Response corpus generation method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention discloses a response corpus generation method and device, electronic equipment and a storage medium. The method comprises the following steps: identifying an inquiry intention of an acquirer of an online inquiry service; for a first large language model which is obtained by learning at least one first cue word in advance and is used for realizing response corpus generation, determining a first target cue word matched with the inquiry intention from the at least one first cue word, and inputting the first target cue word into the first large language model; and in response to detection of incremental question and answer content output by the acquirer, inputting the incremental question and answer content into the first large language model, so as to generate a response corpus used for responding to the incremental question and answer content according to the first large language model based on the first target prompt word and the incremental question and answer content. According to the technical scheme provided by the embodiment of the invention, the response corpus matched with the personalized incremental question and answer content of the acquirer can be generated, and the inquiry efficiency can be effectively improved.

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of Internet medical technology, and in particular, to a method, apparatus, electronic device, and storage medium for generating response corpora. Background Art

[0002] In recent years, with the rapid development of Internet technology and the increasing health demands of the people, online consultation platforms have developed rapidly, and the online consultations provided by them have become one of the main channels for medical consultations.

[0003] It should be noted that doctors usually use fragmented time to provide online consultation services for users. Therefore, it is crucial to allow doctors to have time to answer the condition in detail and serve more users, and improve the consultation efficiency of doctors. In this regard, online consultation platforms provide auxiliary diagnosis tools for doctors to use to improve the consultation efficiency.

[0004] In the process of implementing the present invention, the inventors found the following technical problems in the prior art: The currently used auxiliary diagnosis tools cannot enable doctors to obtain response corpora that match the personalized incremental Q&A content of users, which results in the inability to effectively improve the consultation efficiency of doctors and needs to be improved. Summary of the Invention

[0005] Embodiments of the present invention provide a method, apparatus, electronic device, and storage medium for generating response corpora to generate response corpora that match the personalized incremental Q&A content of the acquirer (such as a user) of online consultation services, thereby effectively improving the consultation efficiency of the provider (such as a doctor) of online consultation services.

[0006] According to one aspect of the present invention, a method for generating response corpora is provided, which may include:

[0007] Identifying the consultation intention of the acquirer of online consultation services;

[0008] For a first large language model obtained by pre-learning at least one first prompt word and used to implement the generation of response corpora, determining a first target prompt word that matches the consultation intention from at least one first prompt word, and inputting the first target prompt word into the first large language model;

[0009] In response to detecting the incremental Q&A content output by the acquirer, inputting the incremental Q&A content into the first large language model to generate a response corpus for answering the incremental Q&A content based on the first target prompt word and the incremental Q&A content according to the first large language model.

[0010] According to another aspect of the present invention, a device for generating response corpora is provided, which may include:

[0011] An inquiry intention recognition module, configured to recognize the inquiry intention of the acquirer of the online inquiry service;

[0012] A first target prompt word input module, configured to determine, for a first large language model obtained by pre-learning at least one first prompt word and used for generating response corpus, a first target prompt word that matches the inquiry intention from at least one first prompt word, and input the first target prompt word into the first large language model;

[0013] A response corpus generation module, configured to, in response to detecting incremental Q&A content output by the acquirer, input the incremental Q&A content into the first large language model, and generate, according to the first large language model, based on the first target prompt word and the incremental Q&A content, a response corpus for responding to the incremental Q&A content.

[0014] According to another aspect of the present invention, there is provided an electronic device, which may include:

[0015] At least one processor; and

[0016] A memory communicatively connected to the at least one processor; wherein,

[0017] The memory stores a computer program executable by the at least one processor. When the computer program is executed by the at least one processor, the at least one processor is caused to implement the response corpus generation method provided in any embodiment of the present invention.

[0018] According to another aspect of the present invention, there is provided a computer-readable storage medium, on which computer instructions are stored, and the computer instructions are used to cause a processor to implement the response corpus generation method provided in any embodiment of the present invention when executed.

[0019] In the technical solution of the embodiments of the present invention, the inquiry intention of the acquirer of the online inquiry service is recognized; then, for a first large language model obtained by pre-learning at least one first prompt word and used for generating response corpus, a first target prompt word that matches the inquiry intention can be determined from at least one first prompt word, and the first target prompt word is input into the first large language model; on this basis, in response to detecting incremental Q&A content output by the acquirer, the incremental Q&A content can be input into the first large language model, so that the first large language model can be used to generate a response corpus based on the first target prompt word and the incremental Q&A content. The above technical solution generates a response corpus that matches the inquiry intention and is used to respond to the incremental Q&A content by using the first large language model based on the first target prompt word that matches the inquiry intention and the incremental Q&A content output by the acquirer. This response corpus matches the acquirer's personalized incremental Q&A content, thus helping to effectively improve the inquiry efficiency.

