter the term after matching "help me find" and search for "Yunnan" in the database.

[0178] For time-related terms (such as time-bound terms), the word representation is as follows:

[0179] Schedule / to-do list keywords: [arrangement, plan, to-do, calendar, schedule, to-do items, to-do matters], etc.

[0180] Time and holiday type words: [tomorrow, tomorrow, yesterday, yesterday, today, today, the day after tomorrow, the day after tomorrow, the day before yesterday, the day before yesterday, morning, afternoon, evening, dusk, night, this week, this week, last week, next week, week [Mon|Tue|Wed|Thu|Fri|Sat|Sat], this week [Mon|Tue|Wed|Thu|Fri|Sat|Sat], this week [Mon|Tue|Wed|Thu|Fri|Sat|Sat], this week, this week, next week, week [Mon|Tue|Wed|Thu|Fri|Sat|Sat], weekend, weekday [Mon|Tue|Wed|Thu|Fri|Sat|Sat|Sun], this month, last month, next month, beginning of the month, end of the month, this year, last year, next year, the year before last, the year after last, New Year's Day, Lunar New Year's Eve, Spring Festival, Lantern Festival, Women's Day, Qingming Festival, Dragon Boat Festival, Labor Day, National Day, Mid-Autumn Festival, Christmas] etc.

[0181] When such terms are detected in the user input, it is considered that the user has the intention to query the schedule or to-do items, and recommended questions for querying the schedule will be recommended. For example: When the user enters the schedule to-do type word "arrange" on August 24th, confirm the dates with schedule / to-do items in the system. If there are schedule to-do items on August 25th, the recommended question will be "What are the arrangements on August 25th / What are the arrangements tomorrow"; when the user enters "next week", confirm whether there is a schedule in the system for next week. If so, the recommended question will be "What are the arrangements for next week". If not, assemble and recommend questions based on the dates when the user actually has schedule to-do items.

[0182] For name terms (such as application names or brand product name generalizations Term), users may ask questions related to applications, brands, merchants, and products. These entities may have some aliases or common names. In the original content, these entities are in standard appellations, but users can enter aliases for retrieval. Such inputs need to be transformed. The example list is as follows:

[0183] {"Bilibili": ["B站", "bilibili"], "Coca-Cola": ["Happy Water for Otaku"], "Porsche": ["Piece-of-junk Used Car"], "Yves Saint Laurent": ["YSL", "Yangshu Lin"], "KFC": ["KFC", "Kaifeng Cuisine"]} etc.

[0184] When a generalized term is detected in the user input, convert it to the corresponding standard appellation and perform a search in the database.

[0185] For letter terms (such as pinyin terms Term), the example list is as follows:

[0186] {"didi": "Didi", "bili": "Bilibili", "shenfenz": "ID Card", "weilai": "NIO", "taobao": "Taobao", "beiwanglu": "Memo", "weibo": "Weibo", "meit": "Meituan", "meitun": "Meituan", "meituan": "Meituan", "meityan": "Meituan", "meitjan": "Meituan", "xiaomisu7": "Xiaomi su7", "hanglv": "Zongheng Airlines", "shunfeng": "SF Express", "pindd": "Pinduoduo", "zhifub": "Alipay", "douyin": "Douyin", "hema": "Hema", "gaode": "Gaode Map"} etc.

[0187] Similar to name terms, when a pinyin term is detected in the user input, convert it to the corresponding standard appellation and perform a search in the database.

[0188] For colloquial terms (such as the colloquial writing of "Term"), when the content involves the amount of an invoice in words (e.g., 9 yuan 87 jiao), 9.87 yuan), temperature symbols (e.g., 32.5℃) 32.5 degrees Celsius), 250g Half a jin (250g), 89.36m² 89.36 square meters, 80 km / h Discrepancies between colloquial writing (e.g., 80-yard) and formal text can be detected using regularization and converted to standard writing.

[0189] Furthermore, regarding electronic devices, based on the first and second problem materials, multiple retrieval methods are used to separately retrieve the clustered input content, resulting in multiple retrieval results. These results are then fused and sorted to obtain a list of problem materials. The search and matching process can include: based on user input, if the aforementioned terms are matched, the respective processing logic can be entered, such as directly assembling questions or filtering terms, and continuing matching based on subsequent input. Optionally, taking text input as an example, such as... Figure 12 As shown, embodiments of this application can employ multiple retrieval methods (e.g., like search, keyword search, and semantic search) to obtain results (Recall List1, Recall List2, Recall List3), then take the Top K from the multiple retrieval results and perform a fusion sort (e.g., RRF fusion sort) to finally obtain a list of potential recommended problem materials. Specifically, the Top K can be taken for each retrieval method, then sorted, and the final result can be the Top K again. For example, the retrieval method may also include a Trie.

