Question and answer method, device and equipment and storage medium thereof

Abstract

The embodiment of the invention belongs to the technical field of question and answer optimization, is applied to an intelligent customer service question and answer prompt scene, and relates to a question and answer method, device and equipment and a storage medium thereof. According to the method, target problem data is input to a preset intention analysis model, the difficulty category to which the target problem data belongs is judged according to a preset category recognition component, difficulty problem category setting and follow-up recognition are performed on all problem data in advance, and only simple and scientific labeling is required, so that the problem data can be accurately analyzed. Therefore, a small amount of complex question data can be searched in a mark index mode, and a large amount of simple question data is searched in an intention analysis mode, so that the dialogue answering efficiency is greatly improved, and the dialogue experience of a user is improved.

Description

Technical Field

[0001] The present application relates to the technical field of question and answer optimization, and is applied to intelligent customer service question and answer prompt scenarios, and in particular to a question and answer method, device, equipment and storage medium thereof. Background Art

[0002] The Q&A prompting project is an intelligent technology that can automatically analyze the customer's voice or text through a robot and give corresponding answers or suggestions. The knowledge in the financial field is wide-ranging and complex, and it is difficult for ordinary customer service to accurately grasp it. The Q&A prompting project can use technology to accurately solve the questions raised by customers and quickly provide customers with satisfactory answers.

[0003] Traditional financial customer service requires manual work, which will lead to low service efficiency if there are a large number of customers. After adopting the question-answering prompt engineering technology, the robot can respond to customers quickly, improve the efficiency of consultation and service, and reduce the operating costs of enterprises. However, the question-answering prompt engineering requires collecting a large number of questions and having experts label them. This labeling process consumes a lot of manpower, and it is often impossible to reasonably distinguish the difficulty of questions in the labeling of question data and answer data, resulting in unscientific and simple labeling. Summary of the invention

[0004] The purpose of the embodiments of the present application is to propose a question-answering method, apparatus, device and storage medium thereof, so as to solve the problem that the prior art is often unable to reasonably distinguish the difficulty categories of questions when labeling question data and answer data, resulting in labeling that is not scientific and simple enough.

[0005] In order to solve the above technical problems, the embodiment of the present application provides a question-answering method, which adopts the following technical solution:

[0006] A question-answering method comprises the following steps:

[0007] Get target problem data;

[0008] Input the target question data into a preset intention analysis model, and determine the difficulty category to which the target question data belongs according to a preset category recognition component, wherein the difficulty category includes a simple question category and a complex question category;

[0009] If the target question data belongs to the simple question category, searching for answer knowledge content from a preset answer knowledge base through the intention analysis result, and generating answer feedback data according to the answer knowledge content;

[0010] If the target question data belongs to the complex question category, the answer knowledge content is searched from the answer knowledge base according to the preset annotation index, and the answer feedback data is generated according to the answer knowledge content.

[0011] Furthermore, the preset intention analysis model includes a language analysis model based on ChatGLM. Before executing the step of determining the difficulty category to which the target question data belongs according to the preset category identification component, the method further includes:

[0012] Obtain all historical input question data through the preset question and answer record log;

[0013] Inputting all the question data into the ChatGLM-based language analysis model to perform intent analysis;

[0014] Obtaining N search results of answer knowledge content corresponding to all question data from the answer knowledge base according to the intention analysis result, where N is a positive integer;

[0015] Based on the N times of answer knowledge content search results corresponding to all the question data, the difficulty categories to which all the question data belong are set.

[0016] Furthermore, before executing the step of obtaining N search results of answer knowledge content corresponding to all question data from the answer knowledge base according to the intention analysis result, the method further includes:

[0017] Presetting a strategy for acquiring the knowledge content of the N answers, wherein the acquisition strategy includes that the knowledge content of the N answers is not repeated and a maximum search time is preset for each search;

[0018] The step of obtaining N search results of answer knowledge content corresponding to all question data from the answer knowledge base according to the intention analysis result specifically includes:

[0019] Step 401, selecting current question data from all the question data;

[0020] Step 402, obtaining N search results of answer knowledge content corresponding to the current question data from the answer knowledge base according to the intention analysis result corresponding to the current question data and the acquisition strategy, wherein the answer knowledge content search results include the answer knowledge content obtained in each search and the search time in each search;

[0021] Step 403, taking the next question data of the current question data as the latest current question data, and re-execute step 402, until all question data have been searched for answer knowledge content N times, the search is completed, and step 402 is stopped.

[0022] Furthermore, the step of acquiring N search results of answer knowledge content corresponding to the current question data from the answer knowledge base according to the intention analysis result corresponding to the current question data and the acquisition strategy specifically includes:

[0023] According to the intention analysis result corresponding to the current question data and the acquisition strategy, the answer knowledge content of N searches is acquired, and it is identified whether the answer knowledge content of each search is a null value;

[0024] If the answer knowledge content of the current search is a null value, then determine whether the current search is the Nth search corresponding to the current question data;

[0025] If the current search is not the Nth search, continue to perform the next search; if the current search is the Nth search, reselect the latest current question data;

[0026] If the answer knowledge content of the current search is not a null value, then determine whether the current search is the Nth search corresponding to the current question data;

[0027] If the current search is the Nth search, then reselect the latest current question data;

[0028] If the current search is not the Nth search, the answer knowledge content is marked as a non-searchable state until the number of searches corresponding to the current question data reaches N times, then the answer knowledge content is modified to a searchable state, and each search result is recorded.

