Topic recommendation method and related device

Abstract

The invention discloses a question recommendation method and a related device, and relates to the technical field of intelligent education, and the question recommendation method comprises the steps: obtaining a specified question; converting the specified question into a text vector to obtain a text vector corresponding to the specified question; performing similarity calculation on the text vector corresponding to the specified question and the text vector corresponding to each question in the question database to obtain questions similar to the specified question, and obtaining a candidate question set; using a large language model to perform semantic understanding on the specified question and each question in the candidate question set, and generating scores corresponding to each question in the candidate question set, the score corresponding to any question in the candidate question set reflecting the similarity and semantic correlation between the question and the specified question; according to the scores corresponding to all the questions in the candidate question set, recommended questions are determined and recommended. The question recommendation method disclosed by the invention has relatively high recommendation accuracy and recommendation efficiency, and relatively good user experience.

Description

Technical Field

[0001] This application relates to the field of intelligent education technology, and particularly relates to a question recommendation method and related devices. Background Art

[0002] By doing a large number of questions, knowledge can be continuously consolidated and supplemented, so that the knowledge can be mastered more firmly. It can be understood that behind each question is an examination of a certain knowledge point. By doing more questions, one can naturally master the knowledge more firmly. At the same time, during the process of doing questions, one can discover one's own knowledge blind spots and conduct targeted reviews in a timely manner. Therefore, during the learning process, students usually need to do a large number of questions, commonly known as brushing questions.

[0003] With the development of computer technology, in order to facilitate students to learn more effectively, methods and devices for question recommendation have emerged as the times require. The current question recommendation methods are mainly question recommendation methods based on keyword matching, that is, similar questions are screened from the question database according to the keywords in the question for recommendation. Although the question recommendation method based on keyword matching can achieve question recommendation, the recommendation accuracy is not high, which in turn leads to a poor user experience. Summary of the Invention

[0004] In view of this, this application provides a question recommendation method and related devices to solve the problem that the recommendation accuracy of the existing question recommendation methods is not high, which in turn leads to a poor user experience. The technical solutions are as follows:

[0005] The first aspect of this application provides a question recommendation method, including:

[0006] Obtain a specified question;

[0007] Convert the specified question into a text vector to obtain the text vector corresponding to the specified question;

[0008] Obtain questions similar to the specified question by calculating the similarity between the text vector corresponding to the specified question and the text vector corresponding to each question in the question database, and obtain a candidate question set;

[0009] Utilize a large language model to perform semantic understanding on the specified question and each question in the candidate question set, and generate scores corresponding to each question in the candidate question set, where the score corresponding to any question in the candidate question set reflects the similarity degree and semantic relevance between this question and the specified question;

[0010] Determine and recommend recommended questions according to the scores corresponding to each question in the candidate question set.

[0011] In a possible implementation, obtaining the questions similar to the specified question by calculating the similarity between the text vector corresponding to the specified question and the text vectors corresponding to each question in the question database includes:

[0012] Calculating the similarity between the text vector corresponding to the specified question and the text vectors corresponding to each question in the question database to obtain the similarities corresponding to each question in the question database;

[0013] Sorting the similarities corresponding to each question in the question database in descending order;

[0014] Obtaining the questions corresponding to the top N similarities as the questions similar to the specified question, where N is an integer greater than 1.

[0015] In a possible implementation, using the large language model to perform semantic understanding on the specified question and each question in the candidate question set to generate the scores corresponding to each question in the candidate question set includes:

[0016] Using the large language model to analyze the specified question and each question in the candidate question set from multiple dimensions to generate the scores corresponding to each question in the candidate question set;

[0017] Among them, the multiple dimensions include some or all of the following dimensions: the knowledge points involved in the question, the logical relationship of the question, and the problem-solving idea of the question.

[0018] In a possible implementation, using the large language model to analyze the specified question and each question in the candidate question set from multiple dimensions to generate the scores corresponding to each question in the candidate question set includes:

[0019] Obtaining a prompt format template, where the prompt format template includes a specified question information slot and a candidate question information slot, and the prompt format template is used to instruct the large language model to analyze the question in the specified question information slot and each question in the candidate question information slot from multiple dimensions, and give a score reflecting the similarity and semantic relevance between each question in the candidate question information slot and the question in the specified question information slot;

[0020] Filling the specified question into the specified question information slot of the prompt format template, and filling the questions in the candidate question set into the candidate question information slot of the prompt format template to obtain a prompt instruction prompt;

[0021] Input the prompt instruction into the large language model to obtain the scores corresponding to each question in the candidate question set output by the large language model.

[0022] In a possible implementation manner, the determining and recommending the recommended questions according to the scores corresponding to each question in the candidate question set includes:

[0023] Obtain the highest score from the scores corresponding to each question in the candidate question set;

[0024] If the highest score is greater than or equal to a preset score threshold, determine the question corresponding to the highest score as the recommended question and recommend it.

[0025] In a possible implementation manner, the converting the specified question into a text vector to obtain the text vector corresponding to the specified question includes:

[0026] Clean the specified question and perform word segmentation on the cleaned question;

[0027] Use a pre-trained word vector model to convert each word obtained by word segmentation into a word vector;

[0028] The text vectors of each word obtained by word segmentation form the text vector corresponding to the specified question.

