Recommended question obtaining method and device, equipment, medium and program product
By obtaining target questions in the terminal, performing field extraction and demand recognition, and generating question recommendation prompt words, the problem of inaccurate answers due to single question methods is solved in the prior art, and a more accurate answer recall is achieved.
Patent Information
- Application Number
- CN202510592932.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-08
- Publication Date
- 2025-08-12
AI Technical Summary
In the prior art, due to the relatively single form of the input question and the method of questioning, the recalled answers are not accurate enough and cannot meet the real needs of users.
By obtaining the target problem input by the terminal, field extraction and demand identification are performed, demand keyword information is obtained, and searching in the preset database to generate problem recommendation prompt words, and finally obtain recommendation questions related to the target problem.
It improves the accuracy of the answer, meets the real needs of users, and improves the user experience.
Smart Images
Figure CN120470091A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the field of computer technology, and in particular to a method, apparatus, device, medium, and program product for obtaining recommendation questions. Background Art
[0002] In intelligent question-answering systems, the accuracy of answers is closely tied to the user's questions. Suboptimal answers can be caused not only by system issues but also by the user's questioning. Different questioning methods can affect both corpus recall and question-answer generation. Therefore, in the question-answering phase, differences in how users ask questions can influence the source of answers selected by the large language model, even if recall is correct, and thus affect the quality of the answers.
[0003] In related technologies, answers are generated based on the corpus recalled from the current question. However, due to the relatively simple form and question method of the input question, the recalled answers are not accurate enough and cannot meet the real needs of users. Summary of the Invention
[0004] In view of this, the present disclosure provides a method, apparatus, device, medium, and program product for obtaining recommended questions to solve the problem that the current input question format and questioning method are relatively simple, resulting in the recalled answers being inaccurate and unable to meet the real needs of users.
[0005] In a first aspect, the present disclosure provides a method for obtaining a recommendation question, the method comprising:
[0006] Get the target question entered in the terminal;
[0007] In response to the terminal's acquisition request, field extraction and demand identification are performed on the target question to obtain demand keyword information;
[0008] Search the preset database based on the required keyword information to obtain the initial recall data;
[0009] Analyze the initial recall data and generate question recommendation prompt words;
[0010] Based on the question recommendation prompt words, recommended questions related to the target question are obtained.
[0011] In the disclosed embodiment, a target question inputted at a terminal is obtained; in response to the terminal's acquisition request, field extraction and demand identification are performed on the target question to obtain demand keyword information; a search is performed within a preset database based on the demand keyword information to obtain initial recall data; based on the initial recall data, question recommendation prompt words are generated; and based on the question recommendation prompt words, recommended questions related to the target question are obtained. In this way, demand keyword information is extracted based on the current question, and recommended questions with relatively good answer effects and helpfulness to the user are generated based on the demand keyword information. This allows the large language model to ask questions based on the recommended questions, helping users further understand relevant knowledge and obtain more accurate answers. This solves the problem in related technologies where the recalled answers are inaccurate and cannot meet the user's real needs due to the relatively simple form and method of asking questions.
[0012] In a second aspect, the present disclosure provides a device for obtaining a recommendation question, the device comprising:
[0013] The acquisition module is used to obtain the target question input in the terminal;
[0014] The first obtaining module is used to extract fields and identify requirements for the target question in response to the terminal's obtaining request, and obtain the required keyword information;
[0015] The retrieval module is used to search the preset database according to the required keyword information to obtain the initial recall data;
[0016] The generation module is used to analyze the initial recall data and generate question recommendation prompt words;
[0017] The second obtaining module is used to obtain recommended questions related to the target question based on the question recommendation prompt words.
[0018] In a third aspect, the present disclosure provides a computer device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to execute the method for obtaining recommended questions according to the first aspect or any corresponding embodiment thereof.
[0019] In a fourth aspect, the present disclosure provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the method for obtaining recommended questions according to the first aspect or any corresponding embodiment thereof.
