Question and answer information determination method and device, computer device, and storage medium

Abstract

The application relates to a question and answer information determination method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring current question information of a user, determining whether the current question information has unique answer information; if the current question information does not have unique answer information, determining a target answer range corresponding to the current question information based on an initial answer range corresponding to the current question information and an answer range corresponding to a preposed question; determining a candidate answer corresponding to the current question information based on the target answer range; performing segment division on the candidate answer to obtain a plurality of candidate answer segments; screening the plurality of candidate answer segments based on whether each candidate answer segment appears in a plurality of rounds of question and answer processes before the current question information to obtain screened answer segments; determining answer information corresponding to the current question information based on the screened answer segments, and outputting the answer information. The method can improve the accuracy of a question and answer result.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, apparatus, computer device, and storage medium for determining question-and-answer information. Background Technology

[0002] With the development of artificial intelligence technology, intelligent question answering has been widely applied in voice assistants, intelligent customer service, and online consultation. For example, intelligent customer service systems for electric vehicle charging are a key component of the electric vehicle charging service ecosystem. Through large-scale modeling technology, they provide users with intelligent question answering services such as charging policy consultation, equipment fault diagnosis, and billing rule explanations, aiming to improve the efficiency of charging services and user experience. However, existing question answering methods have significant shortcomings in controlling the scope of responses, making it difficult to accurately define the boundaries of the answer content. This leads to redundant or insufficient information in the generated responses, severely limiting the accuracy of the question answering results.

[0003] Therefore, how to determine the information in the question and answer process to improve the accuracy of the results is a problem that needs to be solved. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, apparatus, computer device, and storage medium for determining question-and-answer information to address the aforementioned technical problems, so as to improve the accuracy of question-and-answer results.

[0005] Firstly, this application provides a method for determining question-and-answer information, including:

[0006] Obtain the user's current question information and determine whether the current question information has a unique answer.

[0007] If the current question information does not have unique answer information, then the target answer range corresponding to the current question information is determined based on the initial answer range corresponding to the current question information and the answer range corresponding to the preceding question; the preceding question is the question before the current question information in the multi-round question-and-answer process;

[0008] Based on the target answer range, determine the alternative answers corresponding to the current question information;

[0009] The candidate answers are divided into segments to obtain multiple candidate answer segments;

[0010] Based on whether each of the candidate answer segments appeared in the multiple rounds of question and answer process before the current question information, the multiple candidate answer segments are filtered to obtain the filtered answer segments;

[0011] Based on the filtered answer fragments, determine the answer information corresponding to the current question information and output the answer information.

[0012] In one embodiment, determining the target answer range corresponding to the current question information based on the initial answer range corresponding to the current question information and the answer range corresponding to the preceding questions includes:

[0013] Retrieve all preceding questions before the current question information during the multi-round question-and-answer process;

[0014] If the answer ranges corresponding to all the preceding questions are narrowed down sequentially according to the question order, then the target preceding answer range is determined based on the answer range corresponding to the preceding question closest to the current question information;

[0015] If the answer range corresponding to all the prerequisite questions is not narrowed down sequentially according to the question order, then the target prerequisite question is determined from all the prerequisite questions based on the correlation coefficient between any two adjacent prerequisite questions; and the target prerequisite answer range is determined based on the answer range corresponding to the target prerequisite question.

[0016] The target answer range is determined based on the intersection between the initial answer range and the target prerequisite answer range.

[0017] In one embodiment, determining the target preceding problem from all preceding problems based on the correlation coefficients of adjacent preceding problems includes:

[0018] Determine the problem range difference, interval duration, and correlation coefficient between any two adjacent preceding problems in all the preceding problems;

[0019] Based on the difference in the scope of the problem, the interval duration, and the correlation coefficient, determine the content change coefficient of each pair of preceding problems;

[0020] Filter out all adjacent pairwise pre-questions from those whose content change coefficient is greater than or equal to a preset change coefficient threshold.

[0021] The preceding question closest to the current question information among the selected pairwise preceding questions is identified as the target preceding question.

[0022] In one embodiment, determining the target prerequisite answer range based on the answer range corresponding to the target prerequisite question includes:

[0023] Based on the aforementioned target-related questions and the current question information, valid questions are determined.

[0024] Based on whether the user interrupts the output of the corresponding answer to the valid question, and the user's reading time for the answer to each valid question, the range of invalid answers is determined;

[0025] The invalid answer range is removed from the answer range corresponding to the target prerequisite question to obtain the target prerequisite answer range.

[0026] In one embodiment, determining whether the current question information has unique answer information includes:

[0027] The initial answer to the current question information is determined within the range of the initial answers;

[0028] If the initial answer is not unique, then determine whether the current question information is an objective question;

[0029] If the current question information is an objective question, and the acceptance coefficient corresponding to the answer conditions of each initial answer is less than a preset coefficient threshold, then it is determined that the current question information does not have unique answer information.

[0030] In one embodiment, the method further includes:

[0031] Each of the initial answer conditions is determined as the target answer condition;

[0032] Determine the user approval coefficient of the target answer condition and the historical approval coefficient of the target answer condition, wherein the user approval coefficient is used to characterize the degree of user's acceptance of the target answer condition in the context of the multi-round question and answer process, and the historical approval coefficient is used to characterize the degree of acceptance of the target answer condition in the context of historical users in the historical question and answer process;

[0033] Based on the user approval coefficient and the historical approval coefficient, determine the approval coefficient corresponding to the target answer condition;

[0034] If the acceptance coefficient corresponding to the target answer condition is greater than or equal to the preset coefficient threshold, then the answer information for the current question information is determined based on the target answer condition.

[0035] In one embodiment, the step of filtering multiple candidate answer segments based on whether each candidate answer segment appeared in the multi-round question-and-answer process prior to the current question information to obtain filtered answer segments includes:

[0036] Each of the aforementioned alternative answer segments is designated as the target alternative answer segment;

[0037] If the target alternative answer fragment appears in the multi-round question and answer process before the current question information, then it is determined whether the user views the target alternative answer fragment based on the user's reading time of the target alternative answer fragment;

[0038] If the user has viewed the target candidate answer fragment, then based on the user's memory level, it is determined whether the user remembers the target candidate answer fragment. If the user remembers the target candidate answer fragment, the target candidate answer fragment is deleted from multiple candidate answer fragments until all candidate answer fragments have been traversed, resulting in the filtered answer fragment. The user's memory level is used to characterize the degree to which the user remembers the question and answer content during multiple rounds of question and answering.

[0039] If the user does not view the target candidate answer segment, and the user's follow-up question rate for the target candidate answer segment is less than the preset follow-up question rate, then the target candidate answer segment is deleted from the multiple candidate answer segments until all candidate answer segments have been traversed to obtain the filtered answer segment;

[0040] If the target alternative answer fragment did not appear in the previous rounds of question and answer before the current question information, then the guidance coefficient of the target alternative answer fragment is determined based on the degree of influence of the target alternative answer fragment on the understanding of the answer content, the proportion of interruptions when the target alternative answer fragment was output in the historical question and answer process, and the proportion of the target alternative answer fragment not viewed in the historical question and answer process. When the guidance coefficient is less than the preset guidance coefficient, the target alternative answer fragment is deleted from multiple alternative answer fragments until all alternative answer fragments have been traversed to obtain the filtered answer fragment.

[0041] Secondly, this application also provides a question-and-answer information determining device, comprising:

[0042] The question acquisition module is used to acquire the user's current question information and determine whether the current question information has a unique answer.

[0043] The answer range determination module is used to determine the target answer range corresponding to the current question information based on the initial answer range corresponding to the current question information and the answer range corresponding to the preceding question if the current question information does not have unique answer information; the preceding question is the question before the current question information in the multi-round question-and-answer process;

[0044] The alternative answer determination module is used to determine the alternative answers corresponding to the current question information based on the target answer range;

[0045] The segmentation module is used to segment the candidate answers into multiple candidate answer segments;

[0046] The filtering module is used to filter multiple candidate answer segments based on whether each candidate answer segment appears in the multiple rounds of question and answering before the current question information, so as to obtain the filtered answer segments;

[0047] The answer determination module is used to determine the answer information corresponding to the current question information based on the filtered answer fragments, and output the answer information.

[0048] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the question-and-answer information determination method in any of the above embodiments.

[0049] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the question-and-answer information determination method of any of the above embodiments.

[0050] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps in the question-and-answer information determination method of any of the above embodiments.