[0029] Furthermore, the step of setting the difficulty categories to which all the question data respectively belong based on the N times of answer knowledge content search results corresponding to all the question data respectively includes:

[0030] Count the N times of answer knowledge content search results corresponding to all question data;

[0031] According to the statistical results, determine whether there are fewer than N pieces of knowledge content corresponding to the problem data;

[0032] If all the question data correspond to N pieces of answer knowledge content, then calculate the average search time and the search time variance value corresponding to N searches for all the question data, and sort all the question data based on the average search time and the search time variance value to obtain a first sorting result;

[0033] If there are question data whose corresponding answer knowledge contents are less than N, the question data are sorted in ascending order of the number of answer knowledge contents to obtain a second sorting result, and the average search time and the search time variance value are calculated for other question data whose answer knowledge contents are all N, and the other question data are sorted based on the average search time and the search time variance value to obtain a third sorting result, and the second sorting result and the third sorting result are concatenated to obtain a fourth sorting result;

[0034] Simple question data and complex question data are divided according to the first sorting result or the fourth sorting result.

[0035] Furthermore, the step of sorting all the question data based on the average search time and the search time variance value to obtain a first sorting result specifically includes:

[0036] According to the preset comprehensive time-consuming characterization formula:

[0037] a i =b i ×ω1+c i ×ω2,

[0038] Calculate the comprehensive time consumption representation values ​​corresponding to all the problem data, where a i represents the comprehensive time-consuming representation value corresponding to problem data i, b i represents the average search time corresponding to N searches of problem data i, ω1 represents the weight value corresponding to the average search time, c i represents the variance value of the search time corresponding to N searches of the problem data i, ω2 represents the weight value corresponding to the variance value of the search time, and ω1 is greater than ω2;

[0039] The comprehensive time consumption characterization values ​​corresponding to all the problem data are processed in ascending order to obtain a sequence of comprehensive time consumption characterization values ​​corresponding to all the problem data;

[0040] Based on the comprehensive time-consuming characterization value sequence, a corresponding question data sequence is obtained as the first sorting result;

[0041] The step of sorting the other question data based on the average search time and the search time variance value to obtain a third sorting result, and concatenating the second sorting result and the third sorting result to obtain a fourth sorting result specifically includes:

[0042] Calculate the comprehensive time consumption characterization values ​​corresponding to the other problem data respectively through the comprehensive time consumption characterization formula;

[0043] The comprehensive time consumption characterization values ​​corresponding to the other problem data are processed in ascending order to obtain a comprehensive time consumption characterization value sequence corresponding to the other problem data;

[0044] Based on the comprehensive time-consuming characterization value sequence, a corresponding question data sequence is obtained as the third sorting result;

[0045] splicing the end position of the sequence of the second sorting result and the end position of the sequence of the third sorting result;

[0046] The fourth sorting result is obtained by taking the sequence start position of the third sorting result as the sequence start position of the fourth sorting result and taking the sequence start position of the second sorting result as the sequence end position of the fourth sorting result.

[0047] Furthermore, the step of dividing the simple question data and the complex question data according to the first sorting result or the fourth sorting result specifically includes:

[0048] According to a preset segmentation position, segmenting the first sorting result or the fourth sorting result;

[0049] The question data included from the sequence start position to the segmentation position of the first sorting result or the fourth sorting result is used as the simple question data;

[0050] The question data included from the end position of the sequence of the first sorting result or the fourth sorting result to the segmentation position is used as the complex question data;

[0051] After executing the step of dividing the simple question data and the complex question data according to the first sorting result or the fourth sorting result, the method further includes:

[0052] Deploy all the divided complex problem data as recognition knowledge to the category recognition component;

[0053] The corresponding answer knowledge contents of all the complex question data in the answer knowledge base are marked by manual marking, and the marking results are recorded in a preset query file as index knowledge of the answer knowledge base.

[0054] In order to solve the above technical problems, the embodiment of the present application also provides a question-answering device, which adopts the following technical solution:

[0055] A question-answering device, comprising:

[0056] A problem data acquisition module is used to acquire target problem data;

[0057] A difficulty category identification module, used to input the target question data into a preset intention analysis model, and determine the difficulty category to which the target question data belongs according to a preset category identification component, wherein the difficulty category includes a simple question category and a complex question category;

[0058] A first answer feedback module is used for searching answer knowledge content from a preset answer knowledge base according to the intention analysis result if the target question data belongs to the simple question category, and generating answer feedback data according to the answer knowledge content;

[0059] The second answer feedback module is used to search for answer knowledge content from the answer knowledge base according to a preset annotation index if the target question data belongs to a complex question category, and generate answer feedback data according to the answer knowledge content.

[0060] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0061] A computer device includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the above-mentioned question-and-answer method when executing the computer-readable instructions.

[0062] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0063] A computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the question-answering method described above.