[0029] The second aspect of this application provides a question recommendation device, including: a question acquisition module, a question vectorization module, a candidate question set acquisition module, a question scoring module, a recommended question determination module, and a question recommendation module;

[0030] The question acquisition module is used to acquire a specified question;

[0031] The question vectorization module is used to convert the specified question into a text vector to obtain the text vector corresponding to the specified question;

[0032] The candidate question set acquisition module is used to obtain the questions similar to the specified question by calculating the similarity between the text vector corresponding to the specified question and the text vectors corresponding to each question in the question database, and obtain a candidate question set;

[0033] The question scoring module is used to use the large language model to perform semantic understanding on the specified question and each question in the candidate question set, and generate the scores corresponding to each question in the candidate question set, where the score corresponding to any question in the candidate question set reflects the similarity degree and semantic relevance between this question and the specified question;

[0034] The recommended question determination module is used to determine the recommended question according to the scores corresponding to each question in the candidate question set;

[0035] The title recommendation module is used to recommend the recommended titles.

[0036] A third aspect of the present application provides an electronic device, including at least one processor and a memory connected to the processor, wherein:

[0037] The memory is used to store a computer program;

[0038] The processor is used to execute the computer program so that the electronic device can implement the steps of any one of the above-mentioned title recommendation methods.

[0039] A fourth aspect of the present application provides a computer storage medium, which carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement the steps of any one of the above-mentioned title recommendation methods.

[0040] A fifth aspect of the present application provides a computer program product, including computer-readable instructions. When the computer-readable instructions run on an electronic device, the electronic device can implement the steps of any one of the above-mentioned title recommendation methods.

[0041] By means of the above technical solution, for the title recommendation method provided by the present application, after obtaining a specified title, first convert the specified title into a text vector to obtain the text vector corresponding to the specified title, and then calculate the similarity between the text vector corresponding to the specified title and the text vector corresponding to each title in the title database to obtain the titles similar to the specified title, so as to obtain a candidate title set. After obtaining the candidate title set, in order to obtain a higher recommendation accuracy rate, the embodiments of the present application further utilize the semantic understanding ability of the large language model to perform semantic understanding on the specified title and each title in the candidate title set, generate the scores corresponding to each title in the candidate title set respectively, and finally determine and recommend the recommended titles according to the scores corresponding to each title in the candidate title set respectively. The title recommendation method provided by the present application combines the title screening method based on text vectors with the semantic understanding ability of the large language model, can achieve accurate recommendation. In addition, the title recommendation method provided by the present application can effectively filter out irrelevant titles through multi-stage screening, and has a high recommendation efficiency. In summary, the title recommendation method provided by the present application has a high recommendation accuracy rate and a high recommendation efficiency, and the user experience is good. Description of the Drawings

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the provided drawings.

[0043] Figure 1 Schematic diagram of a system architecture related to the present application;

[0044] Figure 2 Schematic diagram of a hardware structure of a terminal provided by an embodiment of the present application;

[0045] Figure 3 Schematic diagram of a hardware structure of a server provided by an embodiment of the present application;

[0046] Figure 4 Schematic diagram of a process flow of a question recommendation method provided by an embodiment of the present application;

[0047] Figure 5 Schematic diagram of a structure of a question recommendation device provided by an embodiment of the present application. Detailed implementation manners

[0048] The following describes the embodiments of the present application in combination with the drawings in the embodiments of the present application. The terms used in the implementation part of the present application are only used to explain the specific embodiments of the present application, rather than intended to limit the present application.

[0049] The following describes the embodiments of the present application in combination with the drawings. Those of ordinary skill in the art know that with the development of technology and the emergence of new scenarios, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems.

[0050] The terms "first", "second", etc. in the specification, claims and drawings of the present application are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that such terms can be interchanged under appropriate circumstances, which is only a way of distinguishing when describing objects with the same attributes in the embodiments of the present application. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion, so that a process, method, system, product or device including a series of units does not necessarily have to be limited to those units, but may include other units not clearly listed or inherent to these process, method, product or device.

[0051] In a possible implementation manner, as Figure 1As shown in the figure, the system architecture involved in this application may include a terminal 101 and a server 102. The terminal 101 can interact with the server 102 through a network (wired network or wireless network). Among them, the server 102 may include one or more servers ( Figure 1 In the following, an example of including one server is used for illustration). The terminal can obtain a specified topic, transmit the specified topic to the server through the network, and the server determines a recommended topic from the topic database according to the specified topic and sends the recommended topic to the terminal.

[0052] In another possible implementation, the system architecture involved in this application may include a terminal. The terminal has strong data processing capabilities. The terminal can obtain a specified topic, determine a recommended topic from the topic database according to the specified topic, and output it.

[0053] Next, the product form of the above terminal will be described.

[0054] The above terminal can be a mobile phone, a tablet computer, a wearable device, a vehicle-mounted device, a robot, an augmented reality (AR) / virtual reality (VR) device, a laptop computer, an ultra-mobile personal computer (UMPC), a netbook, a personal digital assistant (PDA), etc. The embodiments of this application do not impose any restrictions on this.

[0055] Figure 2 Shows a schematic diagram of an optional hardware structure of the terminal.

