Question recommendation method and device, equipment and medium
By reversely extracting reference problems from knowledge materials and saving them to the problem database, the challenges of intelligent problem recommendation systems in terms of data scarcity and problem diversity are solved, and efficient and accurate problem recommendations and system scalability are achieved.
Patent Information
- Application Number
- CN202411447811.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-16
- Publication Date
- 2025-02-07
AI Technical Summary
Existing smart problem recommendation systems have challenges in data scarcity and problem diversity, resulting in poor user experience and lack of diversity in recommendation issues.
By obtaining the knowledge data to be uploaded to the knowledge database and splicing it with the pre-configured text prompts, using the pre-trained text processing model to reverse extract reference questions from the knowledge data, and saving these questions to the problem database for recommendation.
It improves the efficiency and accuracy of problem generation, reduces labor and time costs, enriches the problem database, improves the pertinence and relevance of problem recommendations, and improves the scalability of the system.
Smart Images
Figure CN119396951A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of artificial intelligence, and in particular to a question recommendation method, device, equipment and medium. Background Art
[0002] With the rapid development of artificial intelligence technology, intelligent question recommendation systems have become an important tool for improving user experience and enhancing service efficiency. Such systems collect and analyze user question data and use advanced search engines (such as Elasticsearch) to build a huge pool of question content, aiming to provide users with fast and accurate recommendations of similar questions, thereby accelerating the question-solving process and reducing the burden on manual customer service. However, although intelligent question recommendation systems have shown great potential in many fields, they still face a series of challenges and limitations in practical applications.
[0003] First, the cold start problem is the main difficulty faced by the intelligent question recommendation system in its early stages. When the system was first launched, due to the lack of sufficient user question data and the relatively scarce question content pool, the number of questions that could be recommended was extremely limited, making it difficult to meet the diverse needs of users. This data scarcity directly limits the effectiveness and coverage of the recommendation system, affecting the initial establishment of user experience.
[0004] Secondly, the homogeneity of user questions exacerbates the lack of diversity in recommended questions. In actual applications, users often ask questions around similar or identical questions, resulting in a large number of similar questions being repeated in the question content pool. This homogeneity phenomenon not only wastes storage resources, but also makes the recommendation system tend to recommend the same or highly similar questions, limiting the diversity and novelty of recommended questions and making it difficult to meet users' needs for diverse information.
[0005] In summary, existing intelligent question recommendation systems still face many challenges in terms of data scarcity and question diversity. Summary of the invention
[0006] The present application provides a question recommendation method, apparatus, device and medium for solving the problems of data scarcity and poor question diversity in existing intelligent question recommendation systems.
[0007] In a first aspect, the present application provides a question recommendation method, the method comprising:
[0008] Obtain any knowledge material to be uploaded to the knowledge database;
[0009] The knowledge material and a pre-configured first text prompt are combined to obtain first data; wherein the first text prompt is used to indicate reverse extraction of questions from the knowledge material;
[0010] Obtaining reference questions reversely extracted from the knowledge material based on the first data through a pre-trained text processing large model;
[0011] The reference questions are saved in a question database to perform question recommendations based on the reference questions saved in the question database.
[0012] In a second aspect, the present application further provides a question recommendation device, the device comprising:
[0013] An acquisition unit, used for acquiring any knowledge material to be uploaded to the knowledge database;
[0014] A preprocessing unit, used for splicing the knowledge material and a pre-configured first text prompt to obtain first data; wherein the first text prompt is used to indicate reverse extraction of questions from the knowledge material;
[0015] A question generating unit, configured to obtain reference questions reversely extracted from the knowledge material based on the first data by using a pre-trained text processing model;
[0016] The storage unit is used to save the reference questions to a question database so as to make question recommendations based on the reference questions saved in the question database.
[0017] In a third aspect, the present application provides a computer device, the computer device comprising a processor, the processor being configured to implement the steps of the question recommendation method as described above when executing a computer program stored in a memory.
[0018] In a fourth aspect, the present application provides a computer-readable storage medium storing a computer program, which implements the steps of the question recommendation method as described above when the computer program is executed by a processor.
[0019] The beneficial effects of this application are as follows:
[0020] 1. Through the pre-trained text processing model, it is possible to automatically extract reference questions from a large amount of knowledge materials, greatly improving the efficiency and accuracy of question generation. Compared with traditional manual annotation or question generation methods, this method significantly reduces labor costs and time costs.