[0064] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0065] The question-answering method described in the embodiment of the present application obtains target question data; inputs the target question data into a preset intent analysis model, determines the difficulty category of the target question data according to a preset category recognition component, and sets the difficulty category of all question data in advance, and then identifies them. Only simple and scientific labeling is required to realize the search by marked indexing of a small amount of complex question data, and the search for answer data by intent analysis is performed for a large amount of simple question data, which greatly improves the efficiency of dialogue answering and improves the user's dialogue experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0067] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0068] Figure 2 is a flow chart of an embodiment of the question-answering method according to the present application;

[0069] Figure 3 It is a schematic diagram of a specific embodiment of the method for setting the difficulty category of question data in the question-answering method described in the present application;

[0070] Figure 4 yes Figure 3 A flowchart of a specific embodiment of step 303 is shown;

[0071] Figure 5 yes Figure 4 A flowchart of a specific embodiment of step 402 is shown;

[0072] Figure 6 yes Figure 3 A flowchart of a specific embodiment of step 304 is shown;

[0073] Figure 7 yes Figure 6 A flowchart of a specific embodiment of step 605 is shown;

[0074] Figure 8 is a schematic structural diagram of an embodiment of a question-answering device according to the present application;

[0075] Fig. 9 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0076] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0077] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0078] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0079] like Figure 1 As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0080] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0081] Terminal devices 101, 102, 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, desktop computers, etc.

[0082] The server 105 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal devices 101 , 102 , and 103 .

[0083] It should be noted that the question-and-answer method provided in the embodiment of the present application is generally executed by a terminal device, and accordingly, the question-and-answer device is generally arranged in the terminal device.

[0084] It should be understood that Figure 1The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0085] Continue to refer Figure 2 , shows a flow chart of an embodiment of a question-answering method according to the present application. The question-answering method comprises the following steps:

[0086] Step 201, obtaining target question data.

[0087] Specifically, the target question data, for example, the dialogue question data raised by the user through a mobile terminal or a question-and-answer terminal, or the dialogue question data input by the user through an input terminal. In this embodiment, the question-and-answer method can be used in the field of financial technology, for example, in the dialogue scenario between intelligent AI customer service and financial business customers, or in the business consultation scenario between a banking business intelligent consulting terminal and a banking customer; the question-and-answer method can also be used in the field of smart medical technology, for example, in the dialogue consultation scenario between a medical service consulting terminal and a patient, the target question data includes question data in the form of voice and / or question data in the form of input text. Of course, it is not excluded that the question-and-answer method is used in question-and-answer scenarios in other technical fields. The main purpose is to provide the user terminal with more efficient answer feedback data through the question-and-answer method, improve the efficiency of answer data feedback, and improve the user's question-and-answer experience. Especially in the field of financial technology, the problems involved are more professional, so that the corresponding financial knowledge must be possessed when answering.

[0088] Step 202 , input the target question data into a preset intention analysis model, and determine the difficulty category to which the target question data belongs according to a preset category recognition component.

[0089] The difficulty categories include simple question categories and complex question categories.

[0090] Specifically, the preset intent analysis model includes a language analysis model based on ChatGLM. In this embodiment, the language analysis model based on ChatGLM, for example, ChatGLM large language model, ChatGLM, also known as Zhipu Qingyan, is developed based on the language model GLM-4 jointly trained by Zhipu AI Company in 2024, and can provide appropriate answers to user questions and requirements. Directly using the ChatGLM large language model for language analysis to determine the intent corresponding to the target question data avoids the construction of a large amount of question and answer data and reduces the consumption of manual labor.

[0091] In addition, the preset intent analysis model may also adopt an upgraded version of ChatGLM, for example, ChatGLM-9b, which is an open source conversational language model that supports bilingual question and answer in Chinese and English and is based on the General Language Model (GLM) architecture.

[0092] In this embodiment, before executing the step of determining the difficulty category to which the target question data belongs according to a preset category identification component, the method further includes a method for setting the difficulty category of question data.

[0093] Specifically, refer to Figure 3 , Figure 3 It is a schematic diagram of a specific embodiment of the method for setting the difficulty category of question data in the question-answering method described in the present application, including:

[0094] Step 301, obtaining all question data inputted in the past through a preset question and answer record log;

[0095] Specifically, the preset question and answer record log, for example: the question and answer record log between the customer and the customer service in the financial business intelligent AI customer service dialogue scenario, all the historical input question data refers to all the question data recorded in the question and answer record log in the historical dialogue scenario, and the "input" here is interpreted in an expanded way, including directly input question data, and also including the question data recorded in the question and answer record log after the question and answer voice is converted.

[0096] In this embodiment, all question data inputted historically are obtained through a preset question and answer record log, so as to directly combine with a specific financial business scenario, obtain all question data previously consulted by the customer, and classify all the question data into difficulty categories.

[0097] Step 302, inputting all the question data into the ChatGLM-based language analysis model to perform intent analysis;

[0098] Specifically, all the question data are respectively input into the ChatGLM-based language analysis model to perform intent analysis, with the aim of identifying the question intent of all the question data, thereby providing corresponding answer data therefor.

[0099] Step 303, obtaining N search results of answer knowledge content corresponding to all question data from the answer knowledge base according to the intention analysis result, where N is a positive integer;

[0100] In this embodiment, before executing the step of obtaining N answer knowledge content search results corresponding to all question data from the answer knowledge base according to the intention analysis results, the method also includes: pre-setting an acquisition strategy for the N answer knowledge contents, wherein the acquisition strategy includes that the N answer knowledge contents are not repeated and a maximum search time is preset for each search.

