Question recommendation method, device and storage medium

Abstract

The present application relates to a method, device and storage medium for question recommendation, the method comprising: obtaining at least one candidate document; for any of the candidate documents, determining the key information of the candidate document, wherein the key information represents the main content of the candidate document; obtaining the target question currently being asked; and recommending questions based on the target question and the key information of at least one of the candidate documents. In this way, question recommendations can be made based on the target question currently being asked and the key information corresponding to each of at least one candidate document, avoiding the recommended questions being limited to a certain topic, and avoiding repeated recommendations when the number of recommendations is too many, thereby improving the user experience.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a question recommendation method, device and storage medium. Background Art

[0002] With the continuous development of big model technology, users can interact with big models through simple natural language to obtain the required knowledge. For this reason, big models have been widely used in document question-and-answer scenarios. Users can quickly understand the content of documents through question-and-answer format without spending a lot of time reading documents.

[0003] However, sometimes users do not know how to ask questions to the big model in order to obtain the required knowledge. Therefore, automatically recommending questions to users has become an urgent need. In the existing related technologies, questions are simply recommended based on the last question and answer, which will result in the recommended questions being limited to a certain topic. When the number of recommendations is too many, repeated recommendations will result. Summary of the invention

[0004] In order to solve the above problem of simply recommending questions based on the last question and answer, which will result in the recommended questions being limited to a certain topic. When the number of recommendations is too many, repeated recommendations will be caused. The embodiment of the present application provides a question recommendation method, device, electronic device and storage medium. The specific technical solution is as follows:

[0005] In a first aspect of an embodiment of the present application, a question recommendation method is first provided, the method comprising:

[0006] Obtain at least one candidate document;

[0007] For any of the candidate documents, determine key information of the candidate document, wherein the key information represents the main content of the candidate document;

[0008] Get the target question currently being asked;

[0009] Question recommendation is performed based on the target question and key information of at least one of the candidate documents.

[0010] In an optional embodiment, the method further comprises:

[0011] Obtaining a current question recommendation mode, and determining whether the question recommendation mode is a first recommendation mode;

[0012] When the question recommendation mode is the first recommendation mode, a plurality of first documents are acquired as candidate documents.

[0013] In an optional embodiment, the method further comprises:

[0014] Acquire a current question recommendation mode, and when the question recommendation mode is a second recommendation mode, acquire a second document as a candidate document;

[0015] When the question recommendation mode is the second recommendation mode, performing question recommendation according to the target question and key information of at least one of the candidate documents includes:

[0016] Filtering at least one first document page in the second document according to the key information, and acquiring first page content of at least one first document page;

[0017] Question recommendation is performed according to the target question and the first page content of at least one of the first document pages.

[0018] In an optional embodiment, the method further comprises:

[0019] Obtaining a current question recommendation mode, and when the question recommendation mode is the third recommendation mode, obtaining historical conversation content, and obtaining a current third document as a candidate document;

[0020] When the question recommendation mode is the third recommendation mode, for any candidate document, determining the key information of the candidate document includes:

[0021] Determining at least one key information of the third document;

[0022] When the question recommendation mode is the third recommendation mode, performing question recommendation according to the target question and key information of at least one of the candidate documents includes:

[0023] According to the target question and the historical conversation content, the current third document is screened to obtain a third screened document;

[0024] Recommending questions based on the third screening document, at least one key information of the third document, and the target question;

[0025] The third document is screened according to the target question and the historical conversation content to obtain a third screened document, including:

[0026] According to the target question and the historical conversation content, the content in the current third document is screened to obtain at least one segment.

[0027] In an optional implementation, when the question recommendation mode is the first recommendation mode, determining the key information of any candidate document includes:

[0028] Select multiple candidate documents, and parse each candidate document in turn to obtain the contents of the multiple candidate documents;

[0029] Summarize the content of each candidate document separately to obtain the key information of the candidate document;

[0030] Among them, the content of each candidate document is summarized, including:

[0031] Selecting at least one third document page from the candidate documents, and acquiring third page content corresponding to each of the at least one third document page;

[0032] Screening at least one of the third document pages according to the third page content to obtain at least one candidate document page;

[0033] For any of the candidate document pages, generating key sub-information of the candidate document page according to the candidate page content of the candidate document page;

[0034] The key information of the candidate document is determined according to the key sub-information corresponding to at least one of the candidate document pages.

[0035] In an optional implementation, when the question recommendation mode is the first recommendation mode, the step of performing question recommendation based on the target question and key information of at least one of the candidate documents includes:

[0036] splicing the key information of the plurality of candidate documents to obtain a key information sequence, and generating a plurality of first candidate questions according to the key information sequence;

[0037] According to the target question, a plurality of the first candidate questions are screened to obtain the screened first candidate questions, and a question recommendation is performed;

[0038] After screening the plurality of first candidate questions according to the target question to obtain the screened first candidate questions, the method further includes:

[0039] For a first document type of any candidate document, searching for a first general question corresponding to the first document type;

[0040] The first candidate question and the first general question that have been screened are used together for question recommendation.

[0041] In an optional implementation, screening a plurality of first candidate questions according to the target question to obtain the screened first candidate questions includes:

[0042] For any of the first candidate questions, determining a first similarity between the first candidate question and the target question;

[0043] When the first similarity is greater than a preset first threshold, discarding the first candidate question;

[0044] When the first similarity is not greater than a preset first threshold, the first candidate question is retained to obtain the screened first candidate question.

[0045] In an optional implementation, the step of recommending the first candidate questions and the first general question after screening together includes:

[0046] Selecting a first target general question from the first general questions, inserting the first target general question into the screened first candidate questions, to obtain a first target question;

[0047] The first target question is recommended.

[0048] In an optional implementation, screening at least one first document page in the second document according to the key information includes:

[0049] For any document page in the second document, counting the number of key information contained in the document page;

[0050] The document pages in the second document are sorted according to the quantity, and at least one first document page in the second document is filtered according to the sorting result.