[0056] Refer to Figure 2 As shown in the figure, the terminal may include a radio frequency unit 210, a memory 220, an input unit 230, a display unit 240, a camera 250 (optional), an audio circuit 260 (optional), a speaker 261 (optional), a microphone 262 (optional), a headphone jack 263 (optional), a processor 270, an external interface 280, a power supply 290, and other components. Those skilled in the art can understand that Figure 2 This is just an example of the terminal and does not constitute a limitation on the terminal. It may include more or fewer components than shown in the figure, or combine some components, or different components.

[0057] The input unit 230 can be used to receive input digital or character information and generate key signal inputs related to the user settings and function controls of the terminal. Specifically, the input unit 230 may include a touch screen 231 (optional) and / or other input devices 232. The touch screen 231 can collect touch operations of the user thereon or nearby (such as operations of the user using any suitable object such as a finger, a joint, a stylus, etc. on or near the touch screen), and drive corresponding connection devices according to a preset program. The touch screen can detect the touch actions of the user on the touch screen, convert the touch actions into touch signals and send them to the processor 270, and can receive and execute the commands sent by the processor 270; the touch signals at least include contact coordinate information. The touch screen 231 can provide an input interface and an output interface between the terminal and the user. In addition, various types such as resistive, capacitive, infrared, and surface acoustic wave can be used to implement the touch screen. In addition to the touch screen 231, the input unit 230 may further include other input devices. Specifically, the other input devices 232 may include, but are not limited to, one or more of a physical keyboard, function keys (such as volume control keys, power on / off keys, etc.), a trackball, a mouse, a joystick, etc.

[0058] The display unit 240 can be used to display information input by the user or information provided to the user, various menus of the terminal, an interactive interface, file display, and / or the playback of any multimedia file.

[0059] The memory 220 can be used to store instructions and data. The memory 220 mainly includes a storage instruction area and a storage data area. The storage data area can store various data, such as multimedia files, texts, etc.; the storage instruction area can store software units such as an operating system, applications, instructions required for at least one function, etc., or their subsets and extended sets. It may also include a non-volatile random access memory; it provides the processor 270 with management of the hardware, software, and data resources in the computing processing device, supports control software and applications. It is also used for the storage of multimedia files and the storage of running programs and applications.

[0060] The processor 270 is the control center of the terminal, connecting various parts of the entire terminal using various interfaces and lines. By running or executing instructions stored in the memory 220 and invoking data stored in the memory 220, it performs various functions of the terminal and processes data, thereby exercising overall control over the terminal. Optionally, the processor 270 may include one or more processing units; preferably, the processor 270 may integrate an application processor and a modem processor. Among them, the application processor mainly processes the operating system, user interface, application programs, etc., and the modem processor mainly processes wireless communication. It can be understood that the above-mentioned modem processor may not be integrated into the processor 270. In some embodiments, the processor and the memory may be implemented on a single chip, and in some embodiments, they may also be separately implemented on independent chips. The processor 270 can also be used to generate corresponding operation control signals, send them to corresponding components of the computing and processing device, read and process data in the software, especially read and process the data and programs in the memory 220, so that each functional module therein executes corresponding functions, thereby controlling the corresponding components to act according to the requirements of the instructions.

[0061] Among them, the memory 220 can be used to store software codes related to the question recommendation method, and the processor 270 can execute the software codes in the memory 220 or can also schedule other units (such as the above-mentioned input unit 230 and display unit 240) to implement corresponding functions.

[0062] The radio frequency unit 210 (optional) can be used to receive and transmit information or signals during a call. For example, after receiving the downlink information from the base station, it is given to the processor 270 for processing; in addition, it sends the designed uplink data to the base station. Usually, the radio frequency unit 210 includes, but is not limited to, antennas, at least one amplifier, a transceiver, a coupler, a low noise amplifier (LNA), a duplexer, etc. In addition, the radio frequency unit 210 can also communicate with network devices and other devices through wireless communication. This wireless communication can use any communication standard or protocol, including but not limited to the Global System of Mobile communication (GSM), General Packet Radio Service (GPRS), Code Division Multiple Access (CDMA), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE), email, Short Messaging Service (SMS), etc.

[0063] Among them, in the embodiments of the present application, the radio frequency unit 210 can send data to other devices and can also receive data sent by other devices. It should be understood that the radio frequency unit 210 is optional and can be replaced by other communication interfaces, such as a network interface.

[0064] The terminal further includes a power supply 290 (such as a battery) for supplying power to each component. Preferably, the power supply can be logically connected to the processor 270 through a power management system, so as to realize functions such as management of charging, discharging, and power consumption management through the power management system.

[0065] The terminal further includes an external interface 280. The external interface can be a standard Micro USB interface or a multi-pin connector, and can be used to connect the terminal to other devices for communication and can also be used to connect a charger to charge the terminal.

[0066] Although not shown, the terminal may further include a flash, a wireless fidelity (WiFi) module, a Bluetooth module, sensors with different functions, etc., which will not be elaborated here.

[0067] Next, the product form of the above server will be described.

[0068] Figure 3 A schematic structural diagram of the above server is provided, as Figure 3 shown, the server may include a bus 301, a processor 302, a communication interface 303, and a memory 304. The processor 302, the memory 304, and the communication interface 303 communicate through the bus 301.

[0069] The bus 301 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the sake of representation, Figure 3 only a thick line is used to represent it in the figure, but it does not mean that there is only one bus or one type of bus.