[0020] In a fifth aspect, the present disclosure provides a computer program product, including computer instructions, which are used to enable a computer to execute the method for obtaining recommended questions according to the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] In order to more clearly illustrate the specific embodiments of the present disclosure or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0022] Figure 1 is a flowchart of a method for obtaining recommended questions according to an embodiment of the present disclosure;
[0023] Figure 2 is a complete flowchart of a method for obtaining recommended questions according to an embodiment of the present disclosure;
[0024] Figure 3 is a flowchart of another method for obtaining recommended questions according to an embodiment of the present disclosure;
[0025] Figure 4 It is a simplified diagram of a terminal interface display according to an embodiment of the present disclosure;
[0026] Figure 5 is a flowchart of a method for obtaining a recommendation question according to another embodiment of the present disclosure;
[0027] Figure 6 is a structural block diagram of a device for obtaining recommendation questions according to an embodiment of the present disclosure;
[0028] Figure 7 Schematic diagram of the hardware structure of the computer device according to the embodiment of the present disclosure. DETAILED DESCRIPTION
[0029] To make the purpose, technical solutions, and advantages of the embodiments of the present disclosure more clear, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of the present disclosure.
[0030] In intelligent question-answering systems, the accuracy of answers is closely tied to the user's question. Different question-asking techniques can impact both corpus recall and question-answer generation. For example, in the recall phase, issues like incomplete question descriptions, mismatches between the question and target document, and typos directly impact whether the recalled document meets user needs. Therefore, in the question-answering phase, differences in how the user asks the question can influence the source of the answer selected by the large language model, even if the recall is successful, and thus impact the quality of the answer.
[0031] In the related art, due to the relatively simple form and method of input questions, the recalled answers are not accurate enough and cannot meet the real needs of users. To solve the above problems, according to an embodiment of the present disclosure, an embodiment of a method for obtaining recommended questions is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.
[0032] In this embodiment, a method for obtaining a recommendation question is provided, such as Figure 1 As shown, Figure 1 : is a flowchart of a method for obtaining recommended questions according to an embodiment of the present disclosure. The method can be applied to a backend server side. The method flow includes the following steps:
[0033] Step S101: obtaining a target question inputted at a terminal.
[0034] Optionally, in the embodiment of the present disclosure, the server will monitor the terminal in real time to obtain the target question entered by the user at the terminal. The terminal here can be a PC, tablet, mobile phone, etc. The user can enter the target question in the first area set by the terminal. The first area here can be the upper left corner of the terminal. Figure 2 , it can also be other locations and is not limited to this.
[0035] In addition, the target question here refers to the question that the user wants to ask, such as: What are the recommended places in the closed-loop e-commerce mall? What is the weather like today?
[0036] Step S102 , in response to the acquisition request of the terminal, performing field extraction and demand identification on the target question to obtain demand keyword information.
[0037] Optionally, after the server detects that the user has input the target question at the terminal, in response to the terminal's request to obtain the target question, the server performs field extraction and demand identification on the obtained target question to obtain demand keyword information. Among them, field identification can use regular expressions. By writing regular expression patterns, the fields in the target question can be quickly and accurately identified. For example, in "Search for which rit positions are in the closed-loop e-commerce mall", the identified fields are: "search", "closed loop", "e-commerce mall", "rit position", "yes", and "which ones". The identified fields are then subjected to demand identification again to obtain demand keyword information. For example, key words and phrases are extracted from the fields. These keywords can often reflect the user's needs. For example, "closed loop", "e-commerce mall", and "rit position" are demand keyword information.
[0038] Step S103: searching the preset database according to the required keyword information to obtain initial recall data.
[0039] Optionally, after obtaining the demand keyword information contained in the target question, the server searches for the demand keyword information in a preset database. The search here can be understood as an information matching process, that is, matching data related to the demand keyword information from the preset database as the initial recall data for this recall.
[0040] Step S104: Analyze the initial recall data and generate question recommendation prompt words.
[0041] Optionally, based on the initial recall data, the server generates question recommendation prompts through three key steps: content analysis, demand extraction, and language organization. For example, the initial recall data is categorized by topic, field, product category, and so on. This classification provides a clear understanding of the data distribution and provides precise guidance for generating prompts. The initial recall data is then mined for user needs: Based on the initial recall data, user needs are determined and detailed analysis is conducted, which will be crucial for generating prompts. Finally, the server combines the analyzed information with user needs to generate question recommendation prompts using concise and accurate language.
