Question and answer processing method and device, medium and equipment

Abstract

The invention discloses a question and answer processing method and device, a medium and equipment. The method comprises the following steps: acquiring historical dialogue information of a dialogue with a target user in response to a dialogue truncation instruction; constructing a dialogue knowledge base based on the historical dialogue information; obtaining a plurality of pieces of key information from the dialogue knowledge base based on a current question of a current dialogue with a target user; based on the key information and the current question, a target question and answer model is obtained through pre-training, and a target answer corresponding to the current question is generated and output. According to the method provided by the invention, when the dialogue is cut off, the historical dialogue information is collected and the dialogue knowledge base is constructed, so that the historical dialogue information can be stored, and when the subsequent dialogue with the target user is carried out, the corresponding target answer can be accurately generated based on the current question and in combination with the historical dialogue information in the dialogue knowledge base; and the target answer is more reasonably and accurately output.

Description

Technical Field

[0001] The present invention relates to the technical field of natural language processing, and in particular to a question and answer processing method, device, medium and equipment. Background Art

[0002] With the development of the times, the advancement of science and technology, and the improvement of people's risk awareness, insurance, as an economic security system, has played an unprecedented role, and the AI ​​insurance broker industry has emerged. Today, artificial intelligence technology is developing rapidly, and human-computer interactive dialogue robots are more and more involved in people's daily lives. The quality of the dialogue robot's reply content has become particularly important.

[0003] However, the existing question-answering processing methods have the problem that the answer output is not accurate and reasonable enough. Summary of the invention

[0004] In view of this, the present invention provides a question-answering processing method, device, medium and equipment, the main purpose of which is to solve the problem that the answer output of the current question-answering processing method is not accurate and reasonable enough.

[0005] To solve the above problems, this application provides a question-answering method, including:

[0006] In response to the conversation interruption instruction, collecting historical conversation information with the target user;

[0007] Building a dialogue knowledge base based on the historical dialogue information;

[0008] Based on the current question of the current conversation with the target user, obtain some key information from the conversation knowledge base;

[0009] Based on each of the key information and the current question, a target question-answering model is obtained by using pre-training, and a target answer corresponding to the current question is generated and output.

[0010] Optionally, the constructing a dialogue knowledge base based on historical dialogue information specifically includes:

[0011] Extract key information from historical conversation information to obtain some key information;

[0012] Building the dialogue knowledge base based on each of the key information;

[0013] The key information includes: any one or more of the following: user status information, user identity information and event information.

[0014] Optionally, the question-answer processing method further includes:

[0015] The dialogue knowledge base is updated in real time based on the current dialogue information to obtain an updated current dialogue knowledge base.

[0016] Optionally, the updating of the dialogue knowledge base based on the current dialogue information in real time specifically includes:

[0017] Extracting semantic information from the current dialogue information to obtain some semantic relationship information and syntactic relationship information;

[0018] Based on the semantic relationship information and the syntactic relationship information, key information in the dialogue knowledge base is corrected or supplemented to update the dialogue knowledge base.

[0019] Optionally, the updating of the dialogue knowledge base based on the current dialogue information in real time specifically includes:

[0020] Extracting structural information from the current conversation information to obtain conversation topic information, conversation focus information, and conversation direction information;

[0021] Based on the conversation topic information, conversation focus information and conversation direction information, key information in the conversation knowledge base is corrected or supplemented to update the conversation knowledge base.

[0022] Optionally, the updating of the dialogue knowledge base based on the current dialogue information in real time specifically includes:

[0023] Collect buried point information in real time;

[0024] Based on the embedded point information, key information in the dialogue knowledge base is corrected or supplemented to update the dialogue knowledge base.

[0025] Optionally, based on each of the key information and the current question, using pre-training to obtain a target question-answering model, generating and outputting a target answer corresponding to the current question, specifically includes:

[0026] Construct target prompt sentences based on key information in the dialogue knowledge base;

[0027] Based on the target prompt sentence and the current question, the target question-answering model is used to generate and output a target answer corresponding to the current question.

[0028] To solve the above problems, the present application provides a question-answering device, comprising:

[0029] A collection module, configured to collect historical conversation information with a target user in response to a conversation interruption instruction;

[0030] A construction module, used to construct a dialogue knowledge base based on the historical dialogue information;

[0031] An acquisition module, used to acquire some key information from the dialogue knowledge base based on the current question of the current dialogue with the target user;

[0032] A processing module is used to obtain a target question-answering model based on the key information and the current question, generate a target answer corresponding to the current question and output it.