[0051] The aforementioned question-and-answer information determination method, apparatus, computer equipment, and storage medium determine whether the current question information has a unique answer. If the current question information does not have a unique answer, a target answer range is determined based on the initial answer range corresponding to the current question information and the answer range corresponding to the preceding questions. This defines the answer range for the current question information by using the answer range corresponding to the preceding questions, avoiding the generation of redundant answers and improving the accuracy of subsequent question-and-answer results. Furthermore, determining the answer by defining the answer range effectively reduces the amount of data required for question-and-answer queries, improving efficiency. In addition, alternative answers are identified within the determined target answer range, and redundant alternative answer fragments are filtered by checking if they appear in previous rounds of question-and-answer processes. Filtered answer fragments are obtained, and finally, the answer information corresponding to the current question information is generated and output based on these filtered answer fragments. This further reduces redundant information in the final answer and further improves the accuracy of the question-and-answer results. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1This is a schematic diagram illustrating an application scenario of a question-and-answer information determination method provided in an embodiment of this application;

[0054] Figure 2 This is a flowchart of a question-and-answer information determination method provided in an embodiment of this application;

[0055] Figure 3 This is a flowchart of a method for determining a target prerequisite problem provided in an embodiment of this application;

[0056] Figure 4 This is a schematic diagram of the structure of a question-and-answer information determination device provided in an embodiment of this application;

[0057] Figure 5 This is a schematic diagram of the internal structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0059] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0060] The question-and-answer information determination method provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, Figure 1 This is a schematic diagram illustrating an application scenario of a question-and-answer information determination method provided in this application embodiment. The terminal 102 communicates with the server 104 via a network. A data storage system can store the data that the server 104 needs to process. The data storage system can be integrated onto the server 104, or it can be located in the cloud or on other network servers. As an example, the terminal 102 can be an electric vehicle charging smart terminal, and can be, but is not limited to, various personal computers, laptops, smartphones, tablets, drones, low-altitude aircraft, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, projection devices, etc. The server 104 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0061] In one exemplary embodiment, such as Figure 2 As shown, Figure 2 This is a flowchart of a question-and-answer information determination method provided in an embodiment of this application, which is applied to... Figure 1 Taking server 104 as an example, the method may include the following steps:

[0062] Step 201: Obtain the user's current question information and determine whether the current question information has a unique answer.

[0063] For example, a user can input current problem information in the terminal, and further, the terminal transmits the user's current problem information to the server, so that the server can obtain the user's current problem information.

[0064] As an example, users can input current problem information into the terminal through at least one of the following methods: text input, voice, or handwriting. In the context of electric vehicle charging services, the current problem information can be information related to electric vehicle charging services, such as the current charging fee calculation standard, charging time, or charging recommendations.

[0065] It should be noted that the embodiments of this application only illustrate the application of the intelligent question-and-answer method in the electric vehicle charging service scenario. In practical applications, this method can also be applied to scenarios such as hotel intelligent customer service, learning question-and-answer, and medical question-and-answer, as well as other question-and-answer scenarios, without any limitations.

[0066] After the server obtains the user's current question information, it determines whether the question has a unique answer. As an example, a preliminary answer can be obtained by searching the current question information; whether this preliminary answer is unique is then used to determine if the question has a unique answer. If the preliminary answer is not unique, or if no answer can be found for the current question information, it is determined that the question does not have a unique answer.

[0067] It should be noted that if the preliminary answer is not unique, it may not be output to the terminal; it is only used by the server to determine whether the current question information has a unique answer. If the preliminary answer is unique, it is the answer to the current question information. Furthermore, the server can transmit the unique preliminary answer to the terminal, allowing the terminal to display the answer to the current question information to the user.

[0068] Step 202: If the current question information does not have unique answer information, then determine the target answer range corresponding to the current question information based on the initial answer range corresponding to the current question information and the answer range corresponding to the previous question.

[0069] Among them, the preceding question is the question that precedes the current question information during the multi-round question-and-answer process.

[0070] Specifically, if the current question information does not have a unique answer, then the initial answer range corresponding to the current question information and the answer range corresponding to the preceding question are determined, and the target answer range corresponding to the current question information is determined based on the initial answer range and the answer range corresponding to the preceding question.

[0071] As an example, in the electric vehicle charging service scenario, the initial answer range corresponding to the current question information can be determined from the question and answer knowledge base of the electric vehicle charging service. This question and answer knowledge base can be preset or continuously updated during the question and answer process, without any restrictions.

[0072] It should be noted that in a question-and-answer scenario, the question-and-answer process for the same user is usually multi-round. The preceding question can be a question preceding the current question in the multi-round process. This preceding question can be one question preceding the current question, or multiple questions preceding the current question; there are no restrictions. If the preceding question consists of multiple questions preceding the current question, then the answer range corresponding to the preceding question can be the intersection of the answer ranges corresponding to all the preceding questions. If the current question is the first question in the multi-round process, then there is no preceding question, and correspondingly, no answer range for a preceding question. In this case, the initial answer range corresponding to the current question can be determined as the target answer range corresponding to the current question.

[0073] Furthermore, the intersection of the initial answer range corresponding to the current question information and the answer range corresponding to the previous question can be determined as the target answer range corresponding to the current question information.

[0074] As an example, in the question-and-answer scenario of electric vehicle charging services, multi-round questions and answers typically involve content on the same or related topics. Furthermore, the question-and-answer knowledge base for electric vehicle charging services can include information on different topics, such as charging costs, charging strategies, and charging duration. The answer range of the preceding questions reflects their respective topics. In determining the answer range corresponding to the current question, the answer topics of the preceding questions are referenced. This allows for the precise identification of the topic information corresponding to the current question through diverse topic information, and ultimately, the accurate definition of the answer range for the current question.

[0075] Step 203: Determine the alternative answers corresponding to the current question information based on the target answer range.

[0076] Furthermore, within the defined range of target answers, the current question information can be queried and retrieved to obtain the answer corresponding to the current question information, and this answer can be designated as the alternative answer corresponding to the current question information.

[0077] As an example, if the alternative answer is unique or the content data volume of the alternative answer is less than the content data volume threshold, the alternative answer can be determined as the answer information corresponding to the current question information, and the alternative answer can be transmitted to the terminal so that the alternative answer can be displayed on the terminal to answer the current question information.

[0078] It should be noted that the content data volume of the candidate answers can be obtained by quantifying the candidate answers according to at least one of the data formats of word count, character count, or memory usage, or by quantifying other data formats of the candidate answers; no restrictions are imposed here. The content data volume threshold can be preset based on the data format of the content data volume.

[0079] Step 204: Divide the candidate answers into segments to obtain multiple candidate answer segments.

[0080] As an example, if the alternative answer is not unique or the content data volume of the alternative answer is greater than or equal to the content data volume threshold, the alternative answer can be segmented to obtain multiple alternative answer segments.

[0081] It should be noted that the methods for dividing the answer choices into segments include, but are not limited to, dividing them into multiple answer choice segments based on the number of lines, the number of characters, the number of words, the characteristic separators, and templates.

[0082] Step 205: Based on whether each candidate answer fragment appears in the multiple rounds of question and answer before the current question information, filter the multiple candidate answer fragments to obtain the filtered answer fragments.

[0083] Furthermore, it can be determined whether each alternative answer fragment has appeared in the multiple rounds of question-and-answer process before the current question information, and the multiple alternative answer fragments can be filtered based on the judgment result to obtain the filtered answer fragment.

[0084] Specifically, during the user's current multi-turn question-and-answer process, if a candidate answer segment has appeared in a previous question or answer, its redundancy can be determined based on whether the user has viewed that segment. As an example, whether a user has viewed a candidate answer segment can be determined by the length of time they spend reading it. For instance, if the user's reading time for a candidate answer segment is greater than or equal to a preset reading time, it means the user has viewed the segment; conversely, if the reading time is less than the preset time, it means the user has not viewed it. It should be noted that the preset reading time can be adaptively set according to needs; for example, it could be 1 second, 2 seconds, or 3 seconds, or other durations, without restriction. Furthermore, if the user has not viewed a candidate answer segment, it is considered redundant and will be removed from the list of candidate answer segments.

[0085] Furthermore, during the user's current multi-round question-and-answer process, if a certain candidate answer fragment has not appeared in previous questions or answers, its redundancy can be determined based on its importance to the final answer. The importance of a candidate answer fragment to the final answer refers to its influence on the understanding of the final answer's content. As an example, the importance of a candidate answer fragment to the final answer can be determined by the probability of that fragment appearing in the answers to questions with the same content as the current question during different users' historical question-and-answer processes. Specifically, if the importance of a candidate answer fragment to the final answer is greater than or equal to a preset influence level, the corresponding candidate answer fragment is not redundant; if the importance of a candidate answer fragment to the final answer is less than the preset influence level, the corresponding candidate answer fragment is redundant and will be removed from the list of candidate answer fragments. It should be noted that the preset influence level can be adaptively set according to needs and is not limited here.