[0042] Step S105: obtaining recommended questions related to the target question based on the question recommendation prompt words.
[0043] Optionally, after obtaining the question recommendation prompt word, the question recommendation prompt word is combined with the large language model, and the large language model continuously adjusts and constructs recommended questions related to the target question.
[0044] In the disclosed embodiment, a target question inputted at a terminal is obtained; in response to the terminal's acquisition request, field extraction and demand identification are performed on the target question to obtain demand keyword information; a search is performed within a preset database based on the demand keyword information to obtain initial recall data; based on the initial recall data, question recommendation prompt words are generated; and based on the question recommendation prompt words, recommended questions related to the target question are obtained. In this way, demand keyword information is extracted based on the current question, and recommended questions with relatively good answer effects and helpfulness to the user are generated based on the demand keyword information. This allows the large language model to ask questions based on the recommended questions, helping users further understand relevant knowledge and obtain more accurate answers. This solves the problem in related technologies where the recalled answers are inaccurate and cannot meet the user's real needs due to the relatively simple form and method of asking questions.
[0045] In some optional embodiments, such as Figure 3 As shown, Figure 3FIG. 4 is a flowchart of another method for obtaining recommended questions according to an embodiment of the present disclosure, the process including the following steps:
[0046] Step S201: Get the target question inputted in the terminal. Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.
[0047] Step S202: In response to the terminal's acquisition request, perform field extraction and demand identification on the target question to obtain demand keyword information. Figure 1 Step S102 of the illustrated embodiment will not be described in detail here.
[0048] Step S203: searching the preset database according to the required keyword information to obtain initial recall data.
[0049] Specifically, the preset database includes a business database and a data knowledge base, and the above step S203 includes:
[0050] Step S2031 , performing vectorization processing on the demand keyword to obtain a feature vector of the demand keyword.
[0051] Optionally, the requirement keywords are converted into vector representations. A common method is to use a word vector model (such as Word2Vec (Word to Vector) or GloVe (Global Vectors for Word Representation)) to vectorize the requirement keywords and obtain feature vectors of the requirement keywords.
[0052] Step S2032: Search in a preset database based on the feature vector to obtain first recalled data.
[0053] Optionally, document information and field information are stored in a preset database. At this time, the vectors of these document information and field information are obtained, and then the cosine similarity can be used to calculate the similarity between the user question vector (i.e., the feature vector of the demand keyword) and the document vector and field vector, thereby obtaining the first recall data.
[0054] Step S2033: Using a search tool to search in a preset database based on the required keyword information to obtain second recalled data.
[0055] Alternatively, the server can utilize a search tool, such as the ES search engine, using its full-text search functionality to search within a pre-set database based on the demand keyword information. Specifically, the demand keyword information is stored in the ES index. ES analyzes and processes the demand keyword information, such as performing word segmentation and filtering stop words, and then constructs an inverted index, enabling ES to quickly locate the second recall data containing the demand keyword information.
[0056] Step S2034: combining the first recall data and the second recall data to obtain initial recall data.
[0057] Optionally, after obtaining the first and second recall data, the two can be combined, such as by a simple union merge, combining all the data from the two result sets, or an intersection merge, retaining only the initial recall data that appears in both result sets. Alternatively, different weights can be assigned to the first and second recall data, and then these weights are applied to the recalled first and second recall data for a comprehensive calculation to obtain the initial recall data.
[0058] Step S204: Analyze the initial recall data and generate question recommendation prompt words. Figure 1 Step S104 of the illustrated embodiment will not be described in detail here.
[0059] Step S205: Get the recommended questions related to the target question based on the question recommendation prompt words. Figure 1 Step S105 of the illustrated embodiment will not be described in detail here.