[0190] Furthermore, if the user inputs a semantically meaningless string, and neither like nor keyword search can retrieve the results, and the similarity of the semantic search is also very low (a manual threshold T can be used), then matching can stop when the input exceeds M characters. During the input process, several questions can be randomly selected from the most recently added batch of corresponding question materials (or recommended questions) as recommended questions.

[0191] It should be noted that, Figure 12 The three retrieval methods shown are merely illustrative examples. Other retrieval methods may also be used in this application, as long as they can cover word, string, and semantic retrieval. In practical applications, other retrieval algorithms may be modified or added as needed.

[0192] In some embodiments, the method may further include: when any question material in the question material list belongs to the second question material, if the length of the question statement in any question material is less than the length threshold, then assembling the question statements in any question material to obtain question statements that meet the semantic fluency requirements.

[0193] Specifically, electronic devices can assemble question statements, i.e., question assembly. Furthermore, they can perform question fluency detection. Based on user input, they may match a first question statement, i.e., a sentence-like statement, such as an internet-based sentence-like question, or a second statement, i.e., key information in phrase or short sentence form, such as a question in phrase or short sentence form based on local content.

[0194] However, in some cases, the length of the question statements in the question material is less than the length threshold, such as whether the key information is less than 4 characters. It's understandable that the length threshold can be adjusted, for example, to 5 characters. For instance, the minimum length threshold is 4 characters; since too few characters reduce semantic information, reducing it is not recommended. When the key information is short, directly using this information as the recommendation question results in insufficient information and an unnatural tone. To address this, this application proposes assembling question statements from any question material to obtain question statements that meet semantic fluency requirements. For example, question assembly can be used to transform them into recommendation questions that are more in line with everyday expression.

[0195] In one embodiment, assembling the question statements in any question material to obtain question statements that meet the semantic fluency requirements includes: obtaining the data source type name corresponding to any question material; comparing the last text of the question statement in any question material with the data source type name to obtain a comparison result; and selecting a target template to expand and rewrite the question statement in any question material based on the comparison result.

[0196] Specifically, the text at the end of a question statement can refer to the text at the end of the question statement. Comparing the text at the end of a question statement in any given question material with the data source type name can be understood as determining whether the text at the end of the key information is the data source type text. For example, if the key information of a document is "application document," directly rewriting it would result in "document of the application document," which is semantically awkward. In this case, the text at the end is the same as the content type (document), so it can be changed to "materials of the application document."

[0197] Furthermore, the last character of the question statement can be two characters. In this application embodiment, the data source type can include: document, note, schedule, to-do, SMS, and email, and the data source type names are respectively document, note, schedule, to-do, SMS, and email. For example, as shown... Figure 13As shown, we can first determine whether the key information is less than 4 characters. If it is more than 4 characters, it can be directly used as a recommendation question. If it is less than 4 characters, we can obtain the corresponding data source type (document, notes, text message, calendar, to-do, email). Then we can check whether the last text of the key information is the same as the text of the data source type. We can use different question assembly templates (such as Type I module and Type II module) to expand and rewrite the key information to obtain a natural and fluent recommendation question.

[0198] Regarding the question recommendation in the embodiments of this application, in one embodiment, the method may further include: determining the order of multiple recommended questions based on user personalized information; wherein, the user personalized information includes the distance between the generation time of the recommended question and the display time of the question-and-answer interface, the semantic similarity between the recommended question and the input content, and one or more of the following: question type preference and full user data.

[0199] Specifically, the electronic device can determine the order of multiple recommended questions based on the user's personalized information; the user's personalized information in this application may include the distance between the generation time of the recommended question and the display time of the question-and-answer interface, the semantic similarity between the recommended question and the input content, as well as one or more of the following: question type preference and full user data.