[0101] Through the acquisition strategy, firstly, it is ensured that the knowledge content of the answers obtained N times is different. Secondly, a maximum search time is preset for each search, which avoids the deadlock of continuous execution when no search results are found, so that the system can continue to operate stably and effectively.

[0102] Continue to refer Figure 4 , Figure 4 yes Figure 3 The flowchart of a specific embodiment of step 303 shown includes:

[0103] Step 401, selecting current question data from all the question data;

[0104] Specifically, all the question data may be searched and sorted or numbered in advance, and then different question data are selected in turn as current question data according to the search sorting sequence or the numbering results, thereby ensuring that all the question data are searched for answer data content N times.

[0105] Step 402, obtaining N search results of answer knowledge content corresponding to the current question data from the answer knowledge base according to the intention analysis result corresponding to the current question data and the acquisition strategy, wherein the answer knowledge content search results include the answer knowledge content obtained in each search and the search time in each search;

[0106] Specifically, the answer knowledge content search results can be composed of the answer knowledge content obtained each time the search is performed and the search time consumed each time the search is performed. For example, when the answer knowledge content is obtained, it is output in the form of "answer knowledge content + search time". It is also possible that the answer knowledge content is not obtained, in which case it is output in the form of "null value + maximum search time".

[0107] Continue to refer Figure 5 , Figure 5 yes Figure 4 The flowchart of a specific embodiment of step 402 shown includes:

[0108] Step 501, according to the intention analysis result corresponding to the current question data and the acquisition strategy, obtain the answer knowledge content of N searches, and identify whether the answer knowledge content of each search is a null value;

[0109] Specifically, since question data can often be divided into common question data or uncommon question data, the common question data can be understood as question data that is easier to answer, and corresponding to the simple question category, it often has more answer data; the uncommon question data can be understood as question data that is not easy to answer, and corresponding to the complex question category, it often has less answer data. Therefore, in the answer knowledge base with complex answer knowledge content, common question data can be easily searched for answer data, while uncommon question data has certain limitations in search speed or search breadth, especially when searching N times, the larger the N value, the more likely it is that there will be no N answer data for uncommon question data in the answer knowledge base, and there will be a search null value; furthermore, even if there are N answer data for uncommon question data in the answer knowledge base, due to the maximum search time limit and the complexity of the answer data in the answer knowledge base, a search null value may appear at the end of the maximum search time.

[0110] Step 502: if the answer knowledge content of the current search is a null value, determine whether the current search is the Nth search corresponding to the current question data;

[0111] Step 502a, if the current search is not the Nth search, continue to perform the next search;

[0112] Step 502b, if the current search is the Nth search, then reselect the latest current question data;

[0113] Step 503, if the answer knowledge content of the current search is not a null value, then determine whether the current search is the Nth search corresponding to the current question data;

[0114] Step 503a, if the current search is the Nth search, then reselect the latest current question data;

[0115] Step 503b, if the current search is not the Nth search, mark the answer knowledge content as a non-searchable state until the number of searches corresponding to the current question data reaches N times, modify the answer knowledge content to a searchable state, and record the search results of each time.

[0116] In this embodiment, when the answer knowledge content obtained by each search is not a null value, in order to make the next search meet the acquisition strategy, a marking method is used to mark the answer knowledge content as a state that cannot be searched further, until the number of searches corresponding to the current question data reaches N times, and the answer knowledge content is modified to a searchable state. Specifically, since there is often a correlation or a many-to-many relationship between question data and answer data, when the current question data is searched N times, the answer knowledge content marked as a state that cannot be searched further is modified to a searchable state, which does not affect the search of the next question data, thereby ensuring the accuracy and scientificity of the classification of question data categories.

[0117] Step 403, taking the next question data of the current question data as the latest current question data, and re-execute step 402, until all question data have been searched for answer knowledge content N times, the search is completed, and step 402 is stopped.

[0118] Specifically, the next question data of the current question data is taken as the latest current question data. The search sort sequence or the difference numbering result set in step 401 can be combined to filter out the next question data of the current question data as the latest current question data, thereby ensuring that N times of answer data content search are performed on all the question data.

[0119] Step 304 : based on the N times of answer knowledge content search results corresponding to all the question data, respectively, the difficulty categories to which all the question data respectively belong are set.

[0120] Continue to refer Figure 6 , Figure 6 yes Figure 3 The flowchart of a specific embodiment of step 304 shown includes:

[0121] Step 601, counting the N times of answer knowledge content search results corresponding to all question data;

[0122] Step 602, judging whether there are fewer than N pieces of answer knowledge content corresponding to the question data according to the statistical results;

[0123] Step 603: if all the question data correspond to N pieces of answer knowledge content, then calculate the average search time and the search time variance value corresponding to N searches of all the question data, and sort all the question data based on the average search time and the search time variance value to obtain a first sorting result;

[0124] In this embodiment, the step of sorting all the question data based on the average search time and the search time variance value to obtain a first sorting result specifically includes: according to a preset comprehensive time characterization formula:i =b i ×ω1+c i ×ω2, calculate the comprehensive time-consuming representation values ​​corresponding to all the problem data, where a i represents the comprehensive time-consuming representation value corresponding to problem data i, b i represents the average search time corresponding to N searches of problem data i, ω1 represents the weight value corresponding to the average search time, c i Represents the search time variance value corresponding to N searches for problem data i, ω2 represents the weight value corresponding to the search time variance value, and ω1 is greater than ω2; the comprehensive time consumption characterization values ​​corresponding to all the problem data are processed in ascending order to obtain a sequence of comprehensive time consumption characterization values ​​corresponding to all the problem data; based on the comprehensive time consumption characterization value sequence, a corresponding problem data sequence is obtained as the first sorting result.