[0086] Furthermore, by iterating through each candidate answer segment in the above manner to determine whether it is a redundant answer segment, the redundant answer segments are deleted from the multiple candidate answer segments, thereby obtaining the filtered answer segments.

[0087] Step 206: Determine the answer information corresponding to the current question information based on the filtered answer fragments, and output the answer information.

[0088] Furthermore, the filtered answer fragment can be identified as the final answer information corresponding to the current question information, and this answer information can be transmitted to the terminal, thereby enabling the output of the answer information corresponding to the current question information in the terminal.

[0089] In the above implementation process, by determining whether the current question information has a unique answer, and when the current question information does not have a unique answer, a target answer range is determined based on the initial answer range corresponding to the current question information and the answer range corresponding to the preceding questions. This defines the answer range for the current question information by using the answer range corresponding to the preceding questions, avoiding the generation of redundant answers and improving the accuracy of subsequent question-and-answer results. Furthermore, determining the answer by defining the answer range effectively reduces the amount of data required for question-and-answer queries, improving efficiency. In addition, by identifying alternative answers within the determined target answer range, redundant alternative answer fragments are filtered out by checking if they appear in previous rounds of question-and-answer processes. Finally, the answer information corresponding to the current question information is generated based on the filtered answer fragments and output, further reducing redundancy in the final answer and improving the accuracy of the question-and-answer results.

[0090] In one embodiment, determining the target answer range for the current question information based on the initial answer range corresponding to the current question information and the answer range corresponding to the previous questions may include the following steps:

[0091] Step 1: Obtain all preceding questions before the current question information during the multi-round question-and-answer process.

[0092] Step 2: If the answer ranges for all preceding questions are narrowed down sequentially according to the question order, then the target preceding answer range is determined based on the answer range of the preceding question closest to the current question information.

[0093] Alternatively, in step 2: if the range of answers corresponding to all prerequisite questions does not narrow down sequentially according to the question order, then determine the target prerequisite question from all prerequisite questions based on the correlation coefficient between adjacent prerequisite questions.

[0094] Step 3: Determine the target-prerequisite answer range based on the answer range corresponding to the target-prerequisite question.

[0095] Step 4: Determine the target answer range based on the intersection between the initial answer range and the target prerequisite answer range.

[0096] For example, if the current question is not the first question in a multi-round question-and-answer process, all questions preceding the current question are retrieved and identified as the preceding questions. If the current question is the first question in the multi-round question-and-answer process, then the target answer range corresponding to the current question is the initial answer range corresponding to the current question. Specifically, whether the current question is the first question in the multi-round question-and-answer process can be determined by its position within the process.

[0097] In a multi-round question-and-answer process, questions on the same or related topics are usually consecutive, and users typically add limiting conditions to questions on the same or related topics in the order in which the questions are asked. This causes the range of answers for questions on the same topic to gradually narrow down in the order of the questions. Therefore, it is possible to determine whether multiple preceding questions are on the same topic by judging whether the range of answers for all preceding questions narrows down in the order of the questions.

[0098] Specifically, if the answer ranges for all preceding questions are narrowed down sequentially according to the question order, it can be determined that the themes of all preceding questions are the same or related. In this case, the answer range of the preceding question closest to the current question can be determined as the target preceding answer range. That is, by using the answer range of the preceding question closest to the current question as a reference, the target answer range corresponding to the current question can be determined more accurately.

[0099] Furthermore, if a user raises multiple questions covering different topics during a multi-round Q&A process, the user will not ask the questions sequentially in a progressively narrowing manner. Correspondingly, the answer range for all preceding questions will not decrease sequentially according to the question order. For example, if a user has raised a total of 10 questions, and the current question is the 10th question, with questions 1-3 about charging costs, questions 4-8 about charging time, and questions 9-10 about charging locations, then the answer range for the 10th question only needs to refer to the 9th question. Therefore, the target preceding question can be determined from all preceding questions based on the correlation coefficient between adjacent preceding questions, and the target preceding answer range can be determined based on the answer range corresponding to the target preceding question.

[0100] Specifically, the correlation coefficients of adjacent pre-questions can be determined sequentially. As an example, the correlation coefficient between any two adjacent pre-questions can be determined by the frequency with which different users simultaneously raised the same questions as those two adjacent pre-questions during the historical question-and-answer process. That is, the correlation coefficient between two adjacent pre-questions is the frequency with which the corresponding two questions appear together in the historical question-and-answer process. Further, the two pre-questions with a correlation coefficient less than a preset correlation coefficient and closest to the current question information are selected, and the pre-question that appears later in the list of the two selected pre-questions is determined as the target pre-question.

[0101] It should be noted that the preset correlation coefficient can be adaptively set according to the actual scenario. It represents the lowest coefficient corresponding to two questions having the same or related topics. If the correlation coefficient is less than the preset correlation coefficient, it means that the topics of the two questions are not the same or related, and the topic of the question ranked later has changed from that of the question ranked earlier. Furthermore, among the two preceding questions with a correlation coefficient less than the preset correlation coefficient and closest to the current question information, the preceding question ranked later is identified as the target preceding question, indicating that there is a higher probability that the topic of the current question information is the same or related to the topic of the preceding question ranked later.

[0102] Furthermore, the range of answers corresponding to the identified target prerequisite questions can be defined as the target prerequisite answer range, and the intersection between the initial answer range and the target prerequisite answer range can be defined as the target answer range.

[0103] In the above implementation process, by determining whether the answer ranges corresponding to all preceding questions are the same or related topics in multiple rounds of question-and-answer sessions, and when the questions are the same or related topics, the answer range corresponding to the preceding question closest to the current question is determined as the target preceding answer range, thus enabling more precise limitation of the answer range for the current question. When the topics of the questions in multiple rounds of question-and-answer sessions are different and unrelated, based on the correlation coefficients of adjacent preceding questions, the preceding questions with the same or related topics as the current question can be accurately determined from multiple preceding questions, thus accurately determining the answer range for the current question and further improving the accuracy of the question-and-answer results.

[0104] In one embodiment, Figure 3 This is a flowchart of a method for determining a target prerequisite problem provided in an embodiment of this application. Specifically, it determines the target prerequisite problem from all prerequisite problems based on the correlation coefficients between adjacent prerequisite problems. Figure 3 As shown, the method may include the following steps:

[0105] Step 301: Determine the problem range difference, interval duration, and correlation coefficient between any two adjacent prerequisite problems.

[0106] Step 302: Based on the difference in problem scope, the interval duration, and the correlation coefficient, determine the content change coefficient of each pair of preceding problems.

[0107] Step 303: Select from all adjacent pairwise pre-questions that have a content change coefficient greater than or equal to a preset change coefficient threshold.

[0108] Step 304: Determine the target preceding problem as the preceding problem that is closest to the current problem information among the selected pairwise preceding problems.

[0109] For example, when determining whether the topics of two adjacent preceding questions are the same or related, in addition to considering the correlation between the two questions, the time interval between the questions and the difference in the scope of the questions can also be considered.

[0110] Specifically, the scope of each preceding question can be determined, and the difference between the scopes of any two adjacent preceding questions can be further determined, i.e., the question scope difference. It should be noted that the question scope can be obtained by quantifying the data volume, presentation length, and word count of the corresponding question.

[0111] The process involves determining the time interval between each pair of adjacent preceding questions, and then determining the content change coefficient for each pair of preceding questions based on the difference in question scope, the time interval between questions, and the correlation coefficient. As an example, the weighted average of the difference in question scope, the time interval between questions, and the correlation coefficient between each pair of adjacent preceding questions can be used to determine the content change coefficient for each pair of preceding questions.

[0112] The larger the correlation coefficient, the lower the probability of the two questions changing their topic content, and the smaller the content change coefficient; while the larger the difference in the question range, the higher the probability of changing the topic content, and the higher the content change coefficient; the longer the interval, the higher the probability of changing the topic content, and the higher the content change coefficient.