[0033] In order to solve the above problems, the present application provides a storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of any of the above-mentioned question and answer processing methods are implemented.

[0034] To solve the above problems, the present application provides an electronic device, which includes at least a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program on the memory, the steps of any of the above-mentioned question and answer processing methods are implemented.

[0035] The question-answer processing method in the present application can save the historical conversation information by collecting the historical conversation information and constructing a conversation knowledge base when the conversation is truncated. Therefore, in the subsequent conversation with the target user, the corresponding target answer can be accurately generated based on the current question and the historical conversation information in the conversation knowledge base, thereby avoiding the problem that the subsequent conversation cannot be combined with the historical conversation due to the truncation of the conversation information during a long conversation, resulting in the output of the subsequent conversation answer being inaccurate and unreasonable.

[0036] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0038] Figure 1 A flowchart of a question-and-answer processing method according to an embodiment of the present application;

[0039] Figure 2 This is a structural block diagram of a question-answer processing device according to another embodiment of the present application;

[0040] Figure 3 This is a structural block diagram of an electronic device according to another embodiment of the present application. DETAILED DESCRIPTION

[0041] Various aspects and features of the present application are described herein with reference to the accompanying drawings.

[0042] It should be understood that various modifications may be made to the embodiments of the present application. Therefore, the above description should not be considered as limiting, but only as an example of an embodiment. Other modifications within the scope and spirit of the present application will occur to those skilled in the art.

[0043] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the present application and, together with the general description of the present application given above and the detailed description of the embodiments given below, serve to explain the principles of the present application.

[0044] These and other characteristics of the present application will become apparent from the following description of a preferred form of embodiment given as a non-limiting example with reference to the accompanying drawings.

[0045] It should also be understood that although the present application has been described with reference to some specific examples, those skilled in the art will be able to readily implement many other equivalent forms of the present application.

[0046] The above and other aspects, features and advantages of the present application will become more apparent in view of the following detailed description when taken in conjunction with the accompanying drawings.

[0047] Specific embodiments of the present application are described hereinafter with reference to the accompanying drawings; however, it should be understood that the embodiments applied for are merely examples of the present application, which may be implemented in a variety of ways. Well-known and / or repeated functions and structures are not described in detail to avoid unnecessary or redundant details that obscure the present application. Therefore, the specific structural and functional details applied for herein are not intended to be limiting, but merely serve as a basis and representative basis for the claims to teach those skilled in the art to use the present application in a variety of ways with substantially any suitable detailed structure.

[0048] This specification may use the phrases "in one embodiment," "in another embodiment," "in yet another embodiment," or "in other embodiments," all of which may refer to one or more of the same or different embodiments according to the present application.

[0049] The present application embodiment provides a question-answering processing method, which can be specifically applied to a terminal, a server, or an intelligent question-answering device, such as a question-answering robot. Figure 1 As shown, the method in this embodiment includes the following steps:

[0050] Step S101, in response to a conversation interruption instruction, collecting historical conversation information with a target user;

[0051] The conversation truncation instruction in this step can specifically be a truncation instruction generated based on a predetermined interval time, for example, a truncation instruction is generated every 10 minutes to collect historical conversation information within the time period, or a predetermined conversation sentence, such as when the target user enters "OK, thank you", it can be considered that the current conversation content with the target user has ended, so that this part of the conversation content can be used as historical conversation information and then collected.

[0052] Step S102, constructing a dialogue knowledge base based on the historical dialogue information;

[0053] In this step, key information can be extracted from the historical conversation information, and then a conversation knowledge base corresponding to the target user can be constructed based on the key information. The key information can be any one or more of user status information, user identity information, and event information.

[0054] The user status information may include: the user's operation status on the target application / target page. The target application / target refers to the application / page that the intelligent customer service guides the user to operate, or the page / application that the target user consults the intelligent customer service. Taking the purchase of insurance as an example, the target application may be the application / page corresponding to the insurance product.