[0021] 2. As the extracted reference questions are continuously saved in the question database, the content of the question database will continue to be enriched and expanded. This will provide the question recommendation system with a wider and more diverse question resource, thus meeting the needs of different users in different scenarios.
[0022] 3. Reverse extraction of questions from knowledge materials can make the extracted questions closer to the knowledge materials, which is conducive to improving the pertinence and relevance of question recommendations. Since these questions are directly derived from the core content or key points of knowledge materials, they can more accurately reflect the themes and key points of knowledge materials, thereby providing users with more accurate and valuable question recommendations.
[0023] 4. While adding new knowledge materials to the knowledge database, new questions can be added to the question database, which improves the scalability of question recommendations. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0025] Figure 1 A schematic diagram of a question recommendation process provided in an embodiment of the present application;
[0026] Figure 2 A schematic diagram of a specific problem recommendation process provided in an embodiment of the present application;
[0027] Figure 3 A schematic diagram of the structure of a question recommendation device provided in an embodiment of the present application;
[0028] Figure 4 It is a structural schematic diagram of a computer device provided in an optional embodiment of the present application. DETAILED DESCRIPTION
[0029] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.
[0030] In order to provide more diverse recommended questions when user question data is scarce, the present application provides a question recommendation method, device, equipment and medium.
[0031] Embodiment 1:
[0032] This application provides a question recommendation method. Figure 1 A schematic diagram of a question recommendation process provided in an embodiment of the present application, the process includes:
[0033] S101: Obtain any knowledge material to be uploaded to the knowledge database.
[0034] The method recommended in this application is applied to a computer device, which may be an intelligent terminal, such as a computer, a robot, etc., or a server, such as an application server, a business server, etc.
[0035] In this application, a database for storing knowledge materials is maintained and constructed in advance. These knowledge materials can be documents, articles, books, professional terminology explanations, etc. They are organized into a format that can be accessed and queried by computer programs. Among them, the knowledge materials in the database can come from a single technical field or from different technical fields, which is not limited here. When a certain knowledge material needs to be uploaded to the knowledge database, the question recommendation method provided by this application can be used at the same time to process the knowledge material to expand the content of the questions that can be recommended.
[0036] In one example, the data in the knowledge database is stored in the form of vectors. By converting knowledge materials into vector representations, vector space can be used to measure the similarity and distance between knowledge items, thereby achieving fast and accurate information retrieval and recommendation. Exemplarily, when any knowledge material to be uploaded to the knowledge database is obtained, the knowledge material can be vectorized first, and then the obtained vector can be uploaded to the knowledge database.
[0037] S102: Joining the knowledge material and a pre-configured first text prompt to obtain first data; wherein the first text prompt is used to indicate reverse extraction of questions from the knowledge material.
[0038] A text prompt is a predefined text segment used to guide or instruct the text processing model to process the input data. In this application, the first text prompt is used to instruct the text processing model to reversely extract questions from knowledge materials. The following is an example of the first text prompt template:
[0039] “
[0040] You are a content extraction expert, good at extracting questions from paragraphs. Please extract relevant questions from the following paragraphs.
[0041] Content: {content}
[0042] The question you generate is: ".
[0043] After the knowledge data is acquired based on the above embodiment, the knowledge data and the first text prompt may be spliced to obtain the first data.
[0044] S103: Obtain reference questions reversely extracted from the knowledge material based on the first data through a pre-trained text processing large model.
[0045] In order to achieve a variety of recommendation questions, in this application, a large text processing model is pre-trained, and the large text processing model has the ability to reversely extract related questions (referred to as reference questions) from the input knowledge data. Exemplarily, after obtaining the first data, the first data can be input into the large text processing model. The input first data is processed by the large text processing model to obtain reference questions reversely extracted from the knowledge data.
[0046] In one example, in one example, the step of combining the knowledge material and the pre-configured first text prompt to obtain the first data includes:
[0047] Determine the paragraphs that the knowledge material contains;
[0048] For each of the paragraphs, the paragraph is concatenated with the first text prompt to obtain first data corresponding to the paragraph.