[0051] In an optional implementation, the recommending a question based on the target question and the first page content of at least one first document page includes:

[0052] In the case of acquiring the first page contents of a plurality of first document pages, splicing the first page contents of the plurality of first document pages to obtain a first content sequence;

[0053] generating a plurality of second candidate questions according to the first content sequence;

[0054] According to the target question, a plurality of the second candidate questions are screened to obtain the screened second candidate questions, and a question recommendation is performed;

[0055] After screening a plurality of the second candidate questions according to the target question to obtain the screened second candidate questions, the method further includes:

[0056] Searching for a second general question corresponding to a second document type of the second document;

[0057] The screened second candidate question and the second general question are used together for question recommendation.

[0058] In an optional implementation, screening a plurality of second candidate questions according to the target question to obtain the screened second candidate questions includes:

[0059] For any of the second candidate questions, determining a second similarity between the second candidate question and the target question;

[0060] When the second similarity is greater than a preset second threshold, discarding the second candidate question;

[0061] When the second similarity is not greater than a preset second threshold, the second candidate question is retained to obtain a screened second candidate question.

[0062] In an optional implementation, the step of recommending the screened second candidate questions and the second general question together includes:

[0063] Selecting a second target general question from the second general questions, inserting the second target general question into the screened second candidate questions, to obtain a second target question;

[0064] The second target question is recommended.

[0065] In an optional implementation, the filtering of the content in the current third document according to the target question and the historical conversation content to obtain at least one fragment includes:

[0066] For any segment in the third document, determining a third similarity between the segment and the target question and the historical conversation content;

[0067] The segments in the third document are sorted according to the third similarity, and the segments in the current third document are screened according to the sorting result to obtain at least one segment.

[0068] In an optional implementation, the recommending questions according to the third screening document, at least one key information of the third document, and the target question includes:

[0069] splicing at least one of the key information and at least one of the fragments to obtain a second content sequence;

[0070] generating a plurality of third candidate questions according to the second content sequence;

[0071] According to the target question, a plurality of the third candidate questions are screened to obtain the screened third candidate questions, and a question recommendation is performed;

[0072] After screening the plurality of third candidate questions according to the target question to obtain the screened third candidate questions, the method further includes:

[0073] Searching for a third general question corresponding to a third document type of the third document;

[0074] The screened third candidate question and the third general question are recommended together.

[0075] In an optional implementation, screening the plurality of third candidate questions according to the target question to obtain the screened third candidate question includes:

[0076] For any of the third candidate questions, determining a fourth similarity between the third candidate question and the target question;

[0077] When the fourth similarity is greater than a preset third threshold, discarding the third candidate question;

[0078] When the fourth similarity is not greater than the preset third threshold, the third candidate question is retained to obtain the screened third candidate question.

[0079] In an optional implementation, the step of recommending the screened third candidate question and the third general question together includes:

[0080] Selecting a third target general question from the third general questions, inserting the third target general question into the screened third candidate question, to obtain a third target question;

[0081] The third target problem is recommended.

[0082] In a second aspect of the embodiments of the present application, a question recommendation device is further provided, the device comprising:

[0083] A document acquisition module is used to acquire a candidate document;

[0084] An information determination module, used for determining key information of any candidate document, wherein the key information represents the main content of the candidate document;

[0085] The question acquisition module is used to obtain the target question currently being asked;

[0086] The question recommendation module is used to recommend questions based on the target question and key information of at least one of the candidate documents.

[0087] In a third aspect of the embodiments of the present application, there is further provided an electronic device, comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus;

[0088] Memory, used to store computer programs;

[0089] The processor is used to implement any of the question recommendation methods described in the first aspect when executing the program stored in the memory.

[0090] In a fourth aspect of the embodiments of the present application, a storage medium is further provided, wherein instructions are stored in the storage medium, and when the storage medium is executed on a computer, the computer executes any of the question recommendation methods described in the first aspect above.

[0091] In a fifth aspect of the embodiments of the present application, a computer program product comprising instructions is also provided, which, when executed on a computer, enables the computer to execute any of the above-mentioned question recommendation methods.

[0092] The technical solution provided by the embodiment of the present application obtains at least one candidate document, determines the key information of the candidate document for any candidate document, wherein the key information represents the main content of the candidate document, obtains the target question currently being asked, and recommends questions based on the target question and the key information of at least one candidate document.

[0093] By obtaining at least one candidate document, determining the key information corresponding to each of the at least one candidate document, and recommending questions based on the target question currently being asked and the key information corresponding to each of the at least one candidate document, question recommendations can be made based on the target question currently being asked and the key information corresponding to each of the at least one candidate document, thereby avoiding the recommended questions being limited to a certain topic, and when the number of recommendations is too many, repeated recommendations can be avoided, thereby improving user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0095] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0096] One or more embodiments are exemplarily described by pictures in the corresponding drawings, and these exemplified descriptions do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0097] Figure 1 A schematic diagram of an implementation flow of a question recommendation method shown in an embodiment of the present application;

[0098] Figure 2 A schematic diagram of an implementation flow of another problem recommendation method shown in an embodiment of the present application;

[0099] Figure 3 A schematic diagram of an implementation flow of another problem recommendation method shown in an embodiment of the present application;

[0100] Figure 4 A schematic diagram of an implementation flow of another problem recommendation method shown in an embodiment of the present application;

[0101] Figure 5 A schematic diagram of an implementation flow of another problem recommendation method shown in an embodiment of the present application;

[0102] Figure 6 A schematic diagram of an implementation flow of another problem recommendation method shown in an embodiment of the present application;

[0103] Figure 7 It is a structural schematic diagram of a question recommendation device shown in an embodiment of the present application;

[0104] Figure 8 It is a schematic diagram of the structure of an electronic device shown in an embodiment of the present application. DETAILED DESCRIPTION

[0105] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0106] The disclosure below provides many different embodiments or examples to realize the different structures of the present application. In order to simplify the disclosure of the present application, the parts and settings of specific examples are described below. Of course, they are only examples, and the purpose is not to limit the present application. In addition, the present application can repeat reference numbers and / or letters in different examples. This repetition is for the purpose of simplification and clarity, and does not itself indicate the relationship between the various embodiments and / or settings discussed.

[0107] like Figure 1 FIG. 1 is a schematic diagram of an implementation flow of a question recommendation method provided in an embodiment of the present application. The method is applied to an electronic device and may specifically include the following steps:

[0108] S101: Obtain at least one candidate document.