[0200] like Figure 14 As shown, if N > 1 potential recommendation questions are matched based on user input, the recommendation order needs to be considered when recommending them to the user. This can be based on a comprehensive ranking of four factors: (i) Time factor: This includes the distance between the generation time of the recommendation question and the display time of the question-and-answer interface. For example, the closer the content is added to the database to the user's search time, the higher the score; (ii) User personalized preferences: Based on the data types of recommendation questions clicked historically by the user (e.g., saved online content, documents, notes, calendars, to-do lists, SMS messages, emails), questions with higher click-through rates receive higher scores; (iii) Semantic relevance (e.g., semantic similarity): The higher the cosine similarity between the user's input content and the matched question vectors, the higher the score; (iv) Full statistics: This can be based on the full user data, similar to (ii), where the statistical object is the distribution of all historical user data. Electronic devices can weight the scores of these four factors to obtain the final ranking of multiple recommendation questions. The weights can be adjusted according to actual needs.

[0201] In one embodiment, the question-and-answer interface includes a search bar; the input content includes an input string determined based on the currently entered characters in the search bar. Specifically, the question-and-answer interface displayed by the electronic device includes a search bar, and the input content may include an input string determined based on the currently entered characters in the search bar.

[0202] This application can implement problem recommendation during the user's search process. Traditional question-and-answer systems mainly give relevant problem recommendations after answering. This application proposes to make recommendations in real time during the user's input process. Whenever the user inputs a new character, the most suitable recommended questions will be adjusted in real time. The user can click on the recommended questions without having to complete the input of the question, thus ending the input in advance.

[0203] In one embodiment, the number of characters in the input string is greater than the number threshold. Specifically, the number threshold can be 2 characters. It can be understood that since the semantics of 1 Chinese character is not sufficient, the scope of making in-search problem recommendations proposed in the embodiments of this application starts from the input of 2 characters. The number of characters in the input string being greater than the number threshold can mean starting from 2 Chinese characters. The semantics of a single Chinese character is insufficient, and direct matching is likely to result in a large difference from the user's intention. For example, when the user inputs "花" (flower), their intention is to search for flowers, but the electronic device matches "花销" (expenses) and gives a recommendation, reducing the user experience. In addition, the input string can refer to Chinese characters. The embodiments of this application are not fully applicable to problem recommendations for word-form languages. The reason is similar to that of 1 Chinese character. 2 letters are likely to match many words. It can be understood that the above content can be used as an extended solution.

[0204] The following is illustrated with a specific example. This application proposes a personalized search problem completion and recommendation solution for end-side content question and answer. The example effect of this application is as Figure 15 shown. The interfaces of Examples ① to ④ may include an input area 802 (such as a search bar) and a display area 804. The input area 802 can be used to obtain the user's input content (for example, input question Query), and the display area 804 can be used to display the recommended questions.

[0205] Furthermore, this application uses AI technologies such as AIGC (Artificial Intelligence Content Generation) to analyze and understand the user's favorite content (such as Internet favorite content) and local content (i.e., local content), prepares several relevant questions offline as potential questions, stores the (Query, Content) pairs in the database. When the user conducts a search, questions are matched during the input process, and the matching list is sorted considering multiple factors and recommended to the user. This solves the problem that users have a vague memory of personal knowledge, content, and information and don't know how to search. At the same time, it can also reduce the user's input time, improve efficiency, and enhance the retrieval experience. As Figure 16As shown, this application may include question preparation and search-based question recommendation. Question preparation includes analyzing and processing various types of local user data to obtain potential recommended question materials, which are then stored in a database. The client-side de-identification module can be used for spam filtering. When a user performs a search, search-based question recommendation refers to matching question materials in the database in real-time based on the user's input, and combining this with a series of recommendation logics to provide real-time question recommendations to the user, thus addressing the pain points of users "not remembering, not knowing how to ask, and finding typing troublesome."

[0206] This application proposes a personalized search completion and recommendation system for client-side content question answering, which differs significantly from traditional search engines. In this system, recommended questions are generated based on the user's personal content, while traditional solutions rely on hot keyword statistics for internet search questions, failing to cover most of the user's local content and thus unable to offer user-relevant questions. This system prepares questions based on user content, ensuring that all recommended questions are relevant to the user's personalized information, provided the input matches the user's input. Furthermore, this system can recommend questions during the user's search process, whereas traditional question answering systems primarily provide related question recommendations after the user has completed their answer. This system provides recommendations in real-time during user input; each new character entered by the user adjusts the most suitable recommended question accordingly. Users can click on a recommended question without fully entering the question to end their input early. In addition, this system proposes a recommendation ranking based on user personalized preferences, whereas traditional solutions rank recommended questions based on statistical data. This system combines multiple dimensions of user personalized information to provide a ranking of recommended questions, making it more aligned with user preferences.