[0055] In this embodiment, the user identity information may be: user age, user name, whether it is the user himself, etc. The event information may be the target dialogue object in the dialogue between the intelligent customer service and the target user; taking the scenario where the user wants to buy insurance as an example, the dialogue object may be: user product demand, product name, etc.

[0056] Step S103, based on the current question in the current dialogue with the target user, obtaining some key information from the dialogue knowledge base;

[0057] In this step, after the conversation knowledge base is constructed, that is, after the historical conversation is truncated, when there is a subsequent conversation with the target user, key information can be obtained from the knowledge base, laying the foundation for the subsequent generation of the target answer based on the key information.

[0058] Step S104, based on each of the key information and the current question, a target question-answering model is obtained by using pre-training, and a target answer corresponding to the current question is generated and output.

[0059] In this step, the target question-answering model can be a ChatGPT model. Specifically, a target prompt sentence can be constructed based on key information, and then the target prompt sentence and the target question are input into the pre-trained ChatGPT model, and the target answer is generated and output through the ChatGPT model.

[0060] The question-answer processing method in the present application can save the historical conversation information by collecting the historical conversation information and constructing a conversation knowledge base when the conversation is truncated. Therefore, in the subsequent conversation with the target user, the corresponding target answer can be accurately generated based on the current question and the historical conversation information in the conversation knowledge base, thereby avoiding the problem that the subsequent conversation cannot be combined with the historical conversation due to the truncation of the conversation information during a long conversation, resulting in the output of the subsequent conversation answer being inaccurate and unreasonable.

[0061] Another embodiment of the present application provides a question-answering processing method, which can be specifically applied to the field of insurance sales. The method in this embodiment includes the following steps:

[0062] Step S201, in response to the conversation interruption instruction, collecting historical conversation information with the target user;

[0063] The historical dialogue information in this step is specifically dialogue information about insurance sales.

[0064] Step S202, extract key information from the historical conversation information to obtain a number of key information; construct the conversation knowledge base based on each of the key information; the key information includes: any one or more of the following: user status information, user identity information and event information;

[0065] Step S203, based on the current question in the current dialogue with the target user, obtaining some key information from the dialogue knowledge base;

[0066] In this step, the key information includes: any one or more of the following: user status information, user identity information, and event information. Among them, the user status information can be: the user's operating status for the target application / target page. The target application / target refers to the application / page that the intelligent customer service guides the user to operate, or the page / application that the target user consults the intelligent customer service. Taking the scenario of purchasing insurance as an example, the target application can be the application / page corresponding to the insurance product. User identity information can be: user age, user name, whether it is the user himself, etc. Event information can be the target dialogue object in the dialogue between the intelligent customer service and the target user; taking the scenario of introducing insurance to the user as an example, the dialogue object can be: user product needs, product name, etc.

[0067] Step S204, constructing a target prompt sentence based on each key information in the dialogue knowledge base; based on the target prompt sentence and the current question, using the target question-answering model, generating and outputting a target answer corresponding to the current question.

[0068] In the specific implementation process of this embodiment, after constructing and obtaining the conversation knowledge base corresponding to the target user, the conversation knowledge base can also be updated in real time based on the current conversation information. The specific updating process can adopt any one or more of the following updating methods:

[0069] Method 1: extract semantic information from the current dialogue information to obtain some semantic relationship information and syntactic relationship information; based on the semantic relationship information and the syntactic relationship information, correct or supplement the key information in the dialogue knowledge base to update the dialogue knowledge base.

[0070] In this method, syntactic relationship information may include subject-predicate relationship, semantic relationship, verb-object relationship, attributive-predicate relationship, adverbial-predicate relationship, etc. For example, "I am not xxx, I am his child, I want to consult for my mother...", then according to the subject-predicate relationship in the sentence, it can be determined that the person answering the phone is not the sales object (target user) himself, so the identity information of the target user in the knowledge base can be updated.

[0071] The semantic relationship may specifically include the target object referred to by the pronoun. For example, in the sentence “Okay, so I just want to ask if this insurance also has this deductible…”, we can determine that “this” refers to the insurance product mentioned / discussed in the previous conversation based on the semantic relationship in the sentence.

[0072] Method 2: extract structural information from the current dialogue information to obtain dialogue topic information, dialogue focus information and dialogue direction information; based on the dialogue topic information, dialogue focus information and dialogue direction information, correct or supplement key information in the dialogue knowledge base to update the dialogue knowledge base.