[0109] In the embodiment of the present application, at least one candidate document is obtained, wherein the candidate document may be a variety of documents. For example, N candidate documents are obtained, where N is a positive integer and is ≥1.

[0110] S102: for any candidate document, determine key information of the candidate document, wherein the key information represents the main content of the candidate document.

[0111] In an embodiment of the present application, for at least one candidate document obtained above, for any candidate document, key information of the candidate document is determined, wherein the key information represents the main content of the candidate document, that is, represents what the document content of the candidate document mainly talks about. The key information may be a keyword, a summary of the document, or a core page of the document.

[0112] S103, obtaining the target question currently being asked.

[0113] In an embodiment of the present application, the user is currently asking a question, thereby obtaining the target question currently asked by the user. For example, the target question may be "What quality changes occur after adding dry ice to bulk braised chicken feet?"

[0114] S104: Recommend a question based on the target question and key information of at least one candidate document.

[0115] In an embodiment of the present application, for the target question currently being asked and the key information corresponding to at least one candidate document, question recommendations can be made based on the target question and the key information corresponding to at least one candidate document.

[0116] In this way, question recommendations are made based on the target question currently being asked and the key information corresponding to at least one candidate document, avoiding the recommended questions being limited to a certain topic. When the number of recommendations is too many, repeated recommendations can be avoided, thereby improving the user experience.

[0117] In addition, in the embodiment of the present application, the question recommendation scenario involves a multi-document question-answering scenario, that is, the question recommendation in the multi-document question-answering scenario involves multiple documents, and the questions have clear directions (each question comes from a different document). To this end, a question recommendation method is provided for the question recommendation in the multi-document question-answering scenario, that is, Figure 2 FIG. 1 is a schematic diagram of an implementation flow of another question recommendation method provided in an embodiment of the present application. The method is applied to an electronic device and may specifically include the following steps:

[0118] S201, obtaining the current question recommendation mode, and determining whether the question recommendation mode is the first recommendation mode.

[0119] S202: When the question recommendation mode is the first recommendation mode, obtain a plurality of first documents as candidate documents.

[0120] In an embodiment of the present application, the current question recommendation mode is obtained, and it is determined whether the question recommendation mode is the first recommendation mode. When the question recommendation mode is the first recommendation mode, it indicates that the current scenario is a multi-document question and answer scenario. In this case, multiple first documents are obtained as candidate documents.

[0121] It should be noted that a mode selection button can be provided, and by triggering the mode selection button, a corresponding recommendation mode can be selected. For example, the mode selection button "Recommendation Mode 1" corresponds to the first recommendation mode, so when the user clicks "Recommendation Mode 1", it means that the current recommendation mode is changed to the first recommendation mode, and the user wants to recommend questions in a multi-document question-and-answer scenario, and multiple first documents are obtained as candidate documents. Of course, the first recommendation mode can also be triggered by selecting multiple first documents, and this embodiment of the application is not limited to this.

[0122] S203, selecting multiple candidate documents, and parsing each candidate document in turn to obtain contents of the multiple candidate documents.

[0123] In an embodiment of the present application, multiple candidate documents are selected from the candidate documents obtained above, and each candidate document is parsed in turn to obtain the content corresponding to each of the multiple candidate documents.

[0124] S204, summarizing the content of each candidate document to obtain key information of the candidate document.

[0125] In the embodiment of the present application, the content of each candidate document is summarized to obtain key information corresponding to each candidate document, wherein the key information is the document summary of the candidate document.

[0126] Specifically, the key information of each candidate document can be determined in the following ways:

[0127] For any candidate document, at least one third document page is selected from the candidate document, for example, N third document pages are randomly selected, where N≥1. The third page content corresponding to each of the at least one third document page is obtained.

[0128] For at least one third document page, it is necessary to judge the quality of the third page content. To this end, at least one third document page is screened according to the third page content to obtain at least one candidate document page.

[0129] Among them, for any third document page, if the third page content of the third document page meets at least one of the following conditions, the third document page is retained, otherwise the third document page is discarded, and quality judgment is performed in this way, at least one third document page is screened, and at least one candidate document page is obtained.

[0130] 1. The content of the third page contains the outline title structure; 2. The content of the third page contains specific formats, such as "Abstract", "Research Methods" and other specific formats in the paper; 3. The number of words in the content of the third page is greater than a certain number.

[0131] For any of the candidate document pages, key sub-information of the candidate document page is generated according to the candidate page content of the candidate document page, wherein the key sub-information of the candidate document page can be generated with the aid of a large model.

[0132] Specifically, the candidate page contents of the candidate document pages are spliced ​​together according to fragments (such as paragraphs) to form prompt words, which are then input into the large model to generate key sub-information of the candidate document pages, thereby obtaining the key sub-information corresponding to each candidate document page.

[0133] It should be noted that the key sub-information is the page summary of the candidate document page, that is, it represents what the candidate page content of the candidate document page specifically expresses, and the embodiment of the present application does not limit this.

[0134] After the above processing, the key sub-information corresponding to each candidate document page is obtained, and the key information of the candidate document can be determined according to the key sub-information corresponding to each candidate document page. The key information of the candidate document can be determined with the help of a large model.

[0135] Specifically, the key sub-information corresponding to each candidate document page is concatenated to form a prompt word, which is then input into the large model to generate the key information of the candidate document, thereby obtaining the key information of each candidate document.

[0136] S205, obtaining the target question currently being asked.

[0137] In the embodiment of the present application, this step is similar to the above-mentioned step S103, and the embodiment of the present application will not be described one by one here.

[0138] S206, concatenating key information of multiple candidate documents to obtain a key information sequence, and generating multiple first candidate questions based on the key information sequence.

[0139] S207, screening multiple first candidate questions according to the target question, obtaining screened first candidate questions, and performing question recommendation.

[0140] In an embodiment of the present application, after the above steps, key information corresponding to each of the multiple candidate documents can be obtained, and the key information corresponding to each of the multiple candidate documents can be spliced ​​to obtain a key information sequence, and multiple first candidate questions can be generated based on the key information sequence.