[0113] Furthermore, to more accurately determine the interval between two adjacent questions, it's necessary to exclude the user's required reading time for each question. Therefore, the user's average reading speed can be obtained, and the reading time for the first-ranked preceding question between two adjacent questions can be determined based on this average reading speed. The difference between the interval between the two preceding questions and the corresponding reading time is then defined as the interval duration. The interval duration reflects the time remaining between two adjacent preceding questions excluding the required reading time; therefore, a longer interval duration increases the likelihood of a change in topic content and a higher content change coefficient.

[0114] Finally, based on the difference in scope, interval, and correlation coefficient between adjacent pairs of questions, the content change coefficient of each pair of preceding questions is determined. Specifically, the weighted average of the difference in scope, interval, and correlation coefficient between adjacent pairs of questions is determined as the content change coefficient of each pair of preceding questions.

[0115] Furthermore, from all adjacent pairwise preceding questions, adjacent pairwise preceding questions with a content change coefficient greater than or equal to a preset change coefficient threshold are selected, and the preceding question closest to the current question information among the selected adjacent pairwise preceding questions is determined as the target preceding question.

[0116] It should be noted that the preset change coefficient threshold can be adaptively set according to the actual scenario. It represents the lowest coefficient corresponding to two questions having different and unrelated topics. If the content change coefficient is greater than or equal to the preset change coefficient threshold, it indicates that the topics of the two corresponding questions are different and unrelated, meaning the topic of the later-ranked question has changed compared to the earlier-ranked question. Furthermore, among the two preceding questions whose content change coefficient is greater than or equal to the preset change coefficient threshold and are closest to the current question information, the later-ranked preceding question is identified as the target preceding question, indicating a higher probability that the topic of the current question information is the same as or related to the topic of the later-ranked preceding question.

[0117] In the above implementation process, based on the problem range difference, interval duration, and correlation coefficient between all adjacent pre-questions, the content change coefficient of each pair of pre-questions is determined. Thus, the content change coefficient determined by the problem range difference, interval duration, and correlation coefficient can more accurately determine the correlation between adjacent pre-questions, thereby improving the accuracy of determining the target pre-question and further facilitating the improvement of the accuracy of determining the answer to subsequent questions.

[0118] In one embodiment, determining the target prerequisite answer range based on the answer range corresponding to the target prerequisite question may include the following steps:

[0119] Step 1: Determine the valid questions based on the prior questions and the current question information.

[0120] Step 2: Determine the range of invalid answers based on whether the user interrupts the output of the corresponding answer for each valid question, and the time the user spends reading the answer for each valid question.

[0121] Step 3: Remove invalid answer ranges from the answer range corresponding to the target prerequisite question to obtain the target prerequisite answer range.

[0122] For example, to more accurately determine the range of answers preceding the target, the range of answers corresponding to the target preceding questions can be filtered. Specifically, the target preceding questions and the questions between the target preceding questions and the current question information can all be determined as valid questions.

[0123] Furthermore, iterate through each valid question to determine whether the user interrupts the output of the answer. If a valid question is interrupted during the output process, the range of answers corresponding to that valid question is determined to be an invalid range. If the user interrupts the output of the question's answer, it means that the content already output but interrupted is not what the user wanted; therefore, this content is invalid. This method allows us to determine whether the range of answers corresponding to all valid questions is an invalid range.

[0124] Furthermore, the system iterates through and determines the reading time for each valid question's answer, i.e., the answer reading time. If the reading time for an answer to a valid question is less than a preset reading time threshold, then the range of answers for that valid question is determined as an invalid answer range. It should be noted that the preset reading time threshold can be set based on the user's average reading time and is not limited here. In a real-world scenario, if a user's reading time for an answer is less than their usual reading time, it means the user skipped or ignored that section of content, indicating that the user does not need that content and therefore chose to skip it. This allows the system to determine whether the range of answers for all valid questions is an invalid answer range.

[0125] Furthermore, the invalid answer range consists of content that users interrupt, skip, or ignore. Therefore, this part of the content does not need to be considered when determining the target prerequisite answer range. That is, the invalid answer range is removed from the answer range corresponding to the target prerequisite question to obtain the target prerequisite answer range.

[0126] It should be noted that if a user interrupts or ignores the answers to all valid questions, it means that when determining the target answer range, there is no need to refer to the answer range of the previous questions. In other words, the initial answer range corresponding to the current question information is the target answer range.

[0127] In the above implementation process, the range of invalid answers is determined by judging whether the answers to the target prerequisite question and all questions between the target prerequisite question and the current question information are invalid. Furthermore, the range of invalid answers is precisely filtered by considering whether the user is interrupted during the output of the answer to the valid question and the user's reading time for the valid question. Further, removing invalid answers from the answer range corresponding to the target prerequisite question eliminates content that the user is not interested in from the target answer range, further improving the accuracy of subsequent answer results.

[0128] In one embodiment, determining whether the current question information has unique answer information may include the following steps:

[0129] Step 1: Determine the initial answer to the current question information within the initial answer range.

[0130] Step 2: If the initial answer is not unique, determine whether the current question information is an objective question.

[0131] Step 3: If the current question is an objective question, and the acceptance coefficient corresponding to the answer conditions of each initial answer is less than the preset coefficient threshold, then it is determined that the current question does not have a unique answer.

[0132] For example, the answer corresponding to the current question information can be determined from the initial answer range. This answer is the initial answer, and it is determined whether the initial answer is unique.

[0133] If the initial answer is unique, then that initial answer is determined as the answer to the current question and output. If the initial answer is not unique, then it is determined whether the current question is an objective question.

[0134] As an example, an objective question can be a pre-defined question with a standard answer. If the current question is not an objective question, a prompt can be output to encourage the user to add conditions, thus guiding the user to obtain the desired answer. If the current question is an objective question and the corresponding initial answer is not unique, it means that the user's answer conditions when inputting the current question information are insufficient, resulting in the inability to output a unique answer.

[0135] If the current question is an objective question, then the content data volume of all initial answers to the current question can be determined. If the content data volume of all initial answers is less than the content data volume threshold, then all initial answers are determined as the answer information corresponding to the current question information, and that answer information is output.

[0136] It should be noted that the initial answer's content data volume can be obtained by quantifying data such as the number of characters, words, and memory usage. The content data volume threshold can be adaptively set according to actual conditions. This threshold can be used to limit the maximum output volume of the answer, thereby preventing the answer content from becoming too large and difficult to output completely.

[0137] Excessive answer content leads to numerous redundant answers. Specifically, if the total data volume of all initial answers exceeds or equals a data volume threshold, a recognition coefficient is determined for each initial answer's condition. For example, when a user inquires about charging costs, since costs vary depending on the time and location, multiple answers may exist if the user doesn't specify a time and location. Therefore, the answer conditions for different answers to the charging cost question are time and location; when these conditions are definite, the answer is definite. The applicability of an initial answer to the current question can be determined by comparing its recognition coefficient with a preset coefficient threshold. It's important to note that the preset coefficient threshold can be adaptively set based on the actual scenario and can be used to represent the minimum coefficient corresponding to user acceptance of the answer condition.

[0138] If an initial answer has a recognition coefficient greater than or equal to a preset threshold, then the current question information has a unique answer. For example, the initial answer with a recognition coefficient greater than or equal to the preset threshold can be determined as the answer to the current question information. If the recognition coefficients of all initial answers are less than the preset threshold, it means that none of the initial answers meet the user's requirements, and therefore the current question information does not have a unique answer.

[0139] In multi-round question-and-answer sessions, user questions may receive inaccurate answers due to insufficient information. If the initial answer options are limited, all options can be output to meet the user's needs. For example, if a user asks about charging costs, the initial possible answers are four combinations: area A, area B, peak hours, and off-peak hours. All these options can be output, allowing the user to find the answer they want. However, if the initial answer options are numerous but the conditions are largely default, a unique answer can be output based on these default conditions. For instance, if the user's charging cost is assumed to be the current location, the real-time charging cost can be output. However, if the user's default tolerance level (i.e., acceptance level) is low, indicating that the user's answer conditions are uncertain, the content needs to be reduced, resulting in no unique output.

[0140] In the above implementation process, the initial answer to the current question information is determined within the initial answer range. When the initial answer is not unique, it is determined whether the current question is an objective question. If the current question is an objective question, the recognition coefficient corresponding to the answer conditions of the initial answer is used to determine whether the current question information has unique answer information. This effectively determines whether the current question does not have a unique answer due to insufficient answer conditions, which facilitates prompting the user to supplement the answer conditions.

[0141] In one embodiment, the method may further include the following steps:

[0142] Step 1: Determine the answer conditions for each initial answer as the target answer conditions.

[0143] Step 2: Determine the user approval coefficient of the target answer condition, and determine the historical approval coefficient of the target answer condition.