[0207] In one exemplary embodiment, such as Figure 17 As shown, a question-and-answer method is provided. Taking the application of this method to an electronic device as an example, the method includes the following steps 602 to 604.

[0208] Step 602: Obtain the user's input content on the Q&A interface.

[0209] Specifically, electronic devices can acquire the user's input on the question-and-answer interface.

[0210] Step 604: When the input content is a recommendation question, the corresponding answer is generated based on the content corresponding to the recommendation question using a large language model; wherein, the recommendation question is the recommendation question in the question recommendation method described above.

[0211] Specifically, when the input content is a recommendation question, the electronic device can generate a corresponding answer based on the content corresponding to the recommendation question using a large language model; where the recommendation question is the recommendation question in the question recommendation method mentioned above.

[0212] Since the recommended questions are based on content, this content can be used as input to answer the questions. Based on the question-and-answer method in this application embodiment, the step of searching in the database can be omitted, thereby improving the efficiency of question-and-answering.

[0213] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0214] Based on the same inventive concept, this application also provides a problem recommendation apparatus for implementing the problem recommendation method described above. The solution provided by this apparatus is similar to the solution described in the above method; therefore, the specific limitations in one or more problem recommendation apparatus embodiments provided below can be found in the limitations of the problem recommendation method described above, and will not be repeated here.

[0215] In one exemplary embodiment, such as Figure 18 As shown, a problem recommendation device is provided, applied to an electronic device, the device comprising:

[0216] The input acquisition unit 801 is used to display the question and answer interface and acquire the input content currently displayed on the question and answer interface;

[0217] The question display unit 802 is used to display recommended questions on the question-and-answer interface based on the input content; wherein the recommended questions are generated based on user data; the user data includes at least the user's local content on the electronic device.

[0218] In one embodiment, the question display unit 802 is used to display a question and answer interface in response to a triggering operation of the question and answer entry; the question and answer entry is used to provide question and answer functions after being triggered.

[0219] In one embodiment, the question-and-answer entry point includes a button control; the triggering operation includes one or more of the following: a touch operation on the button control, a voice input operation, a touch operation on the fingerprint sensor, and a key operation.

[0220] In one embodiment, the user data also includes online bookmarks, and the device further includes:

[0221] The first question material generation unit is used to generate questions based on online collections to obtain the first question material.

[0222] The second question material generation unit is used to generate questions based on local content to obtain second question materials;

[0223] The recommended question generation unit is used to generate recommended questions based on at least one of the first question material and the second question material.

[0224] In one embodiment, the first question material includes multiple first pairs, each pair including a first question statement and corresponding online collection content; the first question material generation unit is used to analyze and process the online collection content through a large language model to obtain the first question statement.

[0225] In one embodiment, the first question statement includes a statement in sentence form.

[0226] In one embodiment, the second question material includes multiple second pairs, each pair including a second question statement and corresponding local content; the second question material generation unit is used to obtain permission information corresponding to the local content; the permission information is used to indicate the cloud's access permissions for the local content; if it is determined from the permission information that the cloud has permission to access the local content, the local content is processed through a large language model to obtain the second question statement; if it is determined from the permission information that the cloud does not have permission to access the local content, the local content is processed through an edge model to obtain the second question statement.

[0227] In one embodiment, the second question statement includes key information in the form of a phrase or a short sentence.

[0228] In one embodiment, the problem display unit 802 includes:

[0229] The random display unit is used to display randomly selected recommended questions when the input content does not meet the question-and-answer criteria;

[0230] The recommendation display unit is used to determine the recommended questions to be displayed on the question-and-answer interface based on the input content when the input content meets the question-and-answer conditions.

[0231] In one embodiment, the randomly selected recommendation question is related to the time when the recommendation question was generated.

[0232] In one embodiment, the randomly selected recommended questions include those generated at times closest to the display time of the question-and-answer interface; and / or, the randomly selected recommended questions include those generated at times furthest from the display time of the question-and-answer interface.

[0233] In one embodiment, the question-answering conditions include determining that the input content is meaningful and determining that the input content has a question-answering intent.

[0234] In one embodiment, the recommended display unit includes:

[0235] Clustering unit is used to perform speech clustering on the input content to obtain the clustered input content;

[0236] The retrieval unit is used to retrieve the clustered input content using multiple retrieval methods based on the first question material and the second question material, resulting in multiple retrieval results. The first question material is generated from online collections; the second question material is generated from local content.