[0125] Specifically, if the current question data is common question data, it is possible that different answer data contents can be quickly searched for N times. Therefore, it can be understood that the smaller the average search time corresponding to its N searches, the smaller the variance value of the search time is. If the current question data is uncommon question data, it is possible that the time used to search for different answer data contents N times is longer, or the next search time is longer than the previous search time, and it is in a continuously rising state. At this time, the average search time corresponding to its N searches will be larger, and the variance value of the search time is also larger. Therefore, consider introducing the average search time and the search time variance value to jointly constitute a comprehensive time representation value, and perform ascending processing according to the size of the comprehensive time representation value. At this time, the question data sequence corresponding to all question data should also be sorted from the most common question data to the least common question data.

[0126] Step 604: if there are question data whose corresponding answer knowledge contents are less than N, the question data are sorted in ascending order of the number of answer knowledge contents to obtain a second sorting result, and the average search time and the search time variance value are calculated for other question data whose answer knowledge contents are all N, and the other question data are sorted based on the average search time and the search time variance value to obtain a third sorting result, and the second sorting result and the third sorting result are concatenated to obtain a fourth sorting result;

[0127] In this embodiment, the steps of sorting the other problem data based on the average search time and the search time variance value to obtain a third sorting result, and concatenating the second sorting result and the third sorting result to obtain a fourth sorting result specifically include: calculating the comprehensive time characterization values ​​corresponding to the other problem data respectively through the comprehensive time characterization formula; performing ascending processing on the comprehensive time characterization values ​​corresponding to the other problem data respectively to obtain a sequence of comprehensive time characterization values ​​corresponding to the other problem data; obtaining a corresponding problem data sequence based on the comprehensive time characterization value sequence as the third sorting result; concatenating the sequence end position of the second sorting result and the sequence end position of the third sorting result; using the sequence start position of the third sorting result as the sequence start position of the fourth sorting result, and using the sequence start position of the second sorting result as the sequence end position of the fourth sorting result, to obtain the fourth sorting result.

[0128] Specifically, if there is question data corresponding to less than N pieces of answer knowledge content, the question data is sorted in ascending order according to the number of answer knowledge content pieces to obtain a second sorting result. At this time, the fewer the number of answer knowledge content pieces, the more uncommon the question data is. For other question data with N pieces of answer knowledge content, the above-mentioned comprehensive time-consuming representation value is used to splice out a third sorting result. At this time, the question data at the beginning of the third sorting result sequence is the most common question data. Therefore, consider splicing the sequence end position of the second sorting result and the sequence end position of the third sorting result, and use the sequence start position of the third sorting result as the sequence start position of the fourth sorting result, and use the sequence start position of the second sorting result as the sequence end position of the fourth sorting result, to obtain the fourth sorting result, thereby realizing sorting from the most common question data to the least common question data.

[0129] Step 605: Divide the simple question data and the complex question data according to the first sorting result or the fourth sorting result.

[0130] Continue to refer Figure 7 , Figure 7 yes Figure 6 The flowchart of a specific embodiment of step 605 shown includes:

[0131] Step 701, segmenting the first sorting result or the fourth sorting result according to a preset segmentation position;

[0132] Specifically, the preset split position, for example: starting from the start position of the sequence, splits the current sequence 9 to 1 along the end position of the sequence, and the current sequence includes the sequence corresponding to the first sorting result, or the sequence corresponding to the fourth sorting result; furthermore, the preset split position, for example: starting from the end position of the sequence, filters out M problem data forward along the start position of the sequence, and takes the position between the Mth problem data and the M+1th problem data as the split position.

[0133] Step 702: taking the question data from the sequence start position to the segmentation position of the first sorting result or the fourth sorting result as the simple question data;

[0134] Step 703: Use the question data included from the end position of the sequence of the first sorting result or the fourth sorting result to the split position as the complex question data.

[0135] In this embodiment, after executing the step of dividing simple question data and complex question data according to the first sorting result or the fourth sorting result, the method also includes: deploying all the divided complex question data as identification knowledge to the category identification component; manually labeling the answer knowledge content corresponding to all the complex question data in the answer knowledge base, and recording the labeling results as index knowledge of the answer knowledge base in a preset query file.

[0136] By deploying all the divided complex question data as identification knowledge to the category identification component, it is achieved that when the target question data is actually used to screen the answer data content, the category identification component can preferentially identify whether the target question data is complex question data; accordingly, a manual labeling method is used to label the answer knowledge content corresponding to all the complex question data in the answer knowledge base, and the labeling results are recorded in a preset query file as the index knowledge of the answer knowledge base. Only simple and scientific labeling is required. After identifying that the target question data is complex question data, it can be directly queried through the index without the need for intent search, thereby improving the search efficiency. For simple question data, the intent search method is still used, and a small amount of complex question data is searched by marking and indexing. For a large amount of simple question data, the intent analysis method is used to search for answer data, which greatly improves the efficiency of dialogue answering and improves the user's dialogue experience.