[0070] The processor 302 can be any one or more of a central processing unit (CPU), a graphics processing unit (GPU), a micro processor (MP), or a digital signal processor (DSP), etc.

[0071] The memory 304 may include a volatile memory, such as a random access memory (RAM). The memory 304 may also include a non-volatile memory, such as a read-only memory (ROM), a flash memory, a hard disk drive (HDD), or a solid state drive (SSD).

[0072] The memory 304 may be used to store software codes related to the question recommendation method. The processor 302 may call the software codes stored in the memory 304 or schedule other units to implement corresponding functions.

[0073] The processors in the above-mentioned terminal and server (such as the processor 270 and the processor 302) may be hardware circuits (such as an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a general-purpose processor, a digital signal processor (DSP), a microprocessor, or a microcontroller, etc.), or a combination of these hardware circuits. For example, the processor may be a hardware system with the function of executing instructions, such as a CPU, a DSP, etc., or a hardware system without the function of executing instructions, such as an ASIC, an FPGA, etc., or a combination of the above-mentioned hardware system without the function of executing instructions and the hardware system with the function of executing instructions.

[0074] In view of the fact that the recommendation accuracy of the existing question recommendation method based on keyword matching is not high, and the recommended questions often do not meet the user's needs, the inventor of this case has conducted research. Through continuous research, a question recommendation method with better effects is finally proposed. Next, the question recommendation method provided in this application will be introduced through the following embodiments.

[0075] Please refer to Figure 4 , which shows a schematic flowchart of the question recommendation method provided in the embodiment of this application. The question recommendation method may include:

[0076] Step S401: Obtain a specified question.

[0077] Among them, the specified question may be a question specified by the user, and the specified question may be any type of question.

[0078] This embodiment can obtain the questions input by the user through operations such as searching for questions or taking pictures of questions.

[0079] Step S402: Convert the specified question into a text vector to obtain the text vector corresponding to the specified question.

[0080] After obtaining the specified question, this embodiment converts the specified question into a text vector, that is, vectorizes the specified question to obtain the text vector corresponding to the specified question.

[0081] Step S403: Obtain the questions similar to the specified question by calculating the similarity between the text vector corresponding to the specified question and the text vectors corresponding to each question in the question database, and obtain a candidate question set.

[0082] This embodiment adopts a question screening method based on text vectors to preliminarily screen out the questions similar to the specified question from the question database, and the screened questions form a candidate question set.

[0083] To improve the question recommendation efficiency, each question in the question database can be converted into a text vector in advance to obtain the text vectors corresponding to each question in the question database respectively. The text vectors corresponding to each question in the question database form a question text vector database. In this way, when it is necessary to obtain a candidate question set, the text vectors corresponding to each question in the question database can be directly obtained from the question text vector database. Furthermore, by calculating the similarity between the text vector corresponding to the specified question and the text vectors corresponding to each question in the question database, the questions similar to the specified question can be obtained, thereby obtaining a candidate question set.

[0084] Of course, this embodiment is not limited to this. For example, each question in the question database can also be converted into a text vector when it is necessary to obtain a candidate question set to obtain the text vectors corresponding to each question in the question database respectively. Furthermore, by calculating the similarity between the text vector corresponding to the specified question and the text vectors corresponding to each question in the question database, the questions similar to the specified question can be obtained, thereby obtaining a candidate question set.

[0085] Step S404: Use a large language model to perform semantic understanding on the specified question and each question in the candidate question set, and generate the scores corresponding to each question in the candidate question set.

[0086] Among them, the score corresponding to any question in the candidate question set reflects the similarity degree and semantic relevance between this question and the specified question.

[0087] After obtaining the candidate question set, this embodiment further uses a large language model to perform in-depth semantic understanding on the specified question and each question in the candidate question set, analyze the similarity degree and semantic relevance between the specified question and each question in the candidate question set, and generate the scores corresponding to each question in the candidate question set.

[0088] Step S405: Determine and recommend the recommended questions according to the scores corresponding to each question in the candidate question set.

[0089] After obtaining the scores corresponding to each question in the candidate question set, based on the scores corresponding to each question in the candidate question set, filter and recommend questions from the candidate question set, and then recommend the recommended questions to the user.

[0090] The question recommendation method provided by the embodiment of the present application, after obtaining the specified question, first converts the specified question into a text vector to obtain the text vector corresponding to the specified question, and then calculates the similarity between the text vector corresponding to the specified question and the text vector corresponding to each question in the question database to obtain questions similar to the specified question, so as to obtain a candidate question set. After obtaining the candidate question set, in order to obtain a higher recommendation accuracy, the embodiment of the present application further utilizes the semantic understanding ability of the large language model to perform semantic understanding on the specified question and each question in the candidate question set, generates the scores corresponding to each question in the candidate question set, and finally determines and recommends the recommended questions according to the scores corresponding to each question in the candidate question set. The question recommendation method provided by the embodiment of the present application combines the question screening method based on text vectors with the semantic understanding ability of the large language model, can achieve accurate recommendation. In addition, the question recommendation method provided by the embodiment of the present application can effectively filter out irrelevant questions through multi-stage screening, and has a high recommendation efficiency. In summary, the question recommendation method provided by the embodiment of the present application has a high recommendation accuracy and a high recommendation efficiency, and the user experience is good.