[0144] Among them, the user approval coefficient is used to characterize the degree of user acceptance of the target answer conditions in the context of multi-round question and answer process, and the historical approval coefficient is used to characterize the degree of acceptance of the target answer conditions in the context of historical users in the historical question and answer process.

[0145] Step 3: Determine the recognition coefficient corresponding to the target answer condition based on the user recognition coefficient and the historical recognition coefficient.

[0146] Step 4: If the acceptance coefficient corresponding to the target answer condition is greater than or equal to the preset coefficient threshold, then determine the answer information for the current question based on the target answer condition.

[0147] For example, the acceptance coefficient of the initial answer's conditions can include a user acceptance coefficient and a historical acceptance coefficient. The user acceptance coefficient characterizes the degree to which the user's contextual understanding of the target answer's conditions during multiple rounds of question-and-answer sessions, while the historical acceptance coefficient characterizes the degree to which the contextual understanding of the target answer's conditions by past users during historical question-and-answer sessions.

[0148] As an example, the answer conditions for each initial answer can be determined as the target answer conditions.

[0149] Determining the user approval coefficient for the target answer condition can be achieved through the following methods: If none of the preceding questions contain the target answer condition, the user approval coefficient for the target answer condition is a preset value. If all preceding questions contain the target answer condition, the time interval between the preceding questions containing the target answer condition and the current question information, as well as the number of questions in the interval, are determined. The user approval coefficient for the target answer condition is then determined based on the time interval and the number of questions in the interval.

[0150] During multiple rounds of question and answer, users may have mentioned answer conditions in the preceding questions. For example, before asking how much the charging fee is, the user may have asked questions related to a certain order and given specific time and location. In this case, it is considered that the preceding questions contain answer conditions, and it can be determined whether the time and location can be extracted from the preceding questions.

[0151] Specifically, it is determined whether the target answer condition is contained in all the preceding questions in the multi-round question-and-answer process. If the target answer condition is not contained in all the preceding questions, the target answer condition cannot be determined based on the preceding questions. In this case, the user approval coefficient corresponding to the target answer condition can be a preset value. As an example, the preset value can be 0, 0.1, or other very small values, which are used to represent that the user does not approve of the target answer condition.

[0152] If all preceding questions contain conditions for the target answer, but it cannot be directly determined that the target answer conditions correspond to the answer conditions of the current question, then the interval between the preceding questions containing the target answer conditions and the current question information, as well as the number of such interval questions, can be determined. Based on the interval duration and the number of such interval questions, the user approval coefficient corresponding to the target answer conditions can be determined.

[0153] Specifically, the interval duration and the number of questions at intervals corresponding to the target answer conditions can be normalized, and a weighted average can be calculated based on the normalized interval duration and the number of questions at intervals to obtain the user approval coefficient of the target answer conditions.

[0154] The more questions there are at intervals and the longer the intervals are, the lower the relevance between the previous answer conditions and the current question. Therefore, the probability that the target answer condition is the answer condition corresponding to the current question is lower, which means the user approval coefficient is lower.

[0155] Furthermore, the historical acceptance coefficient of the target answer condition can be determined using the following method:

[0156] Retrieve all historical conditions corresponding to historical questions with the same information as the current question from all historical users; if all historical conditions correspond to the preset matching rules, then determine the acceptance coefficient of the target answer condition in the preset matching rules as the historical acceptance coefficient of the target answer condition; if all historical conditions do not correspond to the preset matching rules, determine the probability of the target answer condition appearing in all historical conditions; if the probability is greater than or equal to the preset probability threshold, then determine the probability as the historical acceptance coefficient of the target answer condition; if the probability is less than the preset probability threshold, then the historical acceptance coefficient of the target answer condition is the preset value.

[0157] Specifically, historical data is used to obtain historical conditions mentioned by all users when they asked the same questions as the current question. For example, if a user asks, "How much does it cost to charge?", the historical conditions would be the historical conditions mentioned by all users who have asked this question, such as "How much does it cost to charge at 3 PM?", "How much does it cost to charge at 5 PM?", and "How much does it cost to charge at 8 AM?".

[0158] Further determine whether historical conditions correspond to preset matching rules. Correspondence can be determined by the regularity of the historical conditions. For example, if most historical users default to asking "What is the charging cost at 8 AM?" in all historical questions, then the historical condition of 8 AM is regular. Therefore, the acceptance coefficient corresponding to the historical condition of 8 AM recorded in the preset matching rules can be determined as the historical acceptance coefficient when the target answer condition is 8 AM. It should be noted that the preset matching rules contain a historical acceptance coefficient for each historical condition. If the target answer condition is 5 AM, then the acceptance coefficient corresponding to the historical condition of 5 AM in the preset matching rules will be determined as the historical acceptance coefficient when the target answer condition is 5 AM. Furthermore, the correspondence between historical conditions and preset matching rules can also be a correspondence between historical conditions and questionnaire results, a correspondence between historical conditions and fixed industry conditions, or other correspondence methods. This correspondence rule can be adaptively set according to the application scenario and is not limited here.

[0159] If all historical conditions do not correspond to the preset matching rules, the probability of the target answer condition appearing in all historical conditions is determined, and the magnitude of this probability is compared with the preset probability threshold. The preset probability value can be set according to the actual scenario and reverse adaptation. If the probability is greater than or equal to the preset probability threshold, it indicates that most historical users used the target answer condition as the answer condition for the same historical question as the current question information, and this probability can be determined as the historical acceptance coefficient of the target answer condition. If the probability is less than the preset probability threshold, it indicates that a minority of historical users used the target answer condition as the answer condition for the same historical question as the current question information. In this case, the historical acceptance coefficient of the target answer condition is the preset value, for example, the preset value is 0.

[0160] Furthermore, the approval coefficient corresponding to the target answer condition is determined based on the user approval coefficient and the historical approval coefficient. As an example, the coefficient obtained by weighted averaging the user approval coefficient and the historical approval coefficient corresponding to the target answer condition can be used as the approval coefficient corresponding to the target answer condition.

[0161] Furthermore, this method can be used to determine the acceptance coefficient for all answer conditions.

[0162] Furthermore, if the acceptance coefficient corresponding to the target answer condition is greater than or equal to the preset coefficient threshold, the answer information for the current question can be determined based on the target answer condition, that is, the answer information for the current question is determined based on the answer corresponding to the target answer condition.

[0163] In the above implementation process, the acceptance coefficient of the corresponding answer condition is determined by combining the user acceptance coefficient and the historical acceptance coefficient. Thus, when the answer conditions in the current question information are insufficient, the possible answer conditions for the current question can be determined by the contextual information of the user in multiple rounds of question and answer and the contextual information of historical users in historical question and answer. This makes it easier to determine the answer corresponding to the current question information based on the determined answer conditions, thereby improving the accuracy of the question and answer results.

[0164] In one embodiment, filtering multiple candidate answer fragments based on whether each candidate answer fragment appeared in multiple rounds of question-and-answer processes prior to the current question information to obtain filtered answer fragments may include the following steps:

[0165] Step 1: Identify each alternative answer segment as the target alternative answer segment.

[0166] Step 2: If the target alternative answer fragment appears in multiple rounds of question and answer before the current question information, determine whether the user should view the target alternative answer fragment based on the reading time of the user.

[0167] Step 3: If the user has already viewed the target answer fragment, determine whether the user remembers the target answer fragment based on the user's memory level. If the user remembers the target answer fragment, delete the target answer fragment from multiple answer fragments until all answer fragments have been traversed to obtain the filtered answer fragment.

[0168] User memory level is used to characterize the degree to which users remember the content of questions and answers during multiple rounds of question-and-answer sessions.

[0169] Alternatively, in step 3: if the user has not viewed the target candidate answer fragment, and the user's follow-up question rate for the target candidate answer fragment is less than the preset follow-up question rate, then delete the target candidate answer fragment from multiple candidate answer fragments until all candidate answer fragments have been traversed to obtain the filtered answer fragment.

[0170] Alternatively, in step 2: if the target alternative answer fragment has not appeared in the previous rounds of question and answer before the current question information, then determine the guidance coefficient of the target alternative answer fragment based on the degree of influence of the target alternative answer fragment on the understanding of the answer content, the proportion of interruptions when the target alternative answer fragment is output in the historical question and answer process, and the proportion of the target alternative answer fragment not viewed in the historical question and answer process; when the guidance coefficient is less than the preset guidance coefficient, delete the target alternative answer fragment from multiple alternative answer fragments until all alternative answer fragments have been traversed to obtain the filtered answer fragment.