[0060] In an embodiment of the present disclosure, first recall data is obtained by searching the feature vector of the demand keyword information, and second recall data is obtained by searching the demand keyword information using a search tool, and then initial recall data is obtained based on the first recall data and the second recall data, so that the final initial recall data can contain more accurate answers and improve user experience.
[0061] In some optional implementations, the preset database includes a business database and a data knowledge base, the first recalled data includes a first initial document and a first initial field, and the above step S2032 includes:
[0062] Step a1, matching the feature vector with the business data vector in the business database to obtain the first initial document after recall;
[0063] Step a2: Match the feature vector with the knowledge data vector in the data knowledge base to obtain the first initial field after recall.
[0064] The second recalled data includes a second initial document and a second initial field. Step S2033 includes:
[0065] Step b1: inputting the demand keyword information into a search tool, so that the search tool matches the demand keyword information with the business data in the business database to obtain a recalled second initial document;
[0066] Step b2: input the demand keyword information into the search tool, so that the search tool matches the demand keyword information with the knowledge data in the data knowledge base to obtain the recalled second initial field.
[0067] Optionally, the embodiment of the present disclosure sets the preset database to be composed of a business database and a data knowledge base, such as Figure 4 As shown. This allows ES retrieval and vector retrieval to achieve two-way recall and obtain initial recall data. Further, the feature vector is matched with the business data vector in the business database. For example, if the business data vector in the business database contains 100, the feature vector of the demand keyword information is matched with these 100 business data vectors respectively for similarity, thereby obtaining the first initial document after recall. Similarly, the feature vector is matched with the knowledge data vector in the data knowledge base. For example, if the knowledge data vector in the data knowledge base contains 100, the feature vector of the demand keyword information is matched with these 100 knowledge data vectors respectively for similarity, thereby obtaining the first initial field after recall.
[0068] The demand keyword information is input into the ES search engine so that the ES search engine matches the business data in the business database based on the demand keyword information, such as performing a full-text search. At this time, the query keyword will be segmented, and then matching documents will be found in the business data to obtain the second initial document after recall; the demand keyword information is input into the ES search engine so that the ES search engine matches the knowledge data in the data knowledge base based on the demand keyword information, such as performing a full-text search. At this time, the query keyword will be segmented, and then matching fields will be found in the knowledge data to obtain the second initial field after recall.
[0069] In the disclosed embodiment, the business database and the data knowledge base are integrated to realize the recall of documents and fields, so that the obtained first initial document, first initial field and second initial document, second initial field contain accurate answers, solving the problem that the answers recalled by related technologies are not accurate enough.
[0070] In some optional implementations, the initial recall data includes an initial document and an initial field, and the above step S2034 includes:
[0071] Step c1, performing a union or intersection merge on the first initial document and the second initial document to obtain an initial document;
[0072] Step c2: performing a union or intersection merge on the first initial field and the second initial field to obtain an initial field.
[0073] Optionally, based on the above embodiment, when obtaining the initial recall data, the result of the intersection or union of the first recall data and the second recall data can be used as the initial recall data. Since the first recall data includes the first initial document and the first initial field, the second recall data includes the second initial document and the second initial field, and the initial recall data includes the initial document and the initial field. Therefore, in the embodiment of the present disclosure, in order to obtain the initial recall data, the first initial document and the second initial document can be taken as the union, and the result of the union can be used as the initial document, or the first initial document and the second initial document can be taken as the intersection, and the result of the intersection can be used as the initial document. Figure 4 The 10 document knowledge shown in the figure is used as the initial document. Alternatively, the first initial field and the second initial field are combined to obtain the union result as the initial field, or the first initial field and the second initial field are combined to obtain the intersection result as the initial field. Figure 4 The 10 fields shown are used as initial fields.
[0074] In some optional embodiments, such as Figure 5 As shown, Figure 5 FIG. 4 is a flowchart of another method for obtaining recommended questions according to an embodiment of the present disclosure, the process including the following steps:
[0075] Step S301: Get the target question inputted in the terminal. Figure 3 Step S201 of the illustrated embodiment will not be described in detail here.
[0076] Step S302: In response to the terminal's acquisition request, perform field extraction and demand identification on the target question to obtain demand keyword information. Figure 3 Step S202 of the illustrated embodiment will not be described in detail here.