[0073] In this method, each turn-based conversation of the main content of the current conversation can be extracted to obtain the topic information, key information and conversation direction information from each turn-based conversation. Taking the conversation scenario of recommending insurance products to target users as an example, structural information can be extracted from the current conversation to obtain the topic information such as product introduction, claim information, etc. of the conversation content; the user's operation status can also be obtained, such as opening WeChat, entering the official account and other key information; the purchase intention, the depth of guiding customer operations (for example, from following the official account to the payment link) and other conversation direction information can also be obtained.

[0074] Method three: collecting tracking information in real time; correcting or supplementing key information in the dialogue knowledge base based on the tracking information to update the dialogue knowledge base.

[0075] In this method, taking the dialogue scenario of recommending / introducing insurance products to target users as an example, when the user status information in the historical dialogue knowledge base is "click me to buy insurance", that is, the user operates to the "click me to buy insurance" page. At this time, it can be judged / determined through the buried point information that the user's current operation page is the "click here to complete" page, that is, the user status is obtained as "click here to complete". Furthermore, the user status information in the dialogue knowledge base can be corrected based on the user status information of "click here to complete", so as to correct the key information in the knowledge base and obtain an updated dialogue knowledge base.

[0076] In this embodiment, by correcting and supplementing the key information in the dialogue knowledge base based on the current dialogue information, the key information in the dialogue knowledge base can be made more comprehensive and accurate, so that the target question and answer model can generate the target answer accurately and reasonably for the current page information when generating the target answer based on the current question query.

[0077] In the specific implementation process of this embodiment, before executing step S201, that is, before collecting historical dialogue information, the target question-answering model may be pre-trained. The specific training process is as follows:

[0078] Step 1: First, select a general domain large model as the base model, then collect sample data to build a training data set, and pre-train the base model based on the training data set to obtain the initial large model in the insurance field, that is, to obtain the initial model in the insurance field.

[0079] The sample data in this step includes: insurance question-and-answer dialogues, insurance product introductions, insurance contract introductions, claims clauses, claims cases and other data information. Specifically, several sample questions and answer labels corresponding to each sample question can be extracted based on the insurance question-and-answer dialogues, insurance product introductions, insurance contract introductions, insurance clauses, claims cases and other data, and then the base model is pre-trained based on each sample question and each answer label to obtain an initial large model.

[0080] Step 2: Perform supervised fine-tuning / training on the initial large model in the insurance field obtained in the first step to obtain the target question-answering model.

[0081] In this step, each sample question, the sample history conversation information corresponding to each sample question, and the answer label corresponding to each sample question can be collected in advance. Thus, based on each sample history conversation information, keywords such as the ones obtained can be extracted, and then sample prompt sentences can be generated based on each keyword. Finally, the sample prompt sentences and the corresponding sample questions are input into the initial large model, and the initial answer is generated by the initial large model; then, the model parameters of the initial large model are adjusted based on each initial answer and the answer label to obtain the adjusted current large model, and the current large model is reused based on the sample prompt sentence and the corresponding sample question to re-output the current answer corresponding to each sample question, until the current large model is determined to meet the predetermined training conditions based on each current answer and the label answer, the model training is stopped, and the current large model is used as the target question-answering model.

[0082] That is to say, in this step, instruction, output, and history in the instruction data set are mainly used to train the target question-answering model. Instruction represents the user's question, output represents the output, that is, the agent's reply, and history represents the historical conversation data between the customer and the agent, that is, the key information set. For example:

[0083] The historical conversation data is history:

[0084] [Customer: Hello.

[0085] Agent: Hello, is this Ms. xxx?

[0086] Customer: Hi, where are you?

[0087] Agent: Hello, ma'am. I'm the follow-up specialist. I'm calling you because I saw that you purchased a Blue Whale million medical insurance policy for yourself on June 30. Do you still remember this insurance product?

[0088] Client: Um, what?

[0089] Agent: Yes, well, we have both electronic and paper policies here. Have you received both?

[0090] Customer: No, I don’t know that.

[0091] Agent: No, um, go ahead. Okay.

[0092] Customer: Yes, where is the electronic insurance policy?

[0093] Agent: Did our service specialist add your WeChat account through the official corporate WeChat account? Customer: Yes, yes, yes. What happened? ]

[0094] The user's question is: [Yes, yes, um, what's wrong? ].