[0171] For example, after dividing the alternative answers into multiple alternative answer segments, each alternative answer segment can be identified as the target alternative answer segment.

[0172] Furthermore, it is determined whether the target alternative answer segment appeared in the multiple rounds of question-and-answer sessions preceding the current question. If the target alternative answer segment appeared in the multiple rounds of question-and-answer sessions preceding the current question, it can be determined whether the user has viewed the target alternative answer segment based on the segment's reading time. Specifically, the average reading time of the user for each alternative answer segment can be calculated based on the user's reading speed. The difference between the average reading time and the actual reading time of each segment can be determined. If the difference is greater than or equal to a preset difference, it is determined that the user has not viewed the target alternative answer segment. If the difference is less than the preset difference, it is determined that the user has viewed the target alternative answer segment.

[0173] If the user has already viewed the target answer candidate segment, it can be determined whether the user remembers the target answer candidate segment based on their memory level. If the user remembers the target answer candidate segment, it indicates that the target answer candidate segment is redundant for the user. Therefore, the target answer candidate segment can be removed from multiple answer candidate segments. This process is repeated to determine whether each target answer candidate segment is redundant for the user until all answer candidate segments have been reviewed, resulting in the filtered answer segment. The user's memory level is used to characterize the degree to which the user remembers the question and answer content during multiple rounds of question and answer.

[0174] If the user does not view the target alternative answer segment, the user's follow-up question rate on the target alternative answer segment can be determined, and the difference between the follow-up question rate on the target alternative answer segment and the preset follow-up question rate can be further judged.

[0175] It should be noted that the follow-up question rate for the target answer option fragment can be the percentage of times a user asks for follow-up questions about the target answer option fragment out of the total number of times the target answer option fragment was not provided in multiple rounds of question and answer. The preset follow-up question rate can be set according to actual conditions and is not limited here.

[0176] Furthermore, if the follow-up question rate is less than the preset follow-up question rate, it indicates that the target candidate answer segment is a redundant answer segment for the user. In this case, the target candidate answer segment can be deleted from multiple candidate answer segments. This process is repeated to determine whether each target candidate answer segment is redundant for the user, until all candidate answer segments have been traversed, resulting in the filtered answer segment. If the follow-up question rate is greater than or equal to the preset follow-up question rate, then the target candidate answer segment is useful information for the user.

[0177] Conversely, if the target answer fragment did not appear in previous rounds of question-and-answer sessions before the current question, then the degree of influence of the target answer fragment on the understanding of the answer content, the percentage of interruptions during the output of the target answer fragment in historical question-and-answer sessions, and the percentage of times the target answer fragment was not viewed in historical question-and-answer sessions should be determined. Specifically, the degree of influence of the target answer fragment on the understanding of the answer content can be determined through a questionnaire survey. The percentage of interruptions can be the ratio of the number of times the target answer fragment was interrupted in historical question-and-answer sessions to the total number of times the target answer fragment was output. The percentage of times the target answer fragment was not viewed can be the ratio of the number of times the target answer fragment was not viewed in historical question-and-answer sessions to the total number of times the target answer fragment was output.

[0178] Furthermore, a weighted average is calculated based on the degree of influence of the target answer fragment on the understanding of the answer content, the percentage of interruptions during the output of the target answer fragment in historical Q&A sessions, and the percentage of target answer fragments not viewed in historical Q&A sessions, to obtain the guidance coefficient of the target answer fragment. The guidance coefficient of the target answer fragment can be used to characterize the degree of user interest in the target answer fragment.

[0179] Furthermore, the algorithm compares the guidance coefficient of the target candidate answer segment with a preset guidance coefficient. The preset guidance coefficient can be adaptively set according to the actual scenario and is not restricted here. The preset guidance coefficient represents the minimum level of interest the user has in the candidate answer segment. When the guidance coefficient is less than the preset guidance coefficient, it indicates that the user is interested in the target candidate answer segment. In this case, the target candidate answer segment is removed from the list of candidate answer segments. This process is repeated to determine whether each target candidate answer segment is of interest to the user until all candidate answer segments have been traversed, resulting in the filtered answer segment.

[0180] In real-world scenarios, a higher percentage of interruptions and unviewed responses indicates a lower user demand for the corresponding alternative answer segments, thus resulting in a lower guidance coefficient. For example, regarding charging suggestions, if the user is not interested after they are provided, all charging suggestions can be disregarded in future responses, effectively narrowing down the range of answers and improving accuracy. Conversely, if the user is interested, a valid range of answers can be identified, and the next response should focus on providing the content the user needs. The degree of influence reflects the importance of the segment content to the current answer; if the current answer cannot be understood without providing the segment content, then it is clearly necessary to provide it. Therefore, a higher degree of influence results in a higher guidance index.

[0181] In the above implementation process, based on whether each alternative answer fragment appeared in the multiple rounds of question-and-answer sessions preceding the current question, and in the scenario where it appeared, it is determined whether the user viewed the corresponding alternative answer fragment. Further, redundant alternative answer fragments remembered by the user are filtered from the viewed fragments. Among the unviewed alternative answer fragments, redundant fragments are filtered out based on the user's follow-up question rate, thus achieving redundant fragment filtering. If an alternative answer fragment did not appear in the multiple rounds of question-and-answer sessions preceding the current question, a guidance coefficient for the target alternative answer fragment is determined based on its influence, interruption rate, and unviewed rate. This guidance coefficient is then used to determine the user's interest in the target alternative answer fragment, identifying alternative answer fragments that the user is not interested in. Finally, redundant and uninteresting alternative answer fragments are removed from the multiple alternative answer fragments, effectively reducing redundant information in the answers and further improving the accuracy of question-and-answer result determination.

[0182] In one embodiment, determining whether a user has remembered a segment of the target answer choices based on their memory level may include the following steps:

[0183] Step 1: Obtain the user's search history for historical questions to determine the user's average memory capacity.

[0184] Specifically, the search history includes the time and duration of a user's browsing of historical questions and answers. After multiple rounds of question-and-answer sessions, users may forget previous content but retain some recollection, so they will use methods such as backtracking and searching to retrieve past records. In the user's search history for historical questions, if a user starts searching for a fragment based on the historical record when asking historical question A, and this fragment is identified as the search fragment, the amount of content data between the position of the search fragment in the historical record and the position of historical question A is determined. This content data can be quantified based on the number of characters or the memory occupied by the data. This method can determine the amount of content data corresponding to different questions, and the average amount of content data for different questions is determined as the user's average memory capacity, which is used to characterize the user's average memory level.

[0185] Step 2: Determine the first position of the target alternative answer fragment in the multiple rounds of question-and-answer process before the current question information, and the second position corresponding to the current question information.

[0186] Step 3: Determine the amount of real-time content data between the first and second locations.

[0187] Furthermore, the first position of the target alternative answer fragment in the multi-round question-and-answer process before the current question information, and the second position corresponding to the current question information are determined, and the amount of content data between the first position and the second position is determined, and this amount of content data is determined as the real-time content data amount.

[0188] Step 4: If the real-time content data volume is greater than or equal to the average memory capacity, determine that the user has not remembered the target alternative answer segment.

[0189] Specifically, the system determines the difference between the real-time content data volume and the average memory capacity. If the real-time content data volume is greater than or equal to the average memory capacity, it indicates that the real-time content data volume exceeds the user's average memory capacity, thus confirming that the user has not remembered the target answer fragment.

[0190] Taking average memory capacity as an example, if the average memory capacity is 100 characters, it means that a user can usually remember 100 characters of content. If the real-time content data between the target candidate answer fragment and the current question information exceeds 100 characters, it means that the user has not remembered the target candidate answer fragment.

[0191] Alternatively, in step 4: if the amount of real-time content data is less than the average memory capacity, then iterate through and determine the correlation coefficient between any two pairs of questions and answers in the multiple rounds of question and answer before the current question information, and determine the two questions and answers in the multiple rounds of question and answer with a correlation coefficient greater than or equal to the preset correlation coefficient threshold as related questions and answers.

[0192] If the real-time content data volume is less than the average memory capacity, it cannot be guaranteed that the user will remember the target answer fragment. This is because the correlation between questions varies across different question scenarios, and consequently, users' memory varies depending on the question scenario. Users tend to remember more of simpler content and less of complex content. For example, if the current question involves a lot of unfamiliar professional knowledge, the user might only remember 50 characters or 120 characters. Therefore, when the real-time content data volume is less than the average memory capacity, the correlation coefficient between any two pairs of questions and answers in previous rounds of question-and-answer sessions can be used to determine whether the user has remembered the target answer fragment.