[0077] Step S303: Search the preset database based on the required keyword information to obtain the initial recall data. Figure 3 Step S203 of the illustrated embodiment will not be described in detail here.
[0078] Step S304: Analyze the initial recall data and generate question recommendation prompt words. Figure 3 Step S204 of the illustrated embodiment will not be described in detail here.
[0079] Step S305: obtaining recommended questions related to the target question based on the question recommendation prompt words.
[0080] Specifically, the above step S305 includes:
[0081] Step S3051: Input the question recommendation prompt words into the large language model and output a recommended question list.
[0082] Alternatively, as Figure 4 As shown in the figure, after obtaining the question recommendation prompt words, they can be combined with the large language model and input into the large language model, which then generates multiple recommended questions. Specifically, the large language model performs word segmentation on the input question recommendation prompt words. After word segmentation, each subword is mapped to the model's pre-trained word vector space and converted into a corresponding vector representation. This vector feature is then input into the large language model's multi-layer neural network for calculation, resulting in the recommended questions.
[0083] Since there are multiple recommendation questions, a recommendation question list is generated from these recommendation questions.
[0084] Step S3052: Obtain the user's historical questions and preset questions stored in the target system.
[0085] Optionally, some historical questions of users are stored in the server. These historical questions may be questions that users frequently use and may represent the user's habit of asking questions. In addition, in the target system, such as a certain system A, some preset questions are stored. The preset questions may be some high-quality questions. These high-quality questions may come from multiple users and are questions obtained after multiple rounds of selection of questions raised by multiple users. Generally speaking, the target system integrates business knowledge question and answer and data knowledge question and answer functions, which can further recommend different types of questions to users. For example, when a user asks a business question, similar questions such as the data table where the business is located can be recommended to expand the scope of the user's knowledge.
[0086] Step S3053: Combine the questions in the recommended question list with the historical questions and the preset questions to obtain recommended questions.
[0087] Alternatively, as Figure 4 As shown, the questions in the recommended question list are combined with historical questions and preset questions to obtain recommended questions. For example, the server selects a first number of questions from the recommended question list, a second number of questions from the user's historical questions, and a third preset number of preset questions from the target system. All selected questions are then deduplicated to obtain the final recommended questions. Alternatively, all questions in the recommended question list, all historical questions, and all preset questions can be directly deduplicated to obtain the final recommended questions.
[0088] It should be noted that the number of recommended questions finally obtained can be one or more, and the user can select one of the recommended questions and ask the question directly.
[0089] In the embodiment of the present disclosure, by combining the user's historical questions and the preset questions of the target system, the large language model can be used to further recommend relevant questions of interest to the user.
[0090] In some optional implementations, step S3053 includes:
[0091] Step d1, obtaining a first weight corresponding to the recommended question list, a second weight corresponding to the historical question, and a third weight corresponding to the preset question;
[0092] Step d2: obtaining a recommendation question based on the first weight, the second weight, and the third weight.
[0093] Optionally, the specific values of the first, second, and third quantities in the above embodiment are determined based on the first weight corresponding to the recommended question list, the second weight corresponding to the historical questions, and the third weight corresponding to the preset questions set by the user. For example, if the first weight > the second weight > the third weight, then the corresponding value of the first quantity > the value of the second quantity > the value of the third quantity. Alternatively, if the values of the first, second, and third quantities are the same but the weights are different, such as the first weight > the second weight > the third weight, when generating recommended questions, they will tend to align with the styles of the questions in the recommended list.
[0094] In the disclosed embodiment, corresponding weights are set for the recommended question list, historical questions, and preset questions, so that the recommended questions generated ultimately have a focus, thereby generating recommended questions that have better answer effects and are helpful to users.
[0095] In some optional embodiments, the above Figure 1 The corresponding embodiment further includes:
[0096] Step S11, generating question and answer prompt words based on the initial recall data;
[0097] Step S12: input the question-answer prompt words into the large language model to obtain the target answer corresponding to the target question and the answer source of the target answer;
[0098] Step S13: sending the target question, target answer, answer source and recommended question to the terminal, so that the terminal displays the target question, target answer, answer source and recommended question.