[0095] The sample answer output is: [If yes, has he ever directed you to follow our insurance official WeChat public account? ].

[0096] In this embodiment, by pre-collecting the historical conversation information of each sample and then performing secondary training on the model based on the historical conversations of each sample and the sample questions, the target question-answering model obtained through training can be made more accurate, laying the foundation for the subsequent accurate and reasonable generation of the target answer based on the target question-answering model.

[0097] Based on the above embodiment, another embodiment of the present application provides a question-answer processing device, such as Figure 2 As shown, including:

[0098] A collection module 11, for collecting historical conversation information with a target user in response to a conversation interruption instruction;

[0099] A construction module 12, configured to construct a dialogue knowledge base based on the historical dialogue information;

[0100] An acquisition module 13, used to acquire some key information from the dialogue knowledge base based on the current question of the current dialogue with the target user;

[0101] The processing module 14 is used to obtain a target question-answering model based on the key information and the current question by using pre-training, generate a target answer corresponding to the current question and output it.

[0102] In the specific implementation process of this embodiment, the construction module is specifically used to: extract key information from historical conversation information to obtain a number of key information; construct the conversation knowledge base based on each of the key information; the key information includes: any one or more of the following: user status information, user identity information and event information.

[0103] In the specific implementation process of this embodiment, the question and answer processing device also includes an updating module, and the updating module is used to: update the dialogue knowledge base based on the current dialogue information in real time to obtain an updated current dialogue knowledge base.

[0104] In the specific implementation process of this embodiment, the update module is specifically used to: extract semantic information from the current dialogue information to obtain a number of semantic relationship information and syntactic relationship information; based on the semantic relationship information and the syntactic relationship information, correct or supplement the key information in the dialogue knowledge base to update the dialogue knowledge base.

[0105] In the specific implementation process of this embodiment, the update module is specifically used to: extract structural information from the current dialogue information to obtain dialogue topic information, dialogue focus information and dialogue direction information; based on the dialogue topic information, dialogue focus information and dialogue direction information, correct or supplement key information in the dialogue knowledge base to update the dialogue knowledge base.

[0106] In the specific implementation process of this embodiment, the update module is specifically used to: collect the tracking point information in real time; correct or supplement the key information in the dialogue knowledge base based on the tracking point information to update the dialogue knowledge base.

[0107] In the specific implementation process of this embodiment, the processing module is specifically used to: construct a target prompt sentence based on the key information in the dialogue knowledge base; based on the target prompt sentence and the current question, use the target question-answering model to generate and output the target answer corresponding to the current question.

[0108] The question and answer processing device in the present application can save the historical conversation information by collecting the historical conversation information and constructing a conversation knowledge base when the conversation is truncated. Therefore, when conducting a subsequent conversation with the target user, it can accurately generate the corresponding target answer based on the current question and the historical conversation information in the conversation knowledge base, thereby avoiding the problem that in the course of a long conversation, due to the truncation of the conversation information, the subsequent conversation cannot be combined with the historical conversation, resulting in the output of the subsequent conversation answer being inaccurate and unreasonable.

[0109] Another embodiment of the present application provides a storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the following method steps are implemented:

[0110] Step 1: In response to the conversation interruption instruction, historical conversation information of the conversation with the target user is collected;

[0111] Step 2: constructing a dialogue knowledge base based on the historical dialogue information;

[0112] Step 3: Based on the current question of the current conversation with the target user, obtain some key information from the conversation knowledge base;

[0113] Step 4: Based on the key information and the current question, use the pre-trained target question-answering model to generate and output the target answer corresponding to the current question.

[0114] The specific implementation process of the above method steps can be found in the embodiments of any of the above-mentioned question and answer processing methods, and this embodiment will not be repeated here.

[0115] The storage medium in the present application can save the historical conversation information by collecting the historical conversation information and constructing a conversation knowledge base when the conversation is truncated. Therefore, in the subsequent conversation with the target user, the corresponding target answer can be accurately generated based on the current question and the historical conversation information in the conversation knowledge base, thereby avoiding the problem that the subsequent conversation cannot be combined with the historical conversation due to the truncation of the conversation information during a long conversation, resulting in the output of the subsequent conversation answer being inaccurate and unreasonable.