[0141] Among them, multiple first candidate questions can be generated with the help of the big model. To this end, the key information sequence is assembled into prompt words, and the prompt words are input into the big model, and the big model generates multiple first candidate questions. Among them, the prompt words are used to guide the big model to generate the required content step by step according to the requirements.

[0142] For multiple first candidate questions, they need to be screened according to the target question currently being asked, and the screened first candidate questions are obtained to recommend questions. The purpose of this is to avoid duplication of the recommended questions with the target question currently being asked by the user.

[0143] Specifically, for any first candidate question, the first similarity between the first candidate question and the target question is determined. If the first similarity is greater than a preset first threshold, the first candidate question is discarded. If the first similarity is not greater than the preset first threshold, the first candidate question is retained to obtain the screened first candidate question.

[0144] In addition, for the first candidate questions after screening, the similarity can be calculated pairwise. If the similarity is greater than a certain threshold, one of them can be discarded. In this way, the first candidate questions after screening can be screened again and question recommendations can be made. The purpose of doing this is to avoid duplication between recommended questions.

[0145] In addition, according to the target question currently being asked, it is screened to obtain the first candidate question after screening, and after the question recommendation, the first general question corresponding to the first document type (the first document type can be, for example, a paper, a contract, an official document, etc.) of any candidate document can be searched, and then the first candidate question after screening and the first general question are recommended together. Among them, the first target general question can be selected from the first general question, and the first target general question can be inserted into the first candidate question after screening to obtain the first target question, and the first target question is recommended.

[0146] It should be noted that, for the first general question corresponding to the first document type of each candidate document, a first target general question can be randomly selected and randomly inserted into the screened first candidate question to obtain the first target question and recommend the first target question, which can increase the globality and diversity of the recommended questions.

[0147] In addition, before inserting the first target general question into the first candidate question after screening, the first candidate question after screening can also be subjected to some additional processing, such as removing questions that are declarative sentences, removing questions with fewer words, removing questions without a clear subject, etc., which is not limited to the embodiments of the present application.

[0148] In this way, in a multi-document question-and-answer scenario, question recommendations can be made based on the target question currently being asked, as well as the key information and first document type corresponding to multiple candidate documents, to avoid the recommended questions being limited to a certain topic. When the number of recommendations is too many, repeated recommendations can also be avoided, thereby improving the user experience.

[0149] In addition, the single-document question-answering scenario corresponds to the multi-document question-answering scenario. In the single-document question-answering scenario, it is also necessary to avoid limiting the recommended questions to a certain topic, and to avoid repeated recommendations when there are too many recommendations, so as to improve the user experience. Figure 3 FIG. 1 is a schematic diagram of an implementation flow of another question recommendation method provided in an embodiment of the present application. The method is applied to an electronic device and may specifically include the following steps:

[0150] S301, obtaining a current question recommendation mode, and when the question recommendation mode is a second recommendation mode, obtaining a second document as a candidate document.

[0151] In an embodiment of the present application, for the current question recommendation mode, when the question recommendation mode is the second recommendation mode, it means that the current scenario is a single document question and answer scenario, and the second document can be obtained as a candidate document.

[0152] It should be noted that the second recommendation mode can be triggered by using a button or the like, or by selecting a single document or enabling question recommendations based on the currently open document, and this embodiment of the present application does not limit this.

[0153] S302: for any candidate document, determine key information of the candidate document, wherein the key information represents the main content of the candidate document.

[0154] In an embodiment of the present application, for a candidate document, that is, the second document mentioned above, key information of the second document may be determined, wherein the key information represents the main content of the second document, and the key information may be a keyword or a key word, etc.

[0155] Among them, for the extraction of key information, the existing technical solutions can be used, and the key information is usually the more important content in the second document, which can describe the subject content of the second document. The embodiments of the present application are not limited to this.

[0156] S303, obtaining the target question currently being asked.

[0157] In the embodiment of the present application, this step is similar to the above-mentioned step S103, and the embodiment of the present application will not be described one by one here.

[0158] S304: Filter at least one first document page in the second document according to the key information, and obtain first page content of the at least one first document page.

[0159] In an embodiment of the present application, for the key information extracted in the above steps, the problem pages in the second document can be screened according to the key information to obtain at least one first document page in the second document, and the first page content of at least one first document page can be obtained.

[0160] It should be noted that the purpose of screening the question pages in the second document according to the key information is essentially to obtain document pages with higher content quality, so as to facilitate subsequent accurate question recommendation.

[0161] Specifically, for any document page in the second document, the number of key information contained in the document page is counted, the document pages in the second document are sorted according to the number, and at least one first document page in the second document is screened according to the sorting result.

[0162] The document pages in the second document may be sorted from large to small according to the quantity, and then the top N document pages are selected according to the sorting result as at least N first document pages in the second document, where N is a positive integer and ≥1.

[0163] S305: Recommend questions based on the target question and the first page content of at least one first document page.

[0164] In an embodiment of the present application, after the above steps, the target question currently being asked and the first page content of at least one first document page can be obtained, thereby making question recommendations based on the target question and the first page content of at least one first document page.

[0165] For the target question and the first page content of at least one first document page, reference may be made to: Figure 4 The method shown in Figure 1 implements question recommendation in a single document question answering scenario. Figure 4 FIG. 1 is a schematic diagram of an implementation flow of another question recommendation method provided in an embodiment of the present application. The method is applied to an electronic device and may specifically include the following steps:

[0166] S401, when first page contents of a plurality of first document pages are acquired, the first page contents of the plurality of first document pages are spliced ​​to obtain a first content sequence.

[0167] S402: Generate a plurality of second candidate questions according to the first content sequence.

[0168] S403, screening multiple second candidate questions according to the target question, obtaining screened second candidate questions, and performing question recommendation.

[0169] In an embodiment of the present application, after the above steps, the first page contents of multiple first document pages can be obtained, the first page contents of multiple first document pages can be spliced ​​to obtain a first content sequence, and multiple second candidate questions can be generated based on the first content sequence.

[0170] The large model can be used to generate multiple second candidate questions. To this end, the first content sequence is assembled into prompt words, which are input into the large model, and the large model generates multiple second candidate questions. The prompt words are used to guide the large model to generate the required content step by step according to the requirements.