[0099] Alternatively, as Figure 4As shown, the input to the large language model also includes question and answer prompt words, based on which the target answer and the source of the target answer corresponding to the target question can be obtained. The target question, target answer and the source of the answer are then sent to the terminal so that the terminal displays the target question, target answer and the source of the answer. Figure 2 .
[0100] in addition, Figure 2 In the embodiment, the terminal also displays the recommended questions obtained in the above embodiment, so that what is displayed on the terminal is the target question, the target answer, the source of the answer, and the recommended questions related to the target question.
[0101] In the disclosed embodiment, the target question, target answer, answer source, and recommended questions are displayed on the terminal, allowing the user to intuitively see the interaction situation and improve the user experience.
[0102] In some optional implementations, the above step S11 includes:
[0103] Step c1, sorting the initial documents based on the first key information contained in the initial documents to obtain a first sorting queue;
[0104] Step c2, selecting a preset number of initial documents in the first sorting queue to obtain a target document;
[0105] Step c3, sorting the initial fields based on the second key information contained in the initial fields to obtain a second sorting queue;
[0106] Step c4, selecting a preset number of initial fields in the second sorting queue to obtain a target field;
[0107] Step c5: construct question-answer prompt words based on the target document, target field, and target question.
[0108] Alternatively, as Figure 4 As shown, after obtaining the initial documents (for example, 10) and initial fields (for example, 10), the initial documents and initial fields are finely sorted to obtain the finely sorted target documents and target fields, and then question and answer prompt words are constructed based on the target documents, target fields and target questions.
[0109] Specifically, the first key information contained in the 10 initial documents (i.e., the intention and characteristics of the documents) is obtained, and then combined with the user's needs, the initial documents are sorted to obtain a first sorting queue, and the first 5 initial documents in the first sorting queue are selected as target documents; the second key information contained in the 10 initial fields (i.e., the intention and characteristics of the fields) is obtained, and then combined with the user's needs, the initial fields are sorted to obtain a second sorting queue, and the first 5 initial fields in the second sorting queue are selected as target fields.
[0110] Then, the target documents, target fields and target questions are processed through three key steps: content analysis, demand extraction and language organization to generate question and answer prompts.
[0111] In the disclosed embodiment, by finely sorting the initial documents and initial fields, the large language model can obtain more accurate target answers based on the final generated question and answer prompt words, thereby improving the quality of the target answers.
[0112] This embodiment also provides a device for obtaining recommended questions, which is used to implement the above-mentioned embodiments and preferred embodiments. Details that have already been described will not be repeated here. As used below, the term "module" may refer to a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation using hardware, or a combination of software and hardware, is also possible and contemplated.
[0113] This embodiment provides a device for obtaining a recommendation question, such as Figure 6 As shown, including:
[0114] An acquisition module 601 is used to acquire a target question inputted at a terminal;
[0115] The first obtaining module 602 is used to extract fields and identify requirements for the target question in response to the terminal's obtaining request, and obtain requirement keyword information;
[0116] Retrieval module 603, used to search in a preset database according to the required keyword information to obtain initial recall data;
[0117] A generation module 604 is used to analyze the initial recall data and generate question recommendation prompt words;
[0118] The second obtaining module 605 is used to obtain recommended questions related to the target question based on the question recommendation prompt words.
[0119] In some optional embodiments, the preset database includes a business database and a data knowledge base, and the retrieval module 603 is used to vectorize the demand keywords to obtain the feature vectors of the demand keywords; search within the preset database based on the feature vectors to obtain the first recall data; use the search tool to search within the preset database based on the demand keyword information to obtain the second recall data; and combine the first recall data and the second recall data to obtain the initial recall data.
[0120] In some optional embodiments, the preset database includes a business database and a data knowledge base, the first recalled data includes a first initial document and a first initial field, and the retrieval module 603 is also used to match the feature vector with the business data vector in the business database to obtain the first initial document after recall; and match the feature vector with the knowledge data vector in the data knowledge base to obtain the first initial field after recall.