[0049] In the present application, the knowledge material can be divided into paragraphs to reduce the number of large paragraphs that generate questions with low quality. Then for each paragraph, the paragraph is spliced with the first text prompt to obtain the first data corresponding to the paragraph. Then, through the pre-trained text processing large model, based on the first data corresponding to the paragraph, the reference questions extracted from the paragraph are obtained in reverse. In this way, the questions that naturally form when humans read and understand the text can be simulated as much as possible, so as to generate high-quality reference questions that are close to actual needs. This method not only improves the efficiency and accuracy of question generation, but also makes the generated questions more diverse and in-depth, which helps to promote in-depth understanding and learning of the text content and realizes the mining of reference questions from each paragraph in the knowledge material.
[0050] In one example, the text processing model is trained as follows:
[0051] Obtain any knowledge material sample in the question recommendation sample set and a list of standard questions corresponding to the knowledge material sample;
[0052] splicing the knowledge material sample and the first text prompt to obtain third data;
[0053] Obtaining a list of expected questions reversely extracted from the knowledge material sample based on the third data through a pre-trained text retrieval model; wherein the text retrieval model has the ability to generate answer content based on the retrieved text;
[0054] Based on the expected question list and the standard question list, the text retrieval model is fine-tuned so that the trained text retrieval model has the ability to reversely extract questions.
[0055] In order to obtain a text processing model that can reversely extract questions from knowledge materials, in this application, a set of recommended sample sets with questions is collected in advance, so as to train the pre-trained text retrieval model through the knowledge material samples in the question recommendation sample set. The knowledge material samples can be text materials containing certain information such as documents, articles, and book chapters. In the question recommendation sample set, any knowledge material sample corresponds to a standard question list, which includes a series of questions associated with the knowledge material sample. These questions can be manually annotated or verified, and can accurately reflect the core content or key points of the knowledge material sample.
[0056] Obtain any knowledge material sample from the question recommendation sample set, and splice the knowledge material sample with the pre-allocated first text prompt to obtain third data. Input the third data into the pre-trained text retrieval model. The third data is processed by the text retrieval model to obtain a list of expected questions reversely extracted from the knowledge material sample. Then, based on the expected question list and the standard question list, determine a loss value to fine-tune the text retrieval model according to the loss value.
[0057] In a possible implementation, the list of standard questions corresponding to any knowledge material sample may include standard questions corresponding to each paragraph contained in the knowledge material sample. When training the text retrieval model, each paragraph contained in the knowledge material sample can be spliced according to the paragraph and the first text prompt to obtain the third data corresponding to the paragraph. Then, through the text retrieval model, based on the third data corresponding to the paragraph, the expected question reversely extracted from the paragraph is obtained. According to the expected question and the standard question corresponding to the paragraph, the loss value corresponding to the paragraph is determined. The loss values corresponding to all paragraphs in the knowledge material sample are accumulated to obtain the total loss value of the entire knowledge material sample. According to the total loss value, the text retrieval model is fine-tuned to optimize its ability to extract questions from paragraphs.
[0058] It should be noted that the text retrieval model is trained with massive amounts of text content, and the text retrieval model already has the ability to generate answer content based on the retrieved text.
[0059] The sample set recommended for this question contains a large number of knowledge material samples. The above operation is performed for each knowledge material sample. When the preset convergence conditions are met, the training of the large text processing model is completed.
[0060] Among them, satisfying the preset convergence condition may be that the sum of the loss values corresponding to the knowledge material samples in the problem recommendation sample set is less than a preset loss value threshold, or the number of iterations for training the large text retrieval model reaches a set maximum number of iterations, etc. In the specific implementation, the setting can be flexibly made and is not specifically limited here.
[0061] As a possible implementation method, when conducting model training, the knowledge material samples in the question recommendation sample set can be divided into training samples and test samples. The text retrieval model is first trained based on the training samples, and then the reliability of the trained text retrieval model is verified based on the test samples.
[0062] S104: Save the reference question to a question database, so as to perform question recommendation based on the reference question saved in the question database.
[0063] After the reference questions are obtained based on the above embodiment, the reference questions may be saved in a question database, thereby enriching the question pool for recommending questions.
[0064] In a possible implementation, the reference question may be saved in a question database in the form of a vector, so as to facilitate subsequent vector-based similarity search algorithms to quickly retrieve and recommend questions related to the user's question data.
[0065] The beneficial effects of this application are as follows:
[0066] 1. Through the pre-trained text processing model, it is possible to automatically extract reference questions from a large amount of knowledge materials, greatly improving the efficiency and accuracy of question generation. Compared with traditional manual annotation or question generation methods, this method significantly reduces labor costs and time costs.