[0091] In another embodiment of the present application, the specific implementation process of "Step S402: Convert the specified question into a text vector to obtain the text vector corresponding to the specified question" in the above embodiment is introduced.

[0092] In a possible implementation manner, the implementation process of converting the specified question into a text vector to obtain the text vector corresponding to the specified question may include:

[0093] Step a1: Clean the specified question.

[0094] In this embodiment, the specified question is cleaned to remove irrelevant information, for example, removing noise characters, special symbols, etc.

[0095] Step a2: Perform word segmentation processing on the cleaned question.

[0096] Existing word segmentation methods can be used to perform word segmentation processing on the cleaned question.

[0097] Step a3: Use a pre-trained word vector model to convert each word obtained through word segmentation into a word vector, and the word vectors of each word obtained through word segmentation form a text vector corresponding to the specified topic.

[0098] After obtaining several words through word segmentation, a pre-trained word vector model can be used to convert each word obtained through word segmentation into a word vector (i.e., a low-dimensional vector representation), that is, to vectorize each word obtained through word segmentation to obtain the word vector of each word.

[0099] Among them, the pre-trained word vector model can be Word2Vec. Of course, this embodiment is not limited thereto. For example, the pre-trained word vector model can also be GloVe. It should be noted that this embodiment does not limit the word vector model to Word2Vec and GloVe, and other models that can convert words into word vectors are applicable to this application.

[0100] After converting each word obtained through word segmentation into a word vector, the word vectors of each word obtained through word segmentation can form a text vector corresponding to the specified topic.

[0101] In another embodiment of the present application, the specific implementation process of "Step S403: Obtain the topics similar to the specified topic by calculating the similarity between the text vector corresponding to the specified topic and the text vector corresponding to each topic in the topic database, and obtain a candidate topic set" in the above embodiment is introduced.

[0102] In a possible implementation manner, the specific implementation process of obtaining the topics similar to the specified topic by calculating the similarity between the text vector corresponding to the specified topic and the text vector corresponding to each topic in the topic database may include:

[0103] Step b1: Calculate the similarity between the text vector corresponding to the specified topic and the text vector corresponding to each topic in the topic database to obtain the similarity corresponding to each topic in the topic database.

[0104] Optionally, for each topic in the topic database, the cosine similarity between the text vector corresponding to the specified topic and the text vector corresponding to the topic can be calculated as the similarity corresponding to the topic.

[0105] Of course, this embodiment is not limited thereto. For example, for each topic in the topic database, the reciprocal of the Euclidean distance between the text vector corresponding to the specified topic and the text vector corresponding to the topic can be calculated as the similarity corresponding to the topic.

[0106] It should be noted that the greater the similarity corresponding to any topic in the topic database, the more similar the topic is to the specified topic.

[0107] Step b2: Sort the similarities corresponding to each question in the question database in descending order.

[0108] After obtaining the similarities corresponding to each question in the question database, the similarities corresponding to each question in the question database can be sorted in descending order.

[0109] Step b3: Obtain the questions corresponding to the top N similarities as the questions similar to the specified question.

[0110] Among them, N is an integer greater than 1, such as 50, 100, etc. It should be noted that the specific value of N is not limited in this embodiment, and the specific value of N can be determined according to the actual application scenario.

[0111] In this embodiment, the questions corresponding to the top N similarities in the sorting result are determined as the questions similar to the specified question, and then a candidate question set is composed of the N questions similar to the specified question.

[0112] It should be noted that this embodiment does not limit the method of obtaining the candidate question set by using steps b1 to b3. For example, the similarity between the text vector corresponding to the specified question and the text vector corresponding to each question in the question database can also be calculated to obtain the similarities corresponding to each question in the question database, and the similarities corresponding to each question in the question database can be sorted in ascending order (i.e., from low to high), and then, the questions corresponding to the last N similarities are obtained as the questions similar to the specified question.

[0113] Through the above process, N questions similar to the specified question can be preliminarily screened from the question database, thus obtaining a candidate question set.

[0114] In another embodiment of the present application, the specific implementation process of "Step S404: Use a large language model to perform semantic understanding on the specified question and each question in the candidate question set, and generate the scores corresponding to each question in the candidate question set" in the above embodiment is introduced.

[0115] In a possible implementation manner, the process of using a large language model to perform semantic understanding on the specified question and each question in the candidate question set and generate the scores corresponding to each question in the candidate question set may include:

[0116] Use a large language model to analyze the specified question and each question in the candidate question set from multiple dimensions, and generate the scores corresponding to each question in the candidate question set.

[0117] Among them, the multiple dimensions include some or all of the following dimensions: the knowledge points involved in the question, the logical relationship of the question, the problem-solving ideas of the question, etc.

[0118] Specifically, the process of using a large language model to analyze a specified topic and each topic in a set of candidate topics from multiple dimensions and generate scores corresponding to each topic in the set of candidate topics may include:

[0119] Step c1: Obtain a prompt format template.

[0120] Among them, the prompt format template includes a specified topic information slot and a candidate topic information slot. The prompt format template is used to instruct the large language model to analyze the topic in the specified topic information slot and each topic in the candidate topic information slot from multiple dimensions, and give a score reflecting the similarity and semantic relevance between each topic in the candidate topic information slot and the topic in the specified topic information slot.