[0193] Specifically, the correlation coefficient between any two pairs of questions and answers can be determined based on the frequency of co-occurrence of the same questions in the historical question-and-answer records. After determining the correlation coefficient between any two pairs of questions and answers, two questions and answers with a correlation coefficient greater than or equal to a preset correlation coefficient threshold are identified as related questions and answers.

[0194] Step 5: Determine the number of related questions and answers for each question and answer in the multi-round question and answer process.

[0195] Step 6: Determine the average number of related questions and answers for all questions and answers as the average number of related questions and answers.

[0196] Step 7: Iterate through and determine the maximum number of interval questions and answers corresponding to each question and answer.

[0197] Step 8: Determine the average of the maximum number of interval questions and answers as the interval question and answer average.

[0198] Step 9: Determine the user's memory load based on the average number of associations and the average number of questions and answers at intervals.

[0199] Specifically, the user's memory load can be obtained by summing the average number of associations with the average number of interval questions and answers.

[0200] A higher average number of associations indicates that users can remember more question-and-answer content, thus resulting in a greater user memory load. The average interval between questions and answers directly reflects the number of question-and-answer rounds a user can remember when asking a question; therefore, a higher average interval between questions and answers indicates a greater user memory load. User memory load characterizes a user's average level of memory regarding the current question-and-answer context.

[0201] Step 10: Determine the memory load of the target candidate answer segment based on the number of questions and answers between the current question information and the target candidate answer segment, as well as the number of related questions and answers to the current question information.

[0202] Specifically, the memory load of the target answer fragment can be determined by the average of the number of questions and answers between the current question information and the target answer fragment and the number of questions and answers associated with the current question information. The memory load of the target answer fragment represents the level of memory required for the user to remember the target answer fragment in the current question and answer scenario.

[0203] Step 11: If the memory load of the target alternative answer fragment is less than or equal to the user's memory load, then the user remembers the target alternative answer fragment.

[0204] Furthermore, if the memory load of the target alternative answer fragment is less than or equal to the user's memory load, then the user will remember the target alternative answer fragment.

[0205] In the above implementation process, an initial determination is made based on the user's average memory capacity to determine whether the user has remembered the target answer fragment. Further, to accurately determine whether the user has remembered the target answer fragment when the real-time content data volume is less than the average memory capacity, the user's memory load for the current question-and-answer scenario is determined based on the correlation coefficient between any two pairs of questions and answers in the previous rounds of question-and-answer sessions. Then, based on the magnitude of the user's memory load and the memory load of the target answer fragment, the user's memory load for the target answer fragment is determined again, thus accurately determining whether the user has remembered the target answer fragment. Furthermore, the user's memory load varies for different question-and-answer scenarios. To accurately determine whether the current user has remembered the target answer fragment, the user's memory load for the current question-and-answer scenario is determined by the average correlation coefficient of multiple rounds of question-and-answer sessions and the average value of interval question-and-answer sessions in the current question-and-answer scenario, thus realizing an assessment of the user's memory capacity for the current question-and-answer scenario. This study employs a principle similar to determining user memory load to ascertain the memory load required for a user to remember a segment of the target answer in the current question-and-answer context (i.e., the memory load of the target answer segment). This allows for the assessment of the user's memory capacity required to remember the target answer segment in the current question-and-answer context. If the memory load of the target answer segment exceeds the user's memory load, it indicates that the user has forgotten the segment even if they have reviewed it. Determining whether the user has forgotten the answer segment helps to identify whether the answer segment content has been repeatedly output. Because repeated output may cause user dissatisfaction and reduce question-and-answer efficiency, determining whether the user has forgotten the answer segment can effectively improve the accuracy of the answer to the question and greatly reduce the probability of negative user emotions.

[0206] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0207] Based on the same inventive concept, this application also provides a question-and-answer information determining device for implementing the question-and-answer information determining method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more embodiments of the question-and-answer information determining device provided below can be found in the limitations of the question-and-answer information determining method described above, and will not be repeated here.

[0208] In one exemplary embodiment, Figure 4 This is a schematic diagram of the structure of a question-and-answer information determination device provided in an embodiment of this application, as shown below. Figure 4 As shown, the device includes:

[0209] The question acquisition module 401 is used to acquire the user's current question information and determine whether the current question information has unique answer information.

[0210] The answer range determination module 402 is used to determine the target answer range corresponding to the current question information based on the initial answer range corresponding to the current question information and the answer range corresponding to the previous question if the current question information does not have unique answer information; the previous question is the question before the current question information in the multi-round question and answer process;

[0211] The alternative answer determination module 403 is used to determine the alternative answers corresponding to the current question information based on the target answer range.

[0212] The segmentation module 404 is used to segment the candidate answers into multiple candidate answer segments.

[0213] The filtering module 405 is used to filter multiple candidate answer fragments based on whether each candidate answer fragment appears in the multiple rounds of question and answer before the current question information, and obtain the filtered answer fragments.

[0214] The answer determination module 406 is used to determine the answer information corresponding to the current question information based on the filtered answer fragments, and output the answer information.

[0215] In one embodiment, the answer range determination module 402 is specifically used for:

[0216] Retrieve all preceding questions in a multi-round question-and-answer process, prior to the current question information.

[0217] If the range of answers to all preceding questions is narrowed down sequentially according to the question order, then the target range of preceding answers is determined based on the range of answers to the preceding question closest to the current question information.

[0218] If the range of answers corresponding to all prerequisite questions does not narrow down sequentially according to the question order, then the target prerequisite question is determined from all prerequisite questions based on the correlation coefficient between any two adjacent prerequisite questions; and the range of answers corresponding to the target prerequisite question is determined based on the range of answers corresponding to the target prerequisite question.

[0219] The target answer range is determined based on the intersection between the initial answer range and the target prerequisite answer range.

[0220] In one embodiment, the answer range determination module 402 is specifically used for:

[0221] Determine the problem range difference, interval duration, and correlation coefficient between all adjacent prerequisite problems;

[0222] Based on the difference in problem scope, the interval duration, and the correlation coefficient, determine the content change coefficient of each pair of preceding problems;

[0223] Filter out all adjacent pairwise pre-questions from those whose content change coefficient is greater than or equal to a preset change coefficient threshold.

[0224] The preceding question that is closest to the current question information among the selected pairwise preceding questions is identified as the target preceding question.

[0225] In one embodiment, the answer range determination module 402 is specifically used for:

[0226] Valid questions are determined based on the prior questions and current problem information;

[0227] The range of invalid answers is determined based on whether the user interrupts the output of the corresponding answer for a valid question, and the user's reading time for the answer to each valid question.

[0228] The invalid answer range is obtained by removing the invalid answer range from the answer range corresponding to the target prerequisite question.

[0229] In one embodiment, the problem acquisition module 401 is specifically used for:

[0230] Determine the initial answer to the current question within the initial answer range;

[0231] If the initial answer is not unique, then determine whether the current question information is an objective question;

[0232] If the current question is an objective question, and the acceptance coefficient corresponding to the answer conditions of each initial answer is less than the preset coefficient threshold, then it is determined that the current question does not have a unique answer.

[0233] In one embodiment, the problem acquisition module 401 is specifically used for:

[0234] Each initial answer condition is defined as the target answer condition.

[0235] The user approval coefficient for the target answer condition and the historical approval coefficient for the target answer condition are determined. The user approval coefficient is used to characterize the degree of user acceptance of the target answer condition in the context of multiple rounds of question and answer, and the historical approval coefficient is used to characterize the degree of acceptance of the target answer condition in the context of historical users in the historical question and answer process.

[0236] Based on user approval coefficients and historical approval coefficients, determine the approval coefficients corresponding to the conditions of the target answer;

[0237] If the acceptance coefficient corresponding to the target answer condition is greater than or equal to the preset coefficient threshold, then the answer information for the current question is determined based on the target answer condition.

[0238] In one embodiment, the filtering module 405 is specifically used for:

[0239] Each of the alternative answer segments is designated as the target alternative answer segment;

[0240] If the target alternative answer fragment appears in multiple rounds of question and answer before the current question information, the user's decision on whether to view the target alternative answer fragment is determined based on the reading time of the user.

[0241] If the user has already viewed the target answer fragment, then based on the user's memory level, it is determined whether the user remembers the target answer fragment. If the user remembers the target answer fragment, the target answer fragment is removed from multiple answer fragments until all answer fragments have been traversed, resulting in the filtered answer fragment. The user's memory level is used to characterize the degree to which the user remembers the question and answer content during multiple rounds of question and answer.