[0020] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present invention, nor is it used to limit the scope of the present invention. Other features of the present invention will become readily understood from the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0022] Figure 1 is a flowchart of a method for generating response corpus according to an embodiment of the present invention;

[0023] Figure 2 is a schematic diagram of an example of displaying response corpus in a method for generating response corpus according to an embodiment of the present invention;

[0024] Figure 3 is a flowchart of another method for generating response corpus according to an embodiment of the present invention;

[0025] Figure 4 is a flowchart of an example of assisting in interrogation in another method for generating response corpus according to an embodiment of the present invention;

[0026] Figure 5 is a flowchart of yet another method for generating response corpus according to an embodiment of the present invention;

[0027] Figure 6 is a flowchart of an example of model training in yet another method for generating response corpus according to an embodiment of the present invention;

[0028] Figure 7 is a flowchart of still another method for generating response corpus according to an embodiment of the present invention;

[0029] Figure 8 is a schematic diagram of an example of displaying explanatory content in still another method for generating response corpus according to an embodiment of the present invention;

[0030] Figure 9 is a schematic diagram of an example of generating explanatory content in still another method for generating response corpus according to an embodiment of the present invention;

[0031] Figure 10 is a block diagram of the structure of a device for generating response corpus according to an embodiment of the present invention;

[0032] Figure 11It is a schematic structural diagram of an electronic device for implementing the response corpus generation method according to an embodiment of the present invention. Detailed implementation manners

[0033] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0034] It should be noted that in the description and claims of the present invention and the above accompanying drawings, the terms "first", "second", etc. are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. The same is true for "target", "original", etc., which will not be elaborated here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0035] It should be noted that in the technical solution of the present invention, in terms of the collection, acquisition, update, analysis, processing, use, transmission, storage, etc. of user personal information, they all comply with the provisions of relevant laws and regulations, are used for legal purposes, and do not violate public order and good customs. Necessary measures are taken for user personal information to prevent illegal access to user personal information data and maintain the security of user personal information and network security.

[0036] Before introducing the embodiments of the present invention, an exemplary explanation will be given first of the reasons for the appearance that the current auxiliary diagnosis tools in the background technology cannot enable doctors to obtain response corpus that matches the user's personalized incremental Q&A content, so as to better understand the reasons why the embodiments of the present invention can enable doctors to obtain such response corpus.

[0037] Exemplarily, the currently used auxiliary diagnosis tools include an interrogation form tool and an inference tool. Among them, the interrogation form tool provides doctors with interrogation forms formulated based on different disease symptoms. Doctors can select an interrogation form that matches the user's disease symptoms from these interrogation forms and send it to the user for step-by-step filling. The inference tool can obtain the interaction content between the doctor and the user, screen out the matching corpus that matches the interaction content from the historical corpus of online consultations, and then push the matching corpus to the doctor.

[0038] Thus, it can be seen that the above-mentioned auxiliary diagnosis tools directly use existing data (such as pre-formulated interrogation forms or historical corpora) as the response corpus for answering users, rather than generating targeted response corpora according to the user's personalized incremental Q&A content. This naturally makes it impossible for doctors to obtain response corpora that are more matched to the user's personalized incremental Q&A content, thus unable to effectively improve the doctor's interrogation efficiency.

[0039] To solve the above problems, the embodiments of the present invention propose an implementation solution for generating a response corpus that matches the user's personalized incremental Q&A content, which helps to better assist doctors in interrogation and thus can effectively improve the interrogation efficiency. The following will elaborate on this in detail.

[0040] Figure 1 It is a flowchart of a response corpus generation method provided by an embodiment of the present invention. This embodiment is applicable to the situation where a provider of an online consultation service generates a response corpus that can be used to answer an acquirer of the online consultation service. In particular, the response corpus is a corpus generated according to the acquirer's personalized incremental Q&A content and belongs to a generative corpus. This method can be executed by a response corpus generation device provided by an embodiment of the present invention. The device can be implemented in software and / or hardware, and the device can be integrated on an electronic device, which can be various user terminals or servers.

[0041] See Figure 1 , the method of the embodiment of the present invention specifically includes the following steps:

[0042] S110. Identify the interrogation intention of the acquirer of the online consultation service.

[0043] Among them, the acquirer can be understood as the party that obtains the online consultation service, that is, the user described above. Combining the application scenarios that the embodiments of the present invention may involve, the user can be, for example, a patient and their associated personnel. The associated personnel can be understood as the personnel related to the patient, such as at least one of the patient's family members, friends, classmates, and colleagues, and no specific limitation is made here.