[0121] In some optional embodiments, the preset database includes a business database and a data knowledge base, the second recalled data includes a second initial document and a second initial field, and the retrieval module 603 is further configured to input the demand keyword information into a search tool, so that the search tool matches the business data in the business database based on the demand keyword information to obtain the recalled second initial document;
[0122] The demand keyword information is input into the search tool, so that the search tool matches the knowledge data in the data knowledge base based on the demand keyword information to obtain the recalled second initial field.
[0123] In some optional embodiments, the initial recall data includes an initial document and an initial field, and the retrieval module 603 is further used to perform a union or intersection merge on the first initial document and the second initial document to obtain the initial document; and perform a union or intersection merge on the first initial field and the second initial field to obtain the initial field.
[0124] In some optional implementations, the second obtaining module 605 is used to input question recommendation prompt words into the large language model and output a recommended question list; obtain the user's historical questions and preset questions stored in the target system; combine the questions in the recommended question list with the historical questions and preset questions to obtain recommended questions.
[0125] In some optional implementations, the second obtaining module 605 is further used to obtain the first weight corresponding to the recommended question list, the second weight corresponding to the historical question, and the third weight corresponding to the preset question; and obtain the recommended question based on the first weight, the second weight, and the third weight.
[0126] In some optional embodiments, the device also includes: generating question and answer prompt words based on initial recall data; inputting the question and answer prompt words into a large language model to obtain a target answer corresponding to the target question and the source of the answer to the target answer; sending the target question, target answer, answer source and recommended question to the terminal so that the terminal displays the target question, target answer, answer source and recommended question.
[0127] In some optional embodiments, the device also includes: sorting the initial documents based on the first key information contained in the initial documents to obtain a first sorting queue; selecting a preset number of initial documents in the first sorting queue to obtain a target document; sorting the initial fields based on the second key information contained in the initial fields to obtain a second sorting queue; selecting a preset number of initial fields in the second sorting queue to obtain a target field; constructing question and answer prompt words based on the target document, target field and target question.
[0128] 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.
[0129] The device for obtaining recommended questions in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.
[0130] The present disclosure also provides a computer device having the above Figure 6 The device for obtaining the recommendation question is shown.
[0131] See also Figure 7 , Figure 7 is a structural diagram of a computer device provided by an optional embodiment of the present disclosure, such as Figure 7 As shown, the computer device includes: one or more processors 10, memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components utilize different buses to communicate with each other 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 the memory 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. Equally, multiple computer devices can be connected, and each device provides part of the necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 7 A processor 10 is taken as an example.
[0132] 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 an application-specific 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.
[0133] The memory 20 stores instructions that can be executed by at least one processor 10, so as to enable at least one processor 10 to execute the method shown in the above embodiment.
[0134] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created based on the use of the computer device, 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 located 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.
[0135] 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.
[0136] The computer device further includes a communication interface 30 for the computer device to communicate with other devices or a communication network.
[0137] The embodiments of the present disclosure also provide a computer-readable storage medium. The above-mentioned method according to the embodiments of the present disclosure can be implemented in hardware, firmware, or implemented as a computer code that can be recorded in a storage medium, or implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method shown in the above embodiment is implemented.
[0138] A portion of the present disclosure may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present disclosure through the operation of the computer. Those skilled in the art should understand that the form in which the computer program instruction exists in a computer-readable medium includes but is not limited to a source file, an executable file, an installation package file, etc. Accordingly, the way in which the computer program instruction is executed by the computer includes but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium that can be accessed by the computer.
[0139] Although the embodiments of the present disclosure have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present disclosure, and such modifications and variations are all within the scope defined by the appended claims.
Claims
1. A method for obtaining a recommendation question, characterized in that: The method comprises: Get the target question entered in the terminal; In response to the acquisition request of the terminal, performing field extraction and demand identification on the target question to obtain demand keyword information; Searching a preset database based on the required keyword information to obtain initial recall data; Analyzing the initial recall data to generate question recommendation prompt words; Based on the question recommendation prompt words, recommended questions related to the target question are obtained.