[0137] Step 203: If the target question data belongs to the simple question category, the answer knowledge content is searched from the preset answer knowledge base through the intention analysis result, and answer feedback data is generated according to the answer knowledge content.

[0138] Step 204: If the target question data belongs to a complex question category, then the answer knowledge content is searched from the answer knowledge base according to a preset annotation index, and answer feedback data is generated according to the answer knowledge content.

[0139] In this embodiment, the question-answering method further includes, through a preset data monitoring component, if it is monitored that the latest updated answer knowledge content in the answer knowledge base has reached a preset first update threshold, the first update threshold is, for example, a preset data volume value, 2GB of answer knowledge content is newly imported into the answer knowledge base, 2GB is the first update threshold; or if it is monitored that the latest input question data entry has reached a preset second update threshold, the second update threshold is, for example, a preset value of newly added question data entries, 200 newly added question data, 200 is the second update threshold, then execute Figure 3 The difficulty category setting method shown in the figure redefines the difficulty categories to which all the question data respectively belong, and continuously updates the difficulty categories of the questions when the answer knowledge content in the answer knowledge base gradually increases and / or the number of input question data entries increases. This enables the redeployment of the recognition knowledge in the category recognition component and the re-labeling of the new answer knowledge content in the answer knowledge base, ensuring continuous optimization and updating of the question and answer system and question and answer related knowledge.

[0140] The present application uses an intent analysis model to set difficulty categories for all question data, and realizes searching by marking and indexing a small amount of complex question data, while searching for answer data by intent analysis for a large amount of simple question data, which greatly improves the efficiency of dialogue and answering and improves the user's dialogue experience. At the same time, as the answer knowledge content in the answer knowledge base gradually increases and / or the number of input question data entries increases, the question difficulty categories are continuously updated, thereby realizing the redeployment of the recognition knowledge in the category recognition component and the re-labeling of the new answer knowledge content in the answer knowledge base, ensuring continuous optimization and updating of the question and answer system and question and answer related knowledge.

[0141] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0142] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0143] In an embodiment of the present application, through the intent analysis model, all question data are classified into difficulty categories, and a small amount of complex question data is searched by marking and indexing, while the answer data search is performed by intent analysis for a large amount of simple question data, which greatly improves the efficiency of dialogue and answering and improves the user's dialogue experience. At the same time, as the answer knowledge content in the answer knowledge base gradually increases and / or the number of input question data entries increases, the question difficulty categories are continuously updated, thereby realizing the redeployment of the recognition knowledge in the category recognition component and the re-labeling of the new answer knowledge content in the answer knowledge base, ensuring continuous optimization and updating of the question and answer system and question and answer related knowledge.

[0144] Further references Figure 8 , as a response to the above Figure 2 In order to realize the method shown in the figure, the present application provides an embodiment of a question-answering device. Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0145] like Figure 8 As shown, the question-answering device 800 of this embodiment includes: a question data acquisition module 801, a difficulty category identification module 802, a first answer feedback module 803 and a second answer feedback module 804. Among them:

[0146] The question data acquisition module 801 is used to acquire target question data;

[0147] The difficulty category identification module 802 is used to input the target question data into a preset intention analysis model, and determine the difficulty category to which the target question data belongs according to a preset category identification component, wherein the difficulty category includes a simple question category and a complex question category;

[0148] The first answer feedback module 803 is used to search for answer knowledge content from a preset answer knowledge base according to the intention analysis result if the target question data belongs to the simple question category, and generate answer feedback data according to the answer knowledge content;

[0149] The second answer feedback module 804 is used to search for answer knowledge content from the answer knowledge base according to a preset annotation index if the target question data belongs to a complex question category, and generate answer feedback data according to the answer knowledge content.

[0150] The question-and-answer device 800 described in this embodiment also includes a difficulty category setting module, which is used to obtain all historical question data input through a preset question-and-answer record log; it is also used to input all the question data into the ChatGLM-based language analysis model to perform intent analysis; it is also used to obtain N answer knowledge content search results corresponding to all the question data from the answer knowledge base according to the intent analysis results, where N is a positive integer; it is also used to set the difficulty categories to which all the question data belong based on the N answer knowledge content search results corresponding to all the question data.

[0151] The question-and-answer device 800 described in this embodiment also includes a difficulty category updating module, which is used to re-execute the difficulty category setting module through a preset data monitoring component, and reset the difficulty categories to which all the question data belong, if it is monitored that the latest updated answer knowledge content in the answer knowledge base has reached a preset first update threshold; or if it is monitored that the latest input question data entry has reached a preset second update threshold.

[0152] The question-and-answer device proposed in the present application belongs to the technical field of question-and-answer optimization, and is applied to intelligent customer service question-and-answer prompt scenarios. Through the intent analysis model, all question data are classified into difficulty categories, and a small amount of complex question data is searched by marking and indexing, while a large amount of simple question data is searched for answer data by adopting the intent analysis method, which greatly improves the efficiency of dialogue answering and improves the user's dialogue experience. At the same time, when the answer knowledge content in the answer knowledge base gradually increases and / or the number of input question data entries increases, the question difficulty category is continuously updated, thereby realizing the redeployment of the recognition knowledge in the category recognition component and the re-labeling of the new answer knowledge content in the answer knowledge base, ensuring the continuous optimization and updating of the question-and-answer system and question-and-answer related knowledge.