[0121] Step c2: Fill the specified topic into the specified topic information slot of the prompt format template, and fill the topic maps in the set of candidate topics into the candidate topic information slot of the prompt format template to obtain a prompt instruction prompt.

[0122] Step c3: Input the prompt instruction prompt into the large language model to obtain the scores corresponding to each topic in the set of candidate topics output by the large language model.

[0123] Inputting the prompt instruction prompt into the large language model, the large language model conducts in-depth semantic understanding of the specified topic and each topic in the set of candidate topics, including analyzing the knowledge points involved in the topic, analyzing the logical relationship of the topic, analyzing the problem-solving ideas of the topic, etc. For each topic in the set of candidate topics, it outputs a score reflecting the similarity and semantic relevance between this topic and the specified topic.

[0124] Through the above process, the scores corresponding to each topic in the set of candidate topics can be obtained.

[0125] In another embodiment of the present application, the specific implementation process of "Step S405: Determine and recommend recommended topics according to the scores corresponding to each topic in the set of candidate topics" in the above embodiment is introduced.

[0126] In a possible implementation manner, the implementation process of determining and recommending recommended topics according to the scores corresponding to each topic in the set of candidate topics may include:

[0127] Step d1: Obtain the highest score from the scores corresponding to each topic in the set of candidate topics.

[0128] The topic corresponding to the highest score is the topic with the greatest similarity and semantic relevance to the specified topic.

[0129] In a possible implementation, the scores corresponding to each question in the candidate question set can be sorted in descending order (i.e., from high to low), and the score ranked first is the highest score. Of course, the scores corresponding to each question in the candidate question set can also be sorted in ascending order (i.e., from low to high), and the score ranked last is the highest score.

[0130] Step d2: Determine whether the highest score is greater than or equal to a preset score threshold.

[0131] In this application, a score threshold can be preset. After obtaining the highest score, the highest score is compared with the preset score threshold.

[0132] Step d3: If the highest score is greater than or equal to the preset score threshold, determine the question corresponding to the highest score as the recommended question and recommend it.

[0133] When the highest score is greater than or equal to the preset score threshold, determine the question corresponding to the highest score in the candidate question set as the recommended question, and then recommend the determined recommended question to the user.

[0134] In practical applications, there may be a situation where the highest score is less than the preset score threshold. If the highest score is less than the preset score threshold, in one possible implementation, an indication message that there is no suitable question to recommend can be output. In another possible implementation, the recall range can be expanded to re-obtain the candidate question set. For example, obtain the similarity corresponding to each question in the question database (i.e., the similarity between the text vector corresponding to the specified question and the text vector corresponding to each question in the question database), sort the similarities corresponding to each question in the question database in descending order, and obtain the questions corresponding to the top M similarities to form the candidate question set, where M is greater than N. After obtaining the candidate question set, use the large language model to perform semantic understanding on the specified question and each question in the new candidate question set, generate the scores corresponding to each question in the new candidate question set, obtain the highest score from the scores corresponding to each question in the new candidate question set. If the highest score is greater than or equal to the preset score threshold, determine the question corresponding to the highest score as the recommended question and recommend it to the user. If the highest score is less than the preset score threshold, continue to expand the recall range, re-obtain the candidate question set and perform subsequent steps such as generating scores and screening recommended questions until a question that meets the score threshold is found.

[0135] It should be noted that this embodiment does not limit the use of the above method to obtain the recommended question. Other methods can also be used to obtain the recommended question. For example, the score threshold may not be preset, and the question corresponding to the highest score can be directly determined as the recommended question.

[0136] The question recommendation method provided by the embodiments of the present application uses a screening method based on text vectors to initially screen similar questions from a question database to obtain a set of candidate questions, and then uses the semantic understanding ability of a large language model to analyze the similarity and semantic relevance between the questions in the candidate question set and a specified question. Furthermore, the recommended questions are determined and recommended according to the analysis results of the large language model. Compared with the question recommendation method based on keyword matching, the question recommendation method provided by the embodiments of the present application combines the question screening method based on text vectors with the semantic understanding ability of a large language model, can more accurately grasp the semantics and internal logic of questions, thereby realizing accurate recommendation, and can provide more valuable questions for users. In addition, the question recommendation method provided by the embodiments of the present application can effectively filter out irrelevant questions through multi-stage screening of questions, improve the recommendation efficiency, reduce the user waiting time. In addition, the question recommendation method provided by the embodiments of the present application can generate personalized similar question recommendations in real time according to the questions specified by the user, can meet the needs of users in different learning scenarios, and thus can improve the user's learning experience and learning effect.

[0137] The embodiments of the present application also provide a device for executing the question recommendation method provided by the above embodiments. Please refer to Figure 5 , Figure 5 which is a schematic structural diagram of a question recommendation device provided by the embodiments of the present application. The question recommendation device may include: a question acquisition module 501, a question vectorization module 502, a candidate question set acquisition module 503, a question scoring module 504, a recommended question determination module 505, and a question recommendation module 506.

[0138] The question acquisition module 501 is used to acquire a specified question.

[0139] The question vectorization module 502 is used to convert the specified question into a text vector to obtain the text vector corresponding to the specified question.

[0140] The candidate question set acquisition module 503 is used to obtain questions similar to the specified question by calculating the similarity between the text vector corresponding to the specified question and the text vector corresponding to each question in the question database, and obtain a candidate question set.