[0171] For multiple second candidate questions, they need to be screened according to the target question currently being asked, and the screened second candidate questions are obtained to recommend questions. The purpose of this is to avoid duplication of the recommended questions with the target question currently being asked by the user.

[0172] Specifically, for any second candidate question, the second similarity between the second candidate question and the target question is determined. If the second similarity is greater than a preset second threshold, the second candidate question is discarded. If the second similarity is not greater than the preset second threshold, the second candidate question is retained to obtain a screened second candidate question.

[0173] In addition, for the second candidate questions after screening, the similarity can be calculated two by two. If the similarity is greater than a certain threshold, one of them can be discarded. In this way, the second candidate questions after screening can be screened again. The purpose of doing this is to avoid duplication between recommended questions.

[0174] In addition, after screening the target question currently being asked to obtain the screened second candidate question and recommending the question, the second general question corresponding to the second document type can be searched, and then the screened second candidate question and the second general question can be recommended together. Among them, the second target general question can be selected from the second general question, and the second target general question can be inserted into the screened second candidate question to obtain the second target question, and the second target question can be recommended.

[0175] It should be noted that, for the second general question corresponding to the second document type of the second document, a second target general question can be randomly selected therefrom and randomly inserted into the screened second candidate question to obtain the second target question and recommend the second target question, thereby increasing the globality and diversity of the recommended questions.

[0176] In addition, before inserting the second target general question into the filtered second candidate question, the filtered second candidate question can also be subjected to some additional processing, such as removing questions that are declarative sentences, removing questions with fewer words, removing questions that do not have a clear subject, etc., which is not limited to the embodiments of the present application.

[0177] In this way, in a single-document question-and-answer scenario, question recommendations can be made based on the target question currently being asked, the first page content of at least one first document page in the second document, and the second document type of the second document, thereby avoiding the recommended questions being limited to a certain topic. When the number of recommendations is too many, repeated recommendations can also be avoided, thereby improving the user experience.

[0178] In addition, the existing technology simply recommends questions based on the last Q&A, which will result in the recommended questions being limited to a certain topic, and when the number of recommendations is too many, repeated recommendations will result, and users will not be able to understand the global content of the document through the recommended questions. To this end, the embodiment of the present invention recommends questions based on the last Q&A and the document content of the current document, so that there are both questions related to the last Q&A and global questions, and they are no longer limited to a certain topic. Based on this, Figure 5 FIG. 1 is a schematic diagram of an implementation flow of another question recommendation method provided in an embodiment of the present application. The method is applied to an electronic device and may specifically include the following steps:

[0179] S501, obtaining the current question recommendation mode, and when the question recommendation mode is the third recommendation mode, obtaining historical conversation content, and obtaining the current third document as a candidate document.

[0180] In an embodiment of the present application, for the current question recommendation mode, when the question recommendation mode is the third recommendation mode, it means that the current scene is a contextual question and answer scene. At this time, the historical conversation content and the current third document are obtained as candidate documents.

[0181] It should be noted that the historical conversation content may refer to the question asked by the user last time and the answer to the question asked by the user last time, and the embodiment of the present application does not limit this.

[0182] In addition, the third recommendation mode may be triggered by a button or the like, or may be triggered based on enabling question recommendations based on the currently opened document, which is not limited in this embodiment of the present application.

[0183] S502: Determine at least one key information of the third document, wherein the key information represents the main content of the third document.

[0184] In an embodiment of the present application, for the third document obtained above, at least one key information of the third document is determined, wherein the key information represents the main content of the third document. The key information refers to the page content of at least one document page in the third document.

[0185] Based on this, for the third document, at least one second document page may be selected from the third document, and the second page content of the at least one second document page may be obtained as at least one key information of the third document.

[0186] S503, obtaining the target question currently being asked.

[0187] In the embodiment of the present application, this step is similar to the above-mentioned step S103, and the embodiment of the present application will not be described one by one here.

[0188] S504: Screen the current third document according to the target question and the historical conversation content to obtain a third screened document.

[0189] In the embodiment of the present application, for the target question currently asked and the historical conversation content, the current third document can be screened according to the target question and the historical conversation content to obtain a third screened document, wherein the third screened document is a certain segment in the third document.

[0190] Based on this, for the target question currently being asked and the historical conversation content, the content in the current third document can be filtered according to the target question and the historical conversation content to obtain at least one fragment (wherein a fragment can be a paragraph, or multiple paragraphs, or a table, or a collection of paragraphs and tables. In the embodiments of the present application, a paragraph is taken as an example for illustration).

[0191] Specifically, for any fragment in the current third document, the similarity between the fragment and the target question and the historical conversation content is determined, and the fragments in the current third document are sorted according to the similarity. According to the sorting result, at least one fragment in the third document is screened and obtained.

[0192] Among them, the fragments in the current third document can be sorted from large to small according to the similarity, and then the first K fragments are selected according to the sorting result as at least K fragments in the third document, where K is a positive integer, for example, K can be 5.

[0193] S505: Recommend questions based on the third screening document, at least one key information of the third document, and the target question.

[0194] In an embodiment of the present application, after the above steps, the target question currently being asked, the third screening document, and at least one key information in the third document can be obtained, and question recommendations can be made based on the target question, the third screening document, and at least one key information in the third document.

[0195] For the target question, the third screening document, and at least one key information in the third document, reference may be made to: Figure 6 The method shown in Figure 1 implements question recommendation in contextual question answering scenarios. Figure 6 FIG. 1 is a schematic diagram of an implementation flow of another question recommendation method provided in an embodiment of the present application. The method is applied to an electronic device and may specifically include the following steps:

[0196] S601: splice at least one key information and at least one segment to obtain a second content sequence.

[0197] S602: Generate a plurality of third candidate questions according to the second content sequence.

[0198] S603, screening multiple third candidate questions according to the target question, obtaining screened third candidate questions, and performing question recommendation.

[0199] In an embodiment of the present application, after the above steps, at least one key information and at least one fragment in the third document can be obtained, the key information and at least one fragment in the third document are spliced ​​to obtain a second content sequence, and multiple third candidate questions are generated based on the second content sequence.