[0153] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0154] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0155] To solve the above technical problems, the present application also provides a computer device. Fig. 9 , Fig. 9 This is a basic structural block diagram of the computer device in this embodiment.

[0156] The computer device 9 includes a memory 9a, a processor 9b, and a network interface 9c that are interconnected through a system bus. It should be noted that the figure only shows a computer device 9 with components 9a-9c, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.

[0157] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0158] The memory 9a includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 9a can be an internal storage unit of the computer device 9, such as a hard disk or memory of the computer device 9. In other embodiments, the memory 9a can also be an external storage device of the computer device 9, such as a plug-in hard disk equipped on the computer device 9, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. Of course, the memory 9a can also include both the internal storage unit of the computer device 9 and its external storage device. In this embodiment, the memory 9a is generally used to store the operating system and various application software installed on the computer device 9, such as a computer-readable instruction of a question-answering method, etc. In addition, the memory 9a can also be used to temporarily store various types of data that have been output or are to be output.

[0159] The processor 9b may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 9b is generally used to control the overall operation of the computer device 9. In this embodiment, the processor 9b is used to run computer-readable instructions stored in the memory 9a or process data, such as computer-readable instructions for running the question-answering method.

[0160] The network interface 9c may include a wireless network interface or a wired network interface. The network interface 9c is generally used to establish a communication connection between the computer device 9 and other electronic devices.

[0161] The computer device proposed in this embodiment belongs to the field of question and answer optimization technology and is applied to intelligent customer service question and answer prompt scenarios. This application uses an intent analysis model to set difficulty and easy question categories for all question data, and realizes the search of a small amount of complex question data by marking and indexing, while the corresponding large amount of simple question data uses intent analysis to search for answer data, which greatly improves the efficiency of dialogue answers and improves the user's dialogue experience. At the same time, when the answer knowledge content in the answer knowledge base gradually increases and / or the input question data entries increase, the question difficulty category is continuously updated, thereby realizing the redeployment of the recognition knowledge in the category recognition component and the re-labeling of the new answer knowledge content in the answer knowledge base, ensuring the continuous optimization and updating of the question and answer system and question and answer related knowledge.

[0162] The present application also provides another implementation, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by a processor to enable the processor to perform the steps of the question-and-answer method as described above.

[0163] The computer-readable storage medium proposed in this embodiment belongs to the field of question-answering optimization technology and is applied to intelligent customer service question-answering prompt scenarios. This application uses an intent analysis model to set difficulty categories for all question data, and realizes the search of a small amount of complex question data by marking and indexing, while the search for answer data using intent analysis for a large amount of simple question data greatly improves the efficiency of dialogue answers and improves the user's dialogue experience. At the same time, as the answer knowledge content in the answer knowledge base gradually increases and / or more and more question data entries are input, the question difficulty category is continuously updated, thereby realizing the redeployment of the recognition knowledge in the category recognition component and the re-labeling of the new answer knowledge content in the answer knowledge base, ensuring the continuous optimization and updating of the question-answering system and question-answering related knowledge.

[0164] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0165] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A question-answering method, characterized in that: The steps include: Get target problem data; Input the target question data into a preset intention analysis model, and determine the difficulty category to which the target question data belongs according to a preset category recognition component, wherein the difficulty category includes a simple question category and a complex question category; If the target question data belongs to the simple question category, searching for answer knowledge content from a preset answer knowledge base through the intention analysis result, and generating answer feedback data according to the answer knowledge content; If the target question data belongs to the complex question category, the answer knowledge content is searched from the answer knowledge base according to the preset annotation index, and the answer feedback data is generated according to the answer knowledge content.

2. The question-answering method according to claim 1, characterized in that: The preset intention analysis model includes a language analysis model based on ChatGLM. Before executing the step of determining the difficulty category to which the target question data belongs according to the preset category identification component, the method further includes: Obtain all historical input question data through the preset question and answer record log; Inputting all the question data into the ChatGLM-based language analysis model to perform intent analysis; Obtaining N search results of answer knowledge content corresponding to all question data from the answer knowledge base according to the intention analysis result, where N is a positive integer; Based on the N times of answer knowledge content search results corresponding to all the question data, the difficulty categories to which all the question data belong are set.

3. The question-answering method according to claim 2, characterized in that: Before executing the step of acquiring N search results of answer knowledge content corresponding to all question data from the answer knowledge base according to the intention analysis result, the method further includes: Presetting a strategy for acquiring the knowledge content of the N answers, wherein the acquisition strategy includes that the knowledge content of the N answers is not repeated and a maximum search time is preset for each search; The step of obtaining N search results of answer knowledge content corresponding to all question data from the answer knowledge base according to the intention analysis result specifically includes: Step 401, selecting current question data from all the question data; Step 402, obtaining N search results of answer knowledge content corresponding to the current question data from the answer knowledge base according to the intention analysis result corresponding to the current question data and the acquisition strategy, wherein the answer knowledge content search results include the answer knowledge content obtained in each search and the search time in each search; Step 403, taking the next question data of the current question data as the latest current question data, and re-execute step 402, until all question data have been searched for answer knowledge content N times, the search is completed, and step 402 is stopped.