[0141] The question scoring module 504 is used to use a large language model to perform semantic understanding on the specified question and each question in the candidate question set, and generate scores corresponding to each question in the candidate question set respectively.

[0142] Among them, the score corresponding to any question in the candidate question set reflects the similarity and semantic relevance between the question and the specified question.

[0143] The recommended question determination module 505 is used to determine the recommended questions according to the scores corresponding to each question in the candidate question set;

[0144] The question recommendation module 506 is used to recommend questions.

[0145] In a possible implementation, when the candidate question set acquisition module 503 obtains questions similar to a specified question by calculating the similarity between the text vector corresponding to the specified question and the text vectors corresponding to each question in the question database, it is specifically used for:

[0146] Calculate the similarity between the text vector corresponding to the specified question and the text vectors corresponding to each question in the question database, and obtain the similarities corresponding to each question in the question database;

[0147] Sort the similarities corresponding to each question in the question database in descending order;

[0148] Obtain the top N questions corresponding to the similarities as the questions similar to the specified question, where N is an integer greater than 1.

[0149] In a possible implementation, when the question scoring module 504 uses a large language model to perform semantic understanding on the specified question and each question in the candidate question set and generates the scores corresponding to each question in the candidate question set, it is specifically used for:

[0150] Use a large language model to analyze the specified question and each question in the candidate question set from multiple dimensions, and generate the scores corresponding to each question in the candidate question set;

[0151] Among them, the multiple dimensions include some or all of the following dimensions: the knowledge points involved in the question, the logical relationship of the question, and the problem-solving idea of the question.

[0152] In a possible implementation, when the question scoring module 504 uses a large language model to analyze the specified question and each question in the candidate question set from multiple dimensions and generates the scores corresponding to each question in the candidate question set, it is specifically used for:

[0153] Obtain a prompt format template, where the prompt format template includes a specified question information slot and a candidate question information slot. The prompt format template is used to instruct the large language model to analyze the question in the specified question information slot and each question in the candidate question information slot from multiple dimensions, and give a score reflecting the similarity and semantic relevance between each question in the candidate question information slot and the question in the specified question information slot;

[0154] Fill the specified question into the specified question information slot of the prompt format template, and fill the questions in the candidate question set into the candidate question information slot of the prompt format template to obtain a prompt instruction prompt;

[0155] Input the prompt instruction into the large language model to obtain the scores corresponding to each question in the set of candidate questions output by the large language model.

[0156] In a possible implementation manner, when the question recommendation module 506 determines and recommends the recommended questions according to the scores corresponding to each question in the set of candidate questions, it is specifically used for:

[0157] Obtain the highest score from the scores corresponding to each question in the set of candidate questions;

[0158] If the highest score is greater than or equal to the preset score threshold, determine the question corresponding to the highest score as the recommended question and recommend it.

[0159] In a possible implementation manner, when the question vectorization module 502 converts the specified question into a text vector to obtain the text vector corresponding to the specified question, it is specifically used for:

[0160] Clean the specified question and perform word segmentation on the cleaned question;

[0161] Use the pre-trained word vector model to convert each word obtained by word segmentation into a word vector;

[0162] The text vectors of the words obtained by word segmentation form the text vector corresponding to the specified question.

[0163] The question recommendation device provided by the embodiments of the present application combines the question screening method based on text vectors with the semantic understanding ability of the large language model, which can achieve accurate recommendation. In addition, the question recommendation device provided by the embodiments of the present application can effectively filter out irrelevant questions through multi-stage screening, and has a high recommendation efficiency. In addition, the question recommendation device provided by the embodiments of the present application can generate personalized similar question recommendations in real time according to the questions specified by the user, which can meet the needs of users in different learning scenarios, thereby improving the learning experience and learning effect of users.

[0164] The embodiments of the present application also provide an electronic device, which may include: at least one processor and a memory connected to the processor.

[0165] The processor may be a central processing unit CPU, or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application, etc.; the memory may include a high-speed RAM memory, and may also include non-volatile memory, such as at least one disk memory.

[0166] Among them, the memory is used to store a computer program, and the processor is used to execute the computer program so that the electronic device can implement the steps of the question recommendation method provided in the above embodiments.

[0167] An embodiment of the present application further provides a computer storage medium, which carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement the steps of the question recommendation method provided in the above embodiments.

[0168] An embodiment of the present application further provides a computer program product, including computer-readable instructions. When the computer-readable instructions run on an electronic device, the electronic device can implement the steps of the question recommendation method provided in the above embodiments.

[0169] In addition, it should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in the attached drawings of the device embodiments provided in the present application, the connection relationship between modules indicates that they have a communication connection, which can be specifically implemented as one or more communication buses or signal lines.

[0170] Through the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software plus necessary general hardware, and of course, it can also be implemented by dedicated hardware including dedicated integrated circuits, dedicated CPUs, dedicated memories, dedicated components, etc. Generally, functions completed by computer programs can be easily implemented by corresponding hardware, and the specific hardware structures for implementing the same function can also be various, such as analog circuits, digital circuits or dedicated circuits. However, for the present application, in more cases, software program implementation is a better implementation method. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a readable storage medium, such as a floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk or optical disc of a computer, etc., and includes several instructions to enable a computer device (which can be a personal computer, training device, or network device, etc.) to execute the methods described in various embodiments of the present application.