[0067] 2. As the extracted reference questions are continuously saved in the question database, the content of the question database will continue to be enriched and expanded. This will provide the question recommendation system with a wider and more diverse question resource, thus meeting the needs of different users in different scenarios.
[0068] 3. Reverse extraction of questions from knowledge materials can make the extracted questions closer to the knowledge materials, which is conducive to improving the pertinence and relevance of question recommendations. Since these questions are directly derived from the core content or key points of knowledge materials, they can more accurately reflect the themes and key points of knowledge materials, thereby providing users with more accurate and valuable question recommendations.
[0069] 4. While adding new knowledge materials to the knowledge database, new questions can be added to the question database, which improves the scalability of question recommendations.
[0070] Embodiment 2:
[0071] In order to diversify the recommendation problem, based on the above embodiment, in this application, the method further includes:
[0072] Get user question data;
[0073] Determine, from the knowledge database, a target knowledge segment associated with the user question data, and determine, from the question database, a recommended question similar to the user question data;
[0074] splicing the target knowledge paragraph, the user question data and a pre-configured second text prompt to obtain second data; wherein the second text prompt is used to indicate answering the user question data according to the target knowledge paragraph;
[0075] Obtaining, by means of the text processing macromodel, an answer content to the user's question data based on the second data;
[0076] The answer content and the recommended question are output.
[0077] When a user uses the intelligent inquiry function, the user can input the user's question data. There are many ways to input the user's question data, such as voice input, text input, or action input, which are not specifically limited here.
[0078] In a possible implementation, the user question data may be preprocessed, such as removing stop words, stemming, spelling correction, etc., to help reduce noise and improve the accuracy and efficiency of subsequent processing.
[0079] In a possible implementation, the time when the user asks the question, context information (such as the user's previous query history), etc. may be recorded to better understand and answer the user's question.
[0080] After the computer device obtains the user's question data, it can retrieve the target knowledge paragraph associated with the user's question data in the pre-built knowledge database. For example, the target knowledge paragraph containing the key information in the user's question data can be retrieved. At the same time, the recommended questions similar to the user's question data can also be determined from the question database.
[0081] In a possible implementation manner, determining, from the question database, recommended questions similar to the user question data includes:
[0082] For each of the reference questions in the question database, determining the similarity between the reference question and the user question data;
[0083] Determine, from each of the reference questions, candidate reference questions whose corresponding similarities are within a preset similarity range;
[0084] The N candidate reference questions with the highest similarity are determined as the recommended questions; wherein N is an integer greater than or equal to 1.
[0085] In the present application, when determining the recommended questions, the similarity between the reference question and the user's question data can be determined for each reference question in the question database, thereby determining the correlation between the reference question and the user's question data. The similarity can be calculated based on a variety of methods, such as text similarity algorithms (such as cosine similarity, Jaccard similarity), keyword matching, semantic analysis, etc. Then, according to a preset similarity range, candidate reference questions with similarities within the similarity range are screened out from all reference questions. From the candidate reference questions, select the N questions with the highest similarity as recommended questions. N is an integer greater than or equal to 1, and the selection of N depends on the needs and scenarios of the actual application, and usually needs to consider the display space of the user interface and the number of recommendations expected by the user.
[0086] It should be noted that when setting the similarity range, if you want to avoid the recommended questions being consistent with the user's question data, you can set the upper threshold of the similarity range lower; if you want to avoid the recommended questions being irrelevant to the user's question data, you can set the lower threshold of the similarity range higher, and the upper threshold is greater than the lower threshold.
[0087] In some possible implementations, if the data in the knowledge database and the question database are both stored in the form of vectors, then after the computer device obtains the user question data, it can also convert the user question data into vectors, so that the knowledge database and the question database can be quickly retrieved through the vectors corresponding to the user question data.
[0088] The target knowledge paragraph, the user question data and the pre-configured second text prompt are concatenated to form new input data (recorded as the second data).
[0089] The input second data is processed by the text processing model to generate answer content for the user's question data.
[0090] After obtaining the answer content and the recommended content based on the above embodiment, the generated answer content and the retrieved recommended questions can be output to the user. In this way, the user can not only get a direct answer to his question, but also see a list of similar questions to further understand or explore related topics.
[0091] In a possible implementation, the method further includes:
[0092] If a confirmation operation on the answer content is received, and the question database does not store a reference question consistent with the user question data, the user question data is updated in the question database.