[0242] If the user does not view the target candidate answer fragment, and the user's follow-up question rate for the target candidate answer fragment is less than the preset follow-up question rate, then the target candidate answer fragment will be deleted from multiple candidate answer fragments until all candidate answer fragments have been traversed to obtain the filtered answer fragment;

[0243] If the target alternative answer fragment has not appeared in the previous rounds of question and answer before the current question information, the guidance coefficient of the target alternative answer fragment is determined based on the degree of influence of the target alternative answer fragment on the understanding of the answer content, the proportion of interruptions when the target alternative answer fragment is output in the historical question and answer process, and the proportion of the target alternative answer fragment that has not been viewed in the historical question and answer process. When the guidance coefficient is less than the preset guidance coefficient, the target alternative answer fragment is deleted from multiple alternative answer fragments until all alternative answer fragments have been traversed to obtain the filtered answer fragment.

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

[0245] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 5 As shown. Figure 5 This is a schematic diagram of the internal structure of a computer device provided in an embodiment of this application, such as... Figure 5 As shown, the computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores question-and-answer data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a question-and-answer information determination method.

[0246] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0247] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the question-and-answer information determination method of any of the above embodiments.

[0248] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the question-and-answer information determination method of any of the above embodiments.

[0249] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0250] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0251] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0252] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for determining question-and-answer information, characterized in that, The method includes: Obtain the user's current question information and determine whether the current question information has a unique answer. If the current question information does not have unique answer information, then the target answer range corresponding to the current question information is determined based on the initial answer range corresponding to the current question information and the answer range corresponding to the preceding question; the preceding question is the question before the current question information in the multi-round question-and-answer process; Based on the target answer range, determine the alternative answers corresponding to the current question information; The candidate answers are divided into segments to obtain multiple candidate answer segments; Based on whether each of the candidate answer segments appeared in the multiple rounds of question and answer process before the current question information, the multiple candidate answer segments are filtered to obtain the filtered answer segments; The step involves filtering multiple candidate answer segments based on whether each of the candidate answer segments appeared in the multi-round question-and-answer process prior to the current question information, to obtain filtered answer segments, including: Each of the aforementioned alternative answer segments is designated as the target alternative answer segment; If the target alternative answer segment appears in the multi-round question and answer process before the current question information, the filtered answer segment is determined based on the user's reading time of the target alternative answer segment; If the target alternative answer fragment has not appeared in the previous rounds of question and answer before the current question information, then a guidance coefficient for the target alternative answer fragment is determined based on the degree of influence of the target alternative answer fragment on the understanding of the answer content, the percentage of interruptions when the target alternative answer fragment is output in the historical question and answer process, and the percentage of the target alternative answer fragment not viewed in the historical question and answer process. When the guidance coefficient is less than a preset guidance coefficient, the target alternative answer fragment is deleted from multiple alternative answer fragments until all alternative answer fragments have been traversed to obtain the filtered answer fragment. Based on the filtered answer fragments, determine the answer information corresponding to the current question information and output the answer information.

2. The method according to claim 1, characterized in that, The step of determining the target answer range corresponding to the current question information based on the initial answer range corresponding to the current question information and the answer range corresponding to the preceding questions includes: Retrieve all preceding questions before the current question information during the multi-round question-and-answer process; If the answer ranges corresponding to all the preceding questions are narrowed down sequentially according to the question order, then the target preceding answer range is determined based on the answer range corresponding to the preceding question closest to the current question information; If the answer range corresponding to all the prerequisite questions is not narrowed down sequentially according to the question order, then the target prerequisite question is determined from all the prerequisite questions based on the correlation coefficient between any two adjacent prerequisite questions; and the target prerequisite answer range is determined based on the answer range corresponding to the target prerequisite question. The target answer range is determined based on the intersection between the initial answer range and the target prerequisite answer range.

3. The method according to claim 2, characterized in that, The step of determining the target preceding problem from all preceding problems based on the correlation coefficients of adjacent preceding problems includes: Determine the problem range difference, interval duration, and correlation coefficient between any two adjacent preceding problems in all the preceding problems; Based on the difference in the scope of the problem, the interval duration, and the correlation coefficient, determine the content change coefficient of each pair of preceding problems; Filter out all adjacent pairwise pre-questions from those whose content change coefficient is greater than or equal to a preset change coefficient threshold. The preceding question closest to the current question information among the selected pairwise preceding questions is identified as the target preceding question.

4. The method according to claim 2 or 3, characterized in that, Determining the target prerequisite answer range based on the answer range corresponding to the target prerequisite question includes: Based on the aforementioned target-related questions and the current question information, valid questions are determined. Based on whether the user interrupts the output of the corresponding answer to the valid question, and the user's reading time for the answer to each valid question, the range of invalid answers is determined; The invalid answer range is removed from the answer range corresponding to the target prerequisite question to obtain the target prerequisite answer range.

5. The method according to claim 1, characterized in that, Determining whether the current question information has a unique answer includes: The initial answer to the current question information is determined within the range of the initial answers; If the initial answer is not unique, then determine whether the current question information is an objective question; If the current question information is an objective question, and the acceptance coefficient corresponding to the answer conditions of each initial answer is less than a preset coefficient threshold, then it is determined that the current question information does not have unique answer information.

6. The method according to claim 5, characterized in that, The method further includes: Each of the initial answer conditions is determined as the target answer condition; Determine the user approval coefficient of the target answer condition and the historical approval coefficient of the target answer condition, wherein the user approval coefficient is used to characterize the degree of user's acceptance of the target answer condition in the context of the multi-round question and answer process, and the historical approval coefficient is used to characterize the degree of acceptance of the target answer condition in the context of historical users in the historical question and answer process; Based on the user approval coefficient and the historical approval coefficient, determine the approval coefficient corresponding to the target answer condition; If the acceptance coefficient corresponding to the target answer condition is greater than or equal to the preset coefficient threshold, then the answer information for the current question information is determined based on the target answer condition.

7. The method according to claim 1, characterized in that, The step of determining the filtered answer segment based on the user's reading time of the target candidate answer segment includes: Whether the user views the target alternative answer segment is determined based on the reading time of the user on the target alternative answer segment; If the user has viewed the target candidate answer fragment, then based on the user's memory level, it is determined whether the user remembers the target candidate answer fragment. If the user remembers the target candidate answer fragment, the target candidate answer fragment is deleted from multiple candidate answer fragments until all candidate answer fragments have been traversed, resulting in the filtered answer fragment. The user's memory level is used to characterize the degree to which the user remembers the question and answer content during multiple rounds of question and answering. If the user does not view the target candidate answer segment, and the user's follow-up question rate for the target candidate answer segment is less than the preset follow-up question rate, then the target candidate answer segment is deleted from the multiple candidate answer segments until all candidate answer segments have been traversed to obtain the filtered answer segment.

8. A question-and-answer information determination device, characterized in that, The device includes: The question acquisition module is used to acquire the user's current question information and determine whether the current question information has a unique answer. The answer range determination module is used to determine the target answer range corresponding to the current question information based on the initial answer range corresponding to the current question information and the answer range corresponding to the preceding question if the current question information does not have unique answer information; the preceding question is the question before the current question information in the multi-round question-and-answer process; The alternative answer determination module is used to determine the alternative answers corresponding to the current question information based on the target answer range; The segmentation module is used to segment the candidate answers into multiple candidate answer segments; The filtering module is used to filter multiple candidate answer segments based on whether each candidate answer segment appears in the multiple rounds of question and answering before the current question information, so as to obtain the filtered answer segments; The filtering module is specifically used to: determine each of the candidate answer segments as the target candidate answer segment; If the target alternative answer segment appears in the multi-round question and answer process before the current question information, the filtered answer segment is determined based on the user's reading time of the target alternative answer segment; If the target alternative answer fragment has not appeared in the previous rounds of question and answer before the current question information, then a guidance coefficient for the target alternative answer fragment is determined based on the degree of influence of the target alternative answer fragment on the understanding of the answer content, the percentage of interruptions when the target alternative answer fragment is output in the historical question and answer process, and the percentage of the target alternative answer fragment not viewed in the historical question and answer process. When the guidance coefficient is less than a preset guidance coefficient, the target alternative answer fragment is deleted from multiple alternative answer fragments until all alternative answer fragments have been traversed to obtain the filtered answer fragment. The answer determination module is used to determine the answer information corresponding to the current question information based on the filtered answer fragments, and output the answer information.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images