[0171] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product.

[0172] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from a website, computer, training device, or data center to another website, computer, training device, or data center by wire (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wirelessly (such as infrared, wireless, microwave, etc.). The computer-readable storage medium may be any available medium that can be stored by a computer or a data storage device such as a training device or a data center that includes one or more integrated available media. The available medium may be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)), etc.

Claims

1. A question recommendation method, characterized in that, Including: Obtain a specified topic; Convert the specified topic into a text vector to obtain the text vector corresponding to the specified topic; Obtain topics similar to the specified topic by calculating the similarity between the text vector corresponding to the specified topic and the text vectors corresponding to each topic in the topic database, and obtain a set of candidate topics; Utilize a large language model to perform semantic understanding on the specified topic and each topic in the set of candidate topics, and generate scores corresponding to each topic in the set of candidate topics, where the score corresponding to any topic in the set of candidate topics reflects the similarity and semantic relevance between this topic and the specified topic; Determine and recommend recommended topics according to the scores corresponding to each topic in the set of candidate topics.

2. The topic recommendation method according to claim 1, characterized in that The obtaining of topics similar to the specified topic by calculating the similarity between the text vector corresponding to the specified topic and the text vectors corresponding to each topic in the topic database includes: Calculate the similarity between the text vector corresponding to the specified topic and the text vectors corresponding to each topic in the topic database to obtain the similarities corresponding to each topic in the topic database; Sort the similarities corresponding to each topic in the topic database in descending order; Obtain the topics corresponding to the top N similarities as the topics similar to the specified topic, where N is an integer greater than 1.

3. The topic recommendation method according to claim 1, wherein The utilization of a large language model to perform semantic understanding on the specified topic and each topic in the set of candidate topics, and generate scores corresponding to each topic in the set of candidate topics includes: Utilize a large language model to analyze the specified topic and each topic in the set of candidate topics from multiple dimensions, and generate scores corresponding to each topic in the set of candidate topics; Among them, the multiple dimensions include some or all of the following dimensions: the knowledge points involved in the topic, the logical relationship of the topic, the problem-solving ideas of the topic.

4. The topic recommendation method according to claim 3, wherein The utilization of a large language model to analyze the specified topic and each topic in the set of candidate topics from multiple dimensions, and generate scores corresponding to each topic in the set of candidate topics includes: Obtain a prompt format template, where the prompt format template includes a specified topic information slot and a candidate topic information slot, and the prompt format template is used to instruct the large language model to analyze the topic in the specified topic information slot and each topic in the candidate topic information slot from multiple dimensions, and give a score reflecting the similarity and semantic relevance between each topic in the candidate topic information slot and the topic in the specified topic information slot; Fill the specified topic into the specified topic information slot of the prompt format template, and fill the topics in the set of candidate topics into the candidate topic information slot of the prompt format template to obtain a prompt instruction prompt; Input the prompt instruction prompt into the large language model to obtain the scores corresponding to each topic in the set of candidate topics output by the large language model.

5. The topic recommendation method according to claim 1, characterized in that The determining and recommending of recommended topics according to the scores corresponding to each topic in the set of candidate topics includes: Obtain the highest score from the scores corresponding to each question in the candidate question set; If the highest score is greater than or equal to a preset score threshold, determine the question corresponding to the highest score as the recommended question and recommend it.

6. The topic recommendation method according to claim 1, characterized in that The converting the specified question into a text vector to obtain the text vector corresponding to the specified question includes: Clean the specified question and perform word segmentation on the cleaned question; Use a pre-trained word vector model to convert each word obtained through word segmentation into a word vector; The text vector corresponding to the specified question is composed of the word vectors of each word obtained through word segmentation.

7. A question recommendation device, characterized in that, Including: A question acquisition module, a question vectorization module, a candidate question set acquisition module, a question scoring module, a recommended question determination module, and a question recommendation module; The question acquisition module is used to acquire a specified question; The question vectorization module is used to convert the specified question into a text vector to obtain the text vector corresponding to the specified question; The candidate question set acquisition module is used to obtain questions similar to the specified question by calculating the similarity between the text vector corresponding to the specified question and the text vector corresponding to each question in the question database, and obtain a candidate question set; The question scoring module is used to use a large language model to perform semantic understanding on the specified question and each question in the candidate question set, and generate scores corresponding to each question in the candidate question set, wherein the score corresponding to any question in the candidate question set reflects the similarity degree and semantic relevance between the question and the specified question; The recommended question determination module is used to determine a recommended question according to the scores corresponding to each question in the candidate question set; The question recommendation module is used to recommend the recommended question.

8. An electronic device, characterized in that, Including at least one processor and a memory connected to the processor, wherein: The memory is used to store a computer program; The processor is used to execute the computer program so that the electronic device can implement the steps of the question recommendation method described in any one of claims 1 to 6.

9. A computer storage medium, characterized in that, The storage medium carries one or more computer programs, and when the one or more computer programs are executed by an electronic device, the electronic device can implement the steps of the question recommendation method described in any one of claims 1 to 6.

10. A computer program product, characterized in that, Including computer-readable instructions, when the computer-readable instructions run on an electronic device, the electronic device can implement the steps of the question recommendation method described in any one of claims 1 to 6.

Images