2. The method according to claim 1, characterized in that The preset database includes a business database and a data knowledge base. The search is performed in the preset database according to the required keyword information to obtain initial recall data, including: Performing vectorization processing on the demand keyword to obtain a feature vector of the demand keyword; Searching the preset database based on the feature vector to obtain first recalled data; Using a search tool to search a preset database based on the demand keyword information to obtain second recall data; The initial recall data is obtained by combining the first recall data and the second recall data.
3. The method according to claim 2, characterized in that The preset database includes a business database and a data knowledge base, the first recalled data includes a first initial document and a first initial field, and the first recalled data is obtained by searching the preset database based on the feature vector, including: Matching the feature vector with a business data vector in a business database to obtain the first recalled initial document; The feature vector is matched with the knowledge data vector in the data knowledge base to obtain the first initial field after recall.
4. The method according to claim 3, characterized in that The preset database includes a business database and a data knowledge base, the second recalled data includes a second initial document and a second initial field, and the second recalled data is obtained by searching the preset database based on the demand keyword information using a search tool, including: inputting the demand keyword information into the search tool, so that the search tool matches the demand keyword information with the business data in the business database to obtain the recalled second initial document; The demand keyword information is input into the search tool, so that the search tool matches the demand keyword information with the knowledge data in the data knowledge base to obtain the recalled second initial field.
5. The method according to claim 4, characterized in that The initial recall data includes an initial document and an initial field. The initial recall data is obtained by combining the first recall data and the second recall data, including: Performing a union or intersection merge on the first initial document and the second initial document to obtain the initial document; The first initial field and the second initial field are union-merged or intersection-merged to obtain the initial field.
6. The method according to claim 1, characterized in that The step of obtaining a recommended question related to the target question based on the question recommendation prompt word includes: Input the question recommendation prompt words into the large language model and output a recommended question list; Obtain the user's historical questions and preset questions stored in the target system; The questions in the recommended question list are combined with the historical questions and the preset questions to obtain the recommended questions.
7. The method according to claim 6, characterized in that Combining the questions in the recommended question list with the historical questions and the preset questions to obtain the recommended questions includes: Obtaining a first weight corresponding to the recommended question list, a second weight corresponding to the historical question, and a third weight corresponding to the preset question; The recommendation question is obtained based on the first weight, the second weight, and the third weight.
8. The method according to claim 5, characterized in that The method further comprises: generating question-answer prompt words based on the initial recall data; Inputting the question-answer prompt words into a large language model to obtain a target answer corresponding to the target question and an answer source of the target answer; The target question, the target answer, the source of the answer, and the recommended question are sent to the terminal, so that the terminal displays the target question, the target answer, the source of the answer, and the recommended question.
9. The method according to claim 8, characterized in that Generating question-answer prompt words based on the initial recall data includes: sorting the initial documents based on the first key information contained in the initial documents to obtain a first sorting queue; Selecting a preset number of initial documents from the first sorting queue to obtain a target document; sorting the initial fields based on the second key information contained in the initial fields to obtain a second sorting queue; Selecting the preset number of initial fields in the second sorting queue to obtain a target field; The question-and-answer prompt words are constructed according to the target document, the target field, and the target question.
10. A device for obtaining a recommendation question, characterized in that: The device comprises: The acquisition module is used to obtain the target question input in the terminal; A first obtaining module is used to extract fields and identify requirements for the target question in response to an obtaining request from the terminal, and obtain requirement keyword information; A retrieval module, configured to search a preset database based on the required keyword information to obtain initial recall data; A generation module, configured to analyze the initial recall data and generate question recommendation prompt words; The second obtaining module is used to obtain recommended questions related to the target question based on the question recommendation prompt words.
11. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method for obtaining a recommendation question according to any one of claims 1 to 9 by executing the computer instructions.
12. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the method for obtaining recommendation questions according to any one of claims 1 to 9.
13. A computer program product, characterized in that The method comprises computer instructions for causing a computer to execute the method for obtaining a recommendation question according to any one of claims 1 to 9.