[0237] The fusion and sorting unit is used to merge and sort multiple search results to obtain a list of question materials; based on the question materials in the list of question materials, recommended questions are generated.

[0238] In one embodiment, a clustering unit is configured to cluster terms in response to detecting terms in the input content to obtain clustered input content; wherein the terms include one or more of prepositional terms, time terms, name terms, alphabetic terms, and colloquial terms; the clustering process includes filtering prepositional terms, a problem material matching strategy for time terms, and standardization processing for name terms, alphabetic terms, and colloquial terms.

[0239] In one embodiment, the apparatus further includes a statement assembly unit, configured to assemble the statement in any question material in the question material list into a statement that satisfies the semantic fluency requirement when any question material in the question material list belongs to the second question material and the length of the statement in any question material is less than a length threshold.

[0240] In one embodiment, the statement assembly unit is used to obtain the data source type name corresponding to any question material; compare the last text of the question statement in any question material with the data source type name to obtain a comparison result; and select a target template to expand and rewrite the question statement in any question material based on the comparison result.

[0241] In one embodiment, the apparatus further includes: a sorting unit, configured to determine the order of multiple recommended questions based on user personalization information; wherein the user personalization information includes one or more of the following: the distance between the generation time of the recommended questions and the display time of the question-and-answer interface, the semantic similarity between the recommended questions and the input content, question type preferences, and full user data.

[0242] In one embodiment, the question-and-answer interface includes a search bar; the input content includes an input string determined based on the characters currently entered in the search bar.

[0243] In one embodiment, the number of characters in the input string is greater than a number threshold.

[0244] The modules in the aforementioned recommended device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of the electronic device in hardware form or independent of it, or stored in the memory of the electronic device in software form, so that the processor can call and execute the operations corresponding to each module.

[0245] Based on the same inventive concept, this application also provides a question-answering device for implementing the question-answering method described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more question-answering device embodiments provided below can be found in the limitations of the question-answering method above, and will not be repeated here.

[0246] In one exemplary embodiment, such as Figure 19 As shown, a question-and-answer device is provided for use in electronic devices. The device includes:

[0247] Content acquisition unit 901 is used to acquire the user's input content on the question-and-answer interface;

[0248] The answer unit 902 is used to generate a corresponding answer based on the content corresponding to the recommendation question when the input content is a recommendation question; wherein, the recommendation question is the recommendation question in the question recommendation method mentioned above.

[0249] Each module in the aforementioned question-and-answer device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of the electronic device in hardware form or independent of it, or stored in the memory of the electronic device in software form, so that the processor can call and execute the operations corresponding to each module.

[0250] In one embodiment, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0251] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0252] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0253] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0254] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0255] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0256] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A question recommendation method characterized by comprising: The method is applied to an electronic device, and the method comprises: displaying a question and answer interface, and obtaining input content currently input in the question and answer interface; based on the input content, displaying a recommended question in the question and answer interface; wherein the recommended question is generated according to user data; and the user data at least comprises local content of a user in the electronic device.

2. The method of claim 1, wherein, displaying a question and answer interface, comprising: in response to a triggering operation on a question and answer entry, displaying the question and answer interface; the question and answer entry is used to provide a question and answer function after being triggered.

3. The method of claim 2, wherein, the question and answer entry comprises a button control; the triggering operation comprises one or more of a touch operation, a voice input operation, a touch operation of a fingerprint sensor, and a key operation on the button control.

4. The method of claim 1, wherein, the user data further comprises network collection content, and the method further comprises: generating a first question material according to the network collection content; generating a second question material according to the local content; generating the recommended question based on at least one of the first question material and the second question material.

5. The method of claim 4, wherein, the first question material comprises a plurality of first combination pairs, and each first combination pair comprises a first question sentence and corresponding network collection content; generating a first question material according to the network collection content comprises: processing the network collection content by a large language model to obtain the first question sentence.

6. The method of claim 5, wherein, the first question sentence comprises a sentence form.

7. The method of claim 4, wherein, the second question material comprises a plurality of second combination pairs, and each second combination pair comprises a second question sentence and corresponding local content; generating a second question material according to the local content comprises: obtaining permission information corresponding to the local content; the permission information is used to indicate an access permission of the cloud to the local content; if it is determined according to the permission information that the cloud has the access permission to the local content, processing the local content by a large language model to obtain the second question sentence; if it is determined according to the permission information that the cloud does not have the access permission to the local content, processing the local content by an end-side model to obtain the second question sentence.