[0200] The large model can be used to generate multiple third candidate questions. To this end, the second content sequence is composed of prompt words, which are input into the large model, and the large model generates multiple third candidate questions. The prompt words are used to guide the large model to generate the required content step by step according to the requirements.

[0201] For multiple third candidate questions, they need to be screened according to the target question currently being asked, and the screened third candidate questions are obtained to recommend questions. The purpose of this is to avoid duplication of the recommended questions with the target question currently being asked by the user.

[0202] Specifically, for any third candidate question, the fourth similarity between the third candidate question and the target question is determined. If the fourth similarity is greater than a preset third threshold, the third candidate question is discarded. If the fourth similarity is not greater than the preset third threshold, the third candidate question is retained to obtain a screened third candidate question.

[0203] In addition, for the third candidate questions after screening, the similarity can be calculated two by two. If the similarity is greater than a certain threshold, one of them can be discarded. In this way, the third candidate questions after screening can be screened again. The purpose of doing this is to avoid duplication between recommended questions.

[0204] In addition, after screening the target question currently being asked to obtain the screened third candidate question and recommending the question, the third general question corresponding to the third document type can be searched, and then the screened third candidate question and the third general question can be recommended together. Among them, the third target general question can be selected from the third general question, and the third target general question can be inserted into the screened third candidate question to obtain the third target question, and the third target question can be recommended.

[0205] It should be noted that, for the third general question corresponding to the third document type of the third document, a third target general question can be randomly selected therefrom, and the third target general question can be randomly inserted into the screened third candidate question to obtain the third target question, and the third target question is recommended, which can increase the globality and diversity of the recommended questions.

[0206] In addition, before inserting the third target general question into the filtered third candidate question, the filtered third candidate question can also be subjected to some additional processing, such as removing questions that are declarative sentences, removing questions with fewer words, removing questions that do not have a clear subject, etc., which is not limited to the embodiments of the present application.

[0207] In this way, question recommendations are made based on the previous question and answer and the content of the current document. This way, there are both questions related to the previous question and answer and global questions, and the questions are no longer limited to a certain topic, thereby improving the user experience.

[0208] Corresponding to the above method embodiment, the present application embodiment also provides a question recommendation device, such as Figure 7 As shown, the device may include: a document acquisition module 710, an information determination module 720, a question acquisition module 730, and a question recommendation module 740.

[0209] A document acquisition module 710, configured to acquire at least one candidate document;

[0210] An information determination module 720, configured to determine key information of any candidate document, wherein the key information represents the main content of the candidate document;

[0211] The question acquisition module 730 is used to acquire the target question currently being asked;

[0212] The question recommendation module 740 is used to recommend questions based on the target question and key information of at least one of the candidate documents.

[0213] In an optional implementation, the device further includes a first candidate document acquisition module, specifically configured to:

[0214] Obtaining a current question recommendation mode, and determining whether the question recommendation mode is a first recommendation mode;

[0215] When the question recommendation mode is the first recommendation mode, a plurality of first documents are acquired as candidate documents.

[0216] In an optional embodiment, the device further comprises:

[0217] A second candidate document acquisition module, used to acquire the current question recommendation mode, and when the question recommendation mode is the second recommendation mode, acquire the second document as a candidate document;

[0218] When the question recommendation mode is the second recommendation mode, the question recommendation module specifically includes:

[0219] A first page screening submodule, configured to screen at least one first document page in the second document according to the key information;

[0220] A first content acquisition submodule, used to acquire first page content of at least one first document page;

[0221] The first question recommendation submodule is used to recommend questions according to the target question and the first page content of at least one of the first document pages.

[0222] In an optional embodiment, the device further comprises:

[0223] A third candidate document acquisition module, used to acquire historical conversation content and acquire the current third document as a candidate document when the question recommendation mode is the third recommendation mode;

[0224] When the question recommendation mode is the third recommendation mode, the information determination module is specifically used to:

[0225] Determining at least one key information of the third document;

[0226] When the question recommendation mode is the second recommendation mode, the question recommendation module specifically includes:

[0227] A third document screening submodule, configured to screen the current third document according to the target question and the historical conversation content to obtain a third screened document;

[0228] A second question recommendation submodule, configured to recommend questions based on the third screening document, at least one key information of the third document, and the target question;

[0229] Wherein, the third document screening submodule is specifically used for:

[0230] According to the target question and the historical conversation content, the content in the current third document is screened to obtain at least one segment.

[0231] In an optional implementation, when the question recommendation mode is the first recommendation mode, the information determination module is specifically configured to:

[0232] Select multiple candidate documents, and parse each candidate document in turn to obtain the contents of the multiple candidate documents;

[0233] Summarize the content of each candidate document separately to obtain the key information of the candidate document;

[0234] Among them, the content of each candidate document is summarized, including:

[0235] Selecting at least one third document page from the candidate documents, and acquiring third page content corresponding to each of the at least one third document page;

[0236] Screening at least one of the third document pages according to the third page content to obtain at least one candidate document page;

[0237] For any of the candidate document pages, generating key sub-information of the candidate document page according to the candidate page content of the candidate document page;

[0238] The key information of the candidate document is determined according to the key sub-information corresponding to at least one of the candidate document pages.

[0239] In an optional implementation, when the question recommendation mode is the first recommendation mode, the question recommendation module specifically includes:

[0240] A key information splicing submodule, used for splicing the key information of a plurality of candidate documents to obtain a key information sequence;

[0241] A first candidate question generating submodule, used for generating a plurality of first candidate questions according to the key information sequence;

[0242] A third question recommendation submodule is used to screen the plurality of first candidate questions according to the target question, obtain the screened first candidate questions, and recommend questions;

[0243] The question recommendation module also includes:

[0244] A first general question search submodule, used for searching, for a first document type of any candidate document, a first general question corresponding to the first document type;

[0245] The fourth question recommendation submodule is used to recommend questions together with the first candidate questions and the first general questions after screening.