[0116] Another embodiment of the present application provides an electronic device, such as Figure 3 As shown, at least a memory 1 and a processor 2 are included, the memory 1 stores a computer program, and the processor 2 implements the following method steps when executing the computer program on the memory 1:

[0117] Step 1: In response to the conversation interruption instruction, historical conversation information of the conversation with the target user is collected;

[0118] Step 2: constructing a dialogue knowledge base based on the historical dialogue information;

[0119] Step 3: Based on the current question of the current conversation with the target user, obtain some key information from the conversation knowledge base;

[0120] Step 4: Based on the key information and the current question, use the pre-trained target question-answering model to generate and output the target answer corresponding to the current question.

[0121] The specific implementation process of the above method steps can be found in the embodiments of any of the above-mentioned question and answer processing methods, and this embodiment will not be repeated here.

[0122] The storage medium in the present application can save the historical conversation information by collecting the historical conversation information and constructing a conversation knowledge base when the conversation is truncated. Therefore, in the subsequent conversation with the target user, the corresponding target answer can be accurately generated based on the current question and the historical conversation information in the conversation knowledge base, thereby avoiding the problem that the subsequent conversation cannot be combined with the historical conversation due to the truncation of the conversation information during a long conversation, resulting in the output of the subsequent conversation answer being inaccurate and unreasonable.

[0123] The above embodiments are only exemplary embodiments of the present application and are not intended to limit the present application. The protection scope of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and protection scope of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present application.

Claims

1. A question-answering processing method, characterized in that: include: In response to the conversation interruption instruction, collecting historical conversation information with the target user; Building a dialogue knowledge base based on the historical dialogue information; Based on the current question of the current conversation with the target user, obtain some key information from the conversation knowledge base; Based on each of the key information and the current question, a target question-answering model is obtained by using pre-training, and a target answer corresponding to the current question is generated and output.

2. The method according to claim 1, characterized in that The step of constructing a conversation knowledge base based on historical conversation information specifically includes: Extract key information from historical conversation information to obtain some key information; Building the dialogue knowledge base based on each of the key information; The key information includes: any one or more of the following: user status information, user identity information and event information.

3. The method according to claim 1, characterized in that The method further comprises: The dialogue knowledge base is updated in real time based on the current dialogue information to obtain an updated current dialogue knowledge base.

4. The method according to claim 3, characterized in that The real-time updating of the dialogue knowledge base based on the current dialogue information specifically includes: Extracting semantic information from the current dialogue information to obtain some semantic relationship information and syntactic relationship information; Based on the semantic relationship information and the syntactic relationship information, key information in the dialogue knowledge base is corrected or supplemented to update the dialogue knowledge base.

5. The method according to claim 3, characterized in that The real-time updating of the dialogue knowledge base based on the current dialogue information specifically includes: Extracting structural information from the current conversation information to obtain conversation topic information, conversation focus information, and conversation direction information; Based on the conversation topic information, conversation focus information and conversation direction information, key information in the conversation knowledge base is corrected or supplemented to update the conversation knowledge base.

6. The method according to claim 1, characterized in that The method further comprises: Collect buried point information in real time; Based on the embedded point information, key information in the dialogue knowledge base is corrected or supplemented to update the dialogue knowledge base.

7. The method according to claim 1, characterized in that The method of obtaining a target question-answering model based on each of the key information and the current question by using pre-training, generating a target answer corresponding to the current question and outputting the answer specifically includes: Construct target prompt sentences based on key information in the dialogue knowledge base; Based on the target prompt sentence and the current question, the target question-answering model is used to generate and output a target answer corresponding to the current question.

8. A question-answer processing device, characterized in that: include: A collection module, configured to collect historical conversation information with a target user in response to a conversation interruption instruction; A construction module, used to construct a dialogue knowledge base based on the historical dialogue information; An acquisition module, used to acquire some key information from the dialogue knowledge base based on the current question of the current dialogue with the target user; A processing module is used to obtain a target question-answering model based on the key information and the current question, generate a target answer corresponding to the current question and output it.

9. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the question and answer processing method described in any one of claims 1 to 7 are implemented.

10. An electronic device, characterized in that: The system at least comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the question-answering processing method according to any one of claims 1 to 7 when executing the computer program on the memory.

Images