8. The method of claim 7, wherein, the second question sentence comprises key information in the form of a phrase or a short sentence.

9. The method according to any one of claims 1 to 8, characterized in that, based on the input content, displaying a recommended question in the question and answer interface, comprising: when the input content does not meet a question and answer condition, displaying a randomly selected recommended question; when the input content meets the question and answer condition, determining a recommended question displayed in the question and answer interface according to the input content.

10. The method of claim 9, wherein, the randomly selected recommended question is related to the generation time of the recommended question.

11. The method of claim 10, wherein, the randomly selected recommended question comprises a recommended question with the closest generation time to the display time of the question and answer interface; and / or the randomly selected recommended question comprises a recommended question with the farthest generation time to the display time of the question and answer interface.

12. The method of claim 9, wherein, the question and answer condition comprises at least one of determining that the input content is meaningful and determining that the input content has a question and answer intention.

13. The method of claim 9, wherein, the user data further comprises network collection content; and determining a recommended question displayed in the question and answer interface according to the input content comprises: performing a technique cluster on the input content to obtain clustered input content; based on the first question material and the second question material, using multiple retrieval methods to respectively retrieve the clustered input content to obtain multiple retrieval results; the first question material is obtained by generating questions according to the network collection content; the second question material is obtained by generating questions according to the local content; performing fusion sorting on the multiple retrieval results to obtain a question material list; generating the recommended question according to the question material in the question material list.

14. The method of claim 13, wherein, performing a technique cluster on the input content to obtain clustered input content, comprising: in response to detecting a term in the input content, performing a clustering process on the term to obtain the clustered input content; wherein the term includes one or more of a prefix term, a time term, a name term, an alphabet term, and a spoken language term; the clustering process includes filtering the prefix term, a question material matching strategy for the time term, and a standardization process for the name term, the alphabet term, and the spoken language term.

15. The method of claim 13, wherein, The method further comprises: when any question material in the question material list belongs to the second question material, if the length of the question sentence in the any question material is less than the length threshold, assembling the question sentence in the any question material to obtain a question sentence that meets the semantic fluency requirement.

16. The method of claim 15, wherein, Assembling the question sentence in the any question material to obtain a question sentence that meets the semantic fluency requirement, comprising: obtaining the data source type name corresponding to the any question material; comparing the last character of the question sentence in the any question material with the data source type name to obtain a comparison result, and selecting a target template according to the comparison result to expand and rewrite the question sentence in the any question material.

17. The method of claim 13, wherein, The method further comprises: determining the ranking of the multiple recommended questions according to user personalized information; wherein the user personalized information includes one or more of the distance between the generation time of the recommended question and the display time of the question and answer interface, the semantic similarity between the recommended question and the input content, and the question type preference, and the full user data.

18. The method of claim 9, wherein, The question and answer interface includes a search bar; the input content includes an input string determined according to the characters currently input in the search bar.

19. The method of claim 18, wherein, The number of characters in the input string is greater than a quantity threshold.

20. A question and answer method, characterized by, Applied to an electronic device, the method comprises: obtaining input content of a user in a question and answer interface; when the input content is a recommended question, generating a corresponding answer through a large language model based on the content corresponding to the recommended question; wherein the recommended question is the recommended question in the question recommendation method of any one of claims 1 to 19.

21. A question recommendation apparatus characterized by comprising: Applied to an electronic device, the device comprises: an input acquisition unit configured to display a question and answer interface and acquire input content currently in the question and answer interface; A question display unit is configured to display a recommended question based on the input content in the Q&A interface; wherein the recommended question is generated according to user data; and the user data at least includes local content of the user on the electronic device.

22. A question and answer apparatus, characterized by The application is applied to an electronic device, and the device comprises: A content acquisition unit is configured to acquire input content of a user in a Q&A interface; An answering unit is configured to, when the input content is a recommended question, generate a corresponding answer through a large language model based on content corresponding to the recommended question; wherein the recommended question is a recommended question in the question recommendation method of any one of claims 1 to 19.

23. An electronic device, comprising a memory and a processor, the memory storing a computer program, characterized in that, The processor executes the computer program to implement the steps of the method of any one of claims 1 to 20.

24. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 20.

25. A computer program product comprising a computer program, characterised in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 20. The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 20.

Images