4. The question-answering method according to claim 3, characterized in that: The step of acquiring N search results of answer knowledge content corresponding to the current question data from the answer knowledge base according to the intention analysis result corresponding to the current question data and the acquisition strategy specifically includes: According to the intention analysis result corresponding to the current question data and the acquisition strategy, the answer knowledge content of N searches is acquired, and it is identified whether the answer knowledge content of each search is a null value; If the answer knowledge content of the current search is a null value, then determine whether the current search is the Nth search corresponding to the current question data; If the current search is not the Nth search, continue to perform the next search; If the current search is the Nth search, then reselect the latest current question data; If the answer knowledge content of the current search is not a null value, then determine whether the current search is the Nth search corresponding to the current question data; If the current search is the Nth search, then reselect the latest current search. Problem data; If the current search is not the Nth search, the answer knowledge content is marked as a non-searchable state until the number of searches corresponding to the current question data reaches N times, then the answer knowledge content is modified to a searchable state, and each search result is recorded.

5. The question-answering method according to claim 2, characterized in that: The step of setting the difficulty categories to which all the question data respectively belong based on the N times of answer knowledge content search results corresponding to all the question data respectively includes: Count the N times of answer knowledge content search results corresponding to all question data; According to the statistical results, determine whether there are fewer than N pieces of knowledge content corresponding to the problem data; If all the question data correspond to N pieces of answer knowledge content, then calculate the average search time and the search time variance value corresponding to N searches for all the question data, and sort all the question data based on the average search time and the search time variance value to obtain a first sorting result; If there are question data whose corresponding answer knowledge contents are less than N, the question data are sorted in ascending order of the number of answer knowledge contents to obtain a second sorting result, and the average search time and the search time variance value are calculated for other question data whose answer knowledge contents are all N, and the other question data are sorted based on the average search time and the search time variance value to obtain a third sorting result, and the second sorting result and the third sorting result are concatenated to obtain a fourth sorting result; According to the first sorting result or the fourth sorting result, simple question data and complex question data are divided.

6. The question-answering method according to claim 5, characterized in that: The step of sorting all the question data based on the average search time and the search time variance value to obtain a first sorting result specifically includes: According to the preset comprehensive time-consuming characterization formula: a i =b i ×ω1+c i ×ω2, Calculate the comprehensive time consumption representation values ​​corresponding to all the problem data, where a i represents the comprehensive time-consuming representation value corresponding to problem data i, b i represents the average search time corresponding to N searches of problem data i, ω1 represents the weight value corresponding to the average search time, c i represents the variance value of the search time corresponding to N searches of the problem data i, ω2 represents the weight value corresponding to the variance value of the search time, and ω1 is greater than ω2; The comprehensive time consumption characterization values ​​corresponding to all the problem data are processed in ascending order to obtain a sequence of comprehensive time consumption characterization values ​​corresponding to all the problem data; Based on the comprehensive time-consuming characterization value sequence, a corresponding question data sequence is obtained as the first sorting result; The step of sorting the other question data based on the average search time and the search time variance value to obtain a third sorting result, and concatenating the second sorting result and the third sorting result to obtain a fourth sorting result specifically includes: Calculate the comprehensive time consumption characterization values ​​corresponding to the other problem data respectively through the comprehensive time consumption characterization formula; The comprehensive time consumption characterization values ​​corresponding to the other problem data are processed in ascending order to obtain a comprehensive time consumption characterization value sequence corresponding to the other problem data; Based on the comprehensive time-consuming characterization value sequence, a corresponding question data sequence is obtained as the third sorting result; splicing the end position of the sequence of the second sorting result and the end position of the sequence of the third sorting result; The fourth sorting result is obtained by taking the sequence start position of the third sorting result as the sequence start position of the fourth sorting result and taking the sequence start position of the second sorting result as the sequence end position of the fourth sorting result.

7. The question-answering method according to claim 5, characterized in that: The step of dividing the simple question data and the complex question data according to the first sorting result or the fourth sorting result specifically includes: According to a preset segmentation position, segmenting the first sorting result or the fourth sorting result; The question data included from the sequence start position to the segmentation position of the first sorting result or the fourth sorting result is used as the simple question data; The question data included from the end position of the sequence of the first sorting result or the fourth sorting result to the segmentation position is used as the complex question data; After executing the step of dividing the simple question data and the complex question data according to the first sorting result or the fourth sorting result, the method further includes: Deploy all the divided complex problem data as recognition knowledge to the category recognition component; The corresponding answer knowledge contents of all the complex question data in the answer knowledge base are marked by manual marking, and the marking results are recorded in a preset query file as index knowledge of the answer knowledge base.

8. A question-answering device, characterized in that: include: A problem data acquisition module is used to acquire target problem data; A difficulty category identification module, used to input the target question data into a preset intention analysis model, and determine the difficulty category to which the target question data belongs according to a preset category identification component, wherein the difficulty category includes a simple question category and a complex question category; A first answer feedback module is used for searching answer knowledge content from a preset answer knowledge base according to the intention analysis result if the target question data belongs to the simple question category, and generating answer feedback data according to the answer knowledge content; The second answer feedback module is used to search for answer knowledge content from the answer knowledge base according to a preset annotation index if the target question data belongs to a complex question category, and generate answer feedback data according to the answer knowledge content.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the question-answering method according to any one of claims 1 to 7 when executing the computer-readable instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, the steps of the question-answering method according to any one of claims 1 to 7 are implemented.

Images