[0093] In order to increase the diversity of reference questions in the question database, in the present application, the user question data input by the user can be updated to the question database. Exemplarily, the user can determine whether the answer content can solve the user question data input by the user based on the output answer content. If the user is satisfied with the answer content, the user can express his approval of the answer content by inputting a confirmation operation on the answer content (such as clicking "OK", "Satisfied" or similar buttons). At this time, after the computer device receives the confirmation operation, it can further check whether there is a reference question in the question database that is completely consistent with the user question data. If it is confirmed through comparison that the user question data is not yet included in the question database, an update operation can be performed to add the user question data as a new reference question to the question database.
[0094] In a possible implementation, if a confirmation operation on the answer content is received, the user question data and the answer content can be used as training data for subsequent optimization of the text processing model.
[0095] Embodiment 3:
[0096] Figure 2 A schematic diagram of a specific question recommendation process provided in an embodiment of the present application, the process includes:
[0097] S201: Obtain any knowledge material to be uploaded to the knowledge database.
[0098] S202: vectorizing the knowledge data to obtain a knowledge vector corresponding to the knowledge data.
[0099] S203: Save the knowledge vector corresponding to the knowledge document into the knowledge database.
[0100] S204: Determine each paragraph included in the knowledge material.
[0101] For each paragraph included in the knowledge material, the following S205 to S208 are executed:
[0102] S205: splicing the paragraph with a pre-configured first text prompt to obtain first data corresponding to the paragraph.
[0103] S206: Obtain reference questions reversely extracted from the current paragraph based on the first data corresponding to the paragraph through a pre-trained text processing large model.
[0104] S207: vectorizing the acquired reference questions to obtain reference question vectors corresponding to the reference questions.
[0105] S208: Save each reference question vector into a question database.
[0106] S209: Obtain user question data and perform vectorization processing.
[0107] S210: Determine a target knowledge segment associated with the user's question data from a pre-built knowledge database, and determine a recommended question similar to the user's question data from a question database.
[0108] In a possible implementation, determining recommended questions similar to the user's question data from a question database includes:
[0109] For each reference question in the question database, determine the similarity between the reference question and the user question data;
[0110] Determine, from each reference question, candidate reference questions whose corresponding similarities are within a preset similarity range;
[0111] The N candidate reference questions with the highest similarity are determined as recommended questions; where N is an integer greater than or equal to 1.
[0112] S211: splicing the target knowledge paragraph, the user question data and the pre-configured second text prompt to obtain second data.
[0113] The second text prompt is used to instruct to answer the user's question data according to the target knowledge paragraph.
[0114] S212: Obtain the answer content of the user's question data based on the second data through the text processing large model.
[0115] S213: Output the answer content and recommended questions.
[0116] In a possible implementation, if a confirmation operation on the answer content is received, and the question database does not store a reference question consistent with the user question data, the user question data is updated in the question database.
[0117] Embodiment 4:
[0118] Based on the same inventive concept, the present application also provides a question recommendation device. Figure 3A schematic diagram of a question recommendation device provided in an embodiment of the present application, the device comprising:
[0119] An acquisition unit 31 is used to acquire any knowledge material to be uploaded to the knowledge database;
[0120] A preprocessing unit 32, configured to combine the knowledge material and a pre-configured first text prompt to obtain first data; wherein the first text prompt is used to indicate reverse extraction of questions from the knowledge material;
[0121] A question generating unit 33 is used to obtain reference questions extracted from the knowledge material in reverse based on the first data by using a pre-trained text processing model;
[0122] The storage unit 34 is used to save the reference questions to a question database, so as to make question recommendations based on the reference questions saved in the question database.
[0123] The question recommendation device in this embodiment is presented in the form of a functional module, where the module refers to an application specific integrated circuit (ASIC), a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.
[0124] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.
[0125] Embodiment 5:
[0126] See also Figure 4 , Figure 4 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present application, such as Figure 4 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 4 A processor 10 is taken as an example.
[0127] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.
[0128] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.
[0129] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created by the use of a computer device based on the presentation of a small program landing page, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0130] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.
[0131] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Figure 4 The example of connecting through bus is taken in the following.
[0132] The input device 30 can receive input digital or character information, and generate key signal input related to the user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, an indicator bar, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (e.g., an LED) and a tactile feedback device (e.g., a vibration motor), etc. The above-mentioned display device includes but is not limited to a liquid crystal display, a light emitting diode, a display and a plasma display. In some optional embodiments, the display device can be a touch screen.