[0246] In an optional implementation, the third question recommendation submodule is specifically used for:

[0247] For any of the first candidate questions, determining a first similarity between the first candidate question and the target question;

[0248] When the first similarity is greater than a preset first threshold, discarding the first candidate question;

[0249] When the first similarity is not greater than a preset first threshold, the first candidate question is retained to obtain the screened first candidate question.

[0250] In an optional implementation, the fourth question recommendation submodule is specifically used for:

[0251] Selecting a first target general question from the first general questions, inserting the first target general question into the screened first candidate questions, to obtain a first target question;

[0252] The first target question is recommended.

[0253] In an optional implementation, the first page screening submodule is specifically used to:

[0254] For any document page in the second document, counting the number of key information contained in the document page;

[0255] The document pages in the second document are sorted according to the quantity, and at least one first document page in the second document is filtered according to the sorting result.

[0256] In an optional implementation, the first question recommendation submodule specifically includes:

[0257] A first page content splicing unit is used to splice the first page contents of a plurality of first document pages to obtain a first content sequence when the first page contents of a plurality of first document pages are acquired;

[0258] A second candidate question generating unit, configured to generate a plurality of second candidate questions according to the first content sequence;

[0259] A second candidate question screening unit, configured to screen a plurality of second candidate questions according to the target question, obtain the screened second candidate questions, and recommend questions;

[0260] The first question recommendation submodule further includes:

[0261] A second general question searching unit, configured to search for a second general question corresponding to a second document type of the second document;

[0262] The first question recommendation unit is used to recommend the screened second candidate question and the second general question together.

[0263] In an optional implementation, the second candidate question screening unit is specifically configured to:

[0264] For any of the second candidate questions, determining a second similarity between the second candidate question and the target question;

[0265] When the second similarity is greater than a preset second threshold, discarding the second candidate question;

[0266] When the second similarity is not greater than a preset second threshold, the second candidate question is retained to obtain a screened second candidate question.

[0267] In an optional implementation, the first question recommendation unit is specifically configured to:

[0268] Selecting a second target general question from the second general questions, inserting the second target general question into the screened second candidate questions, to obtain a second target question;

[0269] The second target question is recommended.

[0270] In an optional implementation, the third document screening submodule is specifically used to:

[0271] For any segment in the third document, determining a third similarity between the segment and the target question and the historical conversation content;

[0272] The segments in the third document are sorted according to the third similarity, and the segments in the current third document are screened according to the sorting result to obtain at least one segment.

[0273] In an optional implementation, the second question recommendation submodule specifically includes:

[0274] A second page content splicing unit, used for splicing at least one of the key information and at least one of the fragments to obtain a second content sequence;

[0275] A third candidate question generating unit, configured to generate a plurality of third candidate questions according to the second content sequence;

[0276] A third candidate question screening unit, configured to screen a plurality of the third candidate questions according to the target question, obtain the screened third candidate questions, and perform question recommendation;

[0277] The second question recommendation submodule further includes:

[0278] A third general question searching unit, configured to search for a third general question corresponding to a third document type of the third document;

[0279] The second question recommendation unit is used to recommend the screened third candidate question and the third general question together.

[0280] In an optional implementation, the third candidate question screening unit is specifically configured to:

[0281] For any of the third candidate questions, determining a fourth similarity between the third candidate question and the target question;

[0282] When the fourth similarity is greater than a preset third threshold, discarding the third candidate question;

[0283] When the fourth similarity is not greater than the preset third threshold, the third candidate question is retained to obtain the screened third candidate question.

[0284] In an optional implementation, the second question recommendation unit is specifically configured to:

[0285] Selecting a third target general question from the third general questions, inserting the third target general question into the screened third candidate question, to obtain a third target question;

[0286] The third target problem is recommended.

[0287] The present application also provides an electronic device, such as Figure 8 As shown, it includes a processor 81, a communication interface 82, a memory 83 and a communication bus 84, wherein the processor 81, the communication interface 82, and the memory 83 communicate with each other through the communication bus 84.

[0288] A memory 83, for storing computer programs;

[0289] The processor 81 is used to execute the program stored in the memory 83 to implement the following steps:

[0290] Obtain at least one candidate document; for any of the candidate documents, determine key information of the candidate document, wherein the key information represents the main content of the candidate document; obtain a target question currently being asked; and recommend questions based on the target question and key information of at least one of the candidate documents.

[0291] The communication bus mentioned in the above electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0292] The communication interface is used for communication between the above electronic device and other devices.

[0293] The memory may include a random access memory (RAM) or a non-volatile memory, such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0294] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0295] In another embodiment provided by the present application, a storage medium is also provided, in which instructions are stored. When the instructions are executed on a computer, the computer executes any one of the question recommendation methods in the above embodiments.

[0296] In another embodiment provided by the present application, a computer program product including instructions is also provided, which, when executed on a computer, enables the computer to execute any one of the question recommendation methods in the above embodiments.

[0297] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a storage medium, or transmitted from one storage medium to another storage medium, for example, the computer instructions may be transmitted from a website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The storage medium may be any available medium that a computer can access or a data storage device such as a server or a data center that includes one or more available media integrations. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive Solid State Disk (SSD)), etc.

[0298] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0299] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0300] The above description is only a preferred embodiment of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application are included in the protection scope of the present application.

Claims

1. A question recommendation method, characterized in that: The method comprises: Obtain at least one candidate document; For any of the candidate documents, determine key information of the candidate document, wherein the key information represents the main content of the candidate document; Get the target question currently being asked; Question recommendation is performed based on the target question and key information of at least one of the candidate documents.

2. The method according to claim 1, characterized in that The method further comprises: Obtaining a current question recommendation mode, and determining whether the question recommendation mode is a first recommendation mode; When the question recommendation mode is the first recommendation mode, a plurality of first documents are acquired as candidate documents.

3. The method according to claim 1, characterized in that The method further comprises: Acquire a current question recommendation mode, and when the question recommendation mode is a second recommendation mode, acquire a second document as a candidate document; When the question recommendation mode is the second recommendation mode, performing question recommendation according to the target question and key information of at least one of the candidate documents includes: Filtering at least one first document page in the second document according to the key information, and acquiring first page content of at least one first document page; Question recommendation is performed according to the target question and the first page content of at least one of the first document pages.