[0133] Embodiment 6:
[0134] On the basis of the above embodiments, an embodiment of the present application further provides a computer-readable storage medium, in which a computer program executable by a processor is stored. When the program runs on the processor, the processor implements the following steps when executing:
[0135] Obtain any knowledge material to be uploaded to the knowledge database;
[0136] The knowledge material and a pre-configured first text prompt are combined to obtain first data; wherein the first text prompt is used to indicate reverse extraction of questions from the knowledge material;
[0137] Obtaining reference questions reversely extracted from the knowledge material based on the first data through a pre-trained text processing large model;
[0138] The reference questions are saved in a question database to perform question recommendations based on the reference questions saved in the question database.
[0139] Since the principle of solving the problem by the above-mentioned computer-readable storage medium is similar to that of the problem recommendation method, the implementation of the above-mentioned computer-readable storage medium can refer to the embodiment of the method, and the repeated parts will not be repeated.
[0140] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.
Claims
1. A question recommendation method, characterized in that: The method comprises: Obtain any knowledge material to be uploaded to the knowledge database; The knowledge material and a pre-configured first text prompt are combined to obtain first data; wherein the first text prompt is used to indicate reverse extraction of questions from the knowledge material; Obtaining reference questions reversely extracted from the knowledge material based on the first data through a pre-trained text processing large model; The reference questions are saved in a question database to perform question recommendations based on the reference questions saved in the question database.
2. The method according to claim 1, characterized in that The step of combining the knowledge material and the pre-configured first text prompt to obtain the first data includes: Determine the paragraphs that the knowledge material contains; For each of the paragraphs, the paragraph is concatenated with the first text prompt to obtain first data corresponding to the paragraph.
3. The method according to claim 1, characterized in that The data in the knowledge database and the question database are both stored in the form of vectors.
4. The method according to claim 1, characterized in that The method further comprises: Get user question data; Determine, from the knowledge database, a target knowledge segment associated with the user question data, and determine, from the question database, a recommended question similar to the user question data; splicing the target knowledge paragraph, the user question data and a pre-configured second text prompt to obtain second data; wherein the second text prompt is used to indicate answering the user question data according to the target knowledge paragraph; Obtaining, by means of the text processing macromodel, an answer content to the user's question data based on the second data; The answer content and the recommended question are output.
5. The method according to claim 4, characterized in that The method further comprises: If a confirmation operation on the answer content is received, and the question database does not store a reference question consistent with the user question data, the user question data is updated in the question database.
6. The method according to claim 4, characterized in that Determining, from the question database, recommended questions similar to the user question data, includes: For each of the reference questions in the question database, determining the similarity between the reference question and the user question data; Determine, from each of the reference questions, candidate reference questions whose corresponding similarities are within a preset similarity range; The N candidate reference questions with the highest similarity are determined as the recommended questions; wherein N is an integer greater than or equal to 1.
7. The method according to claim 1, characterized in that The text processing model is trained in the following way: Obtain any knowledge material sample in the question recommendation sample set and a list of standard questions corresponding to the knowledge material sample; splicing the knowledge material sample and the first text prompt to obtain third data; Obtaining a list of expected questions reversely extracted from the knowledge material sample based on the third data through a pre-trained text retrieval model; wherein the text retrieval model has the ability to generate answer content based on the retrieved text; Based on the expected question list and the standard question list, the text retrieval model is fine-tuned so that the trained text retrieval model has the ability to reversely extract questions.
8. A question recommendation device, characterized in that: The device comprises: An acquisition unit, used for acquiring any knowledge material to be uploaded to the knowledge database; A preprocessing unit, used for splicing the knowledge material and a pre-configured first text prompt to obtain first data; wherein the first text prompt is used to indicate reverse extraction of questions from the knowledge material; A question generating unit, configured to obtain reference questions reversely extracted from the knowledge material based on the first data by using a pre-trained text processing model; The storage unit is used to save the reference questions to a question database so as to make question recommendations based on the reference questions saved in the question database.
9. A computer device, characterized in that: The computer device comprises a processor, and the processor is used to implement the steps of the question recommendation method as described in any one of claims 1 to 7 when executing the computer program stored in the memory.
10. A computer-readable storage medium, characterized in that: It stores a computer program executable by a computer device. When the program is run on the computer device, the computer device executes the steps of the question recommendation method as described in any one of claims 1 to 7.