4. The method according to claim 1, characterized in that: The method further comprises: Acquire the current question recommendation mode, and when the question recommendation mode is the third recommendation mode, acquire the historical conversation content, and acquire the current third document as a candidate document; When the question recommendation mode is the third recommendation mode, for any candidate document, determining the key information of the candidate document includes: Determining at least one key information of the third document; When the question recommendation mode is the third recommendation mode, performing question recommendation according to the target question and key information of at least one of the candidate documents includes: According to the target question and the historical conversation content, the current third document is screened to obtain a third screened document; Recommending questions based on the third screening document, at least one key information of the third document, and the target question; The third document is screened according to the target question and the historical conversation content to obtain a third screened document, including: According to the target question and the historical conversation content, the content in the current third document is screened to obtain at least one segment.

5. The method according to claim 2, characterized in that: When the question recommendation mode is the first recommendation mode, determining the key information of any candidate document includes: Select multiple candidate documents, and parse each candidate document in turn to obtain the contents of the multiple candidate documents; Summarize the content of each candidate document separately to obtain the key information of the candidate document; Among them, the content of each candidate document is summarized, including: Selecting at least one third document page from the candidate documents, and acquiring third page content corresponding to each of the at least one third document page; Screening at least one of the third document pages according to the third page content to obtain at least one candidate document page; For any of the candidate document pages, generating key sub-information of the candidate document page according to the candidate page content of the candidate document page; The key information of the candidate document is determined according to the key sub-information corresponding to at least one of the candidate document pages.

6. The method according to claim 2, characterized in that When the question recommendation mode is the first recommendation mode, performing question recommendation according to the target question and key information of at least one of the candidate documents includes: splicing the key information of the plurality of candidate documents to obtain a key information sequence, and generating a plurality of first candidate questions according to the key information sequence; According to the target question, a plurality of the first candidate questions are screened to obtain the screened first candidate questions, and a question recommendation is performed; After screening the plurality of first candidate questions according to the target question to obtain the screened first candidate questions, the method further includes: For a first document type of any candidate document, searching for a first general question corresponding to the first document type; The first candidate question and the first general question that have been screened are used together for question recommendation.

7. The method according to claim 6, characterized in that The step of screening the plurality of first candidate questions according to the target question to obtain the screened first candidate questions includes: For any of the first candidate questions, determining a first similarity between the first candidate question and the target question; When the first similarity is greater than a preset first threshold, discarding the first candidate question; When the first similarity is not greater than a preset first threshold, the first candidate question is retained to obtain the screened first candidate question.

8. The method according to claim 6, characterized in that The step of recommending the first candidate question and the first general question after screening includes: Selecting a first target general question from the first general questions, inserting the first target general question into the screened first candidate questions, to obtain a first target question; The first target question is recommended.

9. The method according to claim 3, characterized in that: The step of screening at least one first document page in the second document according to the key information includes: For any document page in the second document, counting the number of key information contained in the document page; The document pages in the second document are sorted according to the quantity, and at least one first document page in the second document is filtered according to the sorting result.

10. The method according to claim 3, characterized in that The recommending a question according to the target question and the first page content of at least one first document page includes: In the case of acquiring the first page contents of a plurality of first document pages, splicing the first page contents of the plurality of first document pages to obtain a first content sequence; generating a plurality of second candidate questions according to the first content sequence; According to the target question, a plurality of the second candidate questions are screened to obtain the screened second candidate questions, and a question is recommended; After screening a plurality of the second candidate questions according to the target question to obtain the screened second candidate questions, the method further includes: Searching for a second general question corresponding to a second document type of the second document; The screened second candidate question and the second general question are used together for question recommendation.

11. The method according to claim 10, characterized in that The step of screening a plurality of the second candidate questions according to the target question to obtain the screened second candidate questions includes: For any of the second candidate questions, determining a second similarity between the second candidate question and the target question; When the second similarity is greater than a preset second threshold, discarding the second candidate question; When the second similarity is not greater than a preset second threshold, the second candidate question is retained to obtain a screened second candidate question.

12. The method according to claim 10, characterized in that The step of recommending the screened second candidate question and the second general question together includes: Selecting a second target general question from the second general questions, inserting the second target general question into the screened second candidate questions, to obtain a second target question; The second target question is recommended.

13. The method according to claim 4, characterized in that The step of screening the content in the current third document according to the target question and the historical conversation content to obtain at least one fragment includes: For any segment in the third document, determining a third similarity between the segment and the target question and the historical conversation content; The segments in the third document are sorted according to the third similarity, and the segments in the current third document are screened according to the sorting result to obtain at least one segment.

14. The method according to claim 4, characterized in that The step of recommending a question based on the third screening document, at least one key information of the third document, and the target question includes: splicing at least one of the key information and at least one of the fragments to obtain a second content sequence; generating a plurality of third candidate questions according to the second content sequence; According to the target question, a plurality of the third candidate questions are screened to obtain the screened third candidate questions, and a question recommendation is performed; After screening the plurality of third candidate questions according to the target question to obtain the screened third candidate questions, the method further includes: Searching for a third general question corresponding to a third document type of the third document; The screened third candidate question and the third general question are recommended together.

15. The method according to claim 14, characterized in that The step of screening the plurality of third candidate questions according to the target question to obtain the screened third candidate questions includes: For any of the third candidate questions, determining a fourth similarity between the third candidate question and the target question; When the fourth similarity is greater than a preset third threshold, discarding the third candidate question; When the fourth similarity is not greater than the preset third threshold, the third candidate question is retained to obtain the screened third candidate question.

16. The method according to claim 14, characterized in that The step of recommending the screened third candidate question and the third general question together includes: Selecting a third target general question from the third general questions, inserting the third target general question into the screened third candidate question, to obtain a third target question; The third target problem is recommended.

17. A question recommendation device, characterized in that: The device comprises: A document acquisition module, used to acquire at least one candidate document; An information determination module, used for determining key information of any candidate document, wherein the key information represents the main content of the candidate document; The question acquisition module is used to obtain the target question currently being asked; The question recommendation module is used to recommend questions based on the target question and key information of at least one of the candidate documents.

18. A storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 16 is implemented.

Images