[0044] Corresponding to the acquirer, the provider can be understood as the party providing the online consultation service, that is, the party that interacts with the acquirer online during the online consultation stage. Considering the application scenarios that the embodiments of the present invention may involve, optionally, the provider may be a doctor, for example.

[0045] The consultation intention can be understood as the specific requirement of the acquirer to initiate the current online consultation service. Considering the application scenarios that the embodiments of the present invention may involve, optionally, it may be at least one of symptom consultation, disease consultation, test report interpretation, medication consultation, and life advice, etc.

[0046] During the process where the provider provides the online consultation service so that the acquirer can obtain the online consultation service, the consultation intention is identified. In practical applications, the consultation intention can be identified in various ways. Optionally, the main complaint content of the acquirer is obtained, and the consultation intention is identified by analyzing the main complaint content; additionally, optionally, the interaction content between the acquirer and the provider is obtained, and the consultation intention is identified by analyzing the interaction content; of course, the consultation intention can also be identified by other means, which are not specifically limited herein.

[0047] It should be noted that in practical applications, the number of consultation intentions identified based on this step may be one or more, which is related to the actual situation and is not specifically limited herein. In the case of multiple consultation intentions, optionally, the following steps can be executed based on a single consultation intention, some consultation intentions, or all consultation intentions among the multiple consultation intentions, which are not specifically limited herein.

[0048] S120. For a first large language model obtained by pre-learning at least one first prompt and used to generate response corpus, determine a first target prompt that matches the consultation intention from the at least one first prompt, and input the first target prompt into the first large language model.

[0049] Among them, the first large language model can be understood as a large language model (i.e., a large model) obtained by pre-learning at least one first prompt. The first prompt can be understood as a prompt (i.e., a prompt) that enables the first large language model to have the function of generating response corpus through learning. The number of the first prompts can be one or more, and various first prompts can be preset respectively based on the corresponding consultation intentions.

[0050] It should be noted that compared with machine learning models with a small number of parameters, machine learning models with a large number of parameters (i.e., large language models) can better output generative response corpus due to their powerful context semantic understanding ability. Therefore, the embodiments of the present invention generate response corpus based on large language models.

[0051] On this basis, it further needs to be explained that in large language models, the main role of the prompt is to provide the context of the input information and the parameter information input into the model to the large language model. The setting of the prompt has a great effect and influence on the content output by the large language model. An appropriate prompt can make the large language model output more appropriate and accurate content.

[0052] Therefore, after identifying the interrogation intention, a first prompt word that matches the interrogation intention can be determined from at least one first prompt word. Here, this first prompt word is referred to as the first target prompt word. Then, the first target prompt word is input into the first large language model so that the first large language model generates a response corpus that matches the interrogation intention based on the first target prompt word.

[0053] S130. In response to detecting the incremental Q&A content output by the acquirer, the incremental Q&A content is input into the first large language model, so as to generate, according to the first large language model, a response corpus for answering the incremental Q&A content based on the first target prompt word and the incremental Q&A content.

[0054] Among them, in the online interrogation stage, the acquirer can output Q&A content at intervals or continuously according to its own interrogation needs. This Q&A content can be understood as the question of the inquirer or the answer to the question raised by the answer provider. On this basis, in order to distinguish the Q&A content that has been answered by the provider and the Q&A content that has not been answered by the provider, so that the first large language model can generate a response corpus corresponding to the Q&A content that has not been answered by the provider. Considering that compared with the Q&A content that has been answered by the provider, the Q&A content that has not been answered by the provider is the incremental Q&A content newly added by the acquirer. Therefore, here this newly added Q&A content can be referred to as incremental Q&A content. Further, in order to assist the provider in answering the acquirer, in response to detecting the incremental Q&A content output by the acquirer, the incremental Q&A content can be input into the first large language model, so that the first large language model can be used to generate a response corpus that matches the interrogation intention and is used to answer the incremental Q&A content based on the first target prompt word and the incremental Q&A content.

[0055] It can be understood that since the response corpus is generated according to the incremental Q&A content output by the acquirer, the response corpus can be updated in real time according to the change of the incremental Q&A content.

[0056] In the technical solution of the embodiment of the present invention, by identifying the consultation intention of the acquirer of the online consultation service; then, for the first large language model obtained by pre-learning at least one first prompt word and used to implement the generation of response corpus, the first target prompt word matching the consultation intention can be determined from at least one first prompt word, and the first target prompt word is input into the first large language model; on this basis, in response to detecting the incremental question-and-answer content output by the acquirer, the incremental question-and-answer content can be input into the first large language model, so that the first large language model can be used to generate a response corpus based on the first target prompt word and the incremental question-and-answer content. The above technical solution generates a response corpus that matches the consultation intention for answering the incremental question-and-answer content by using the first large language model based on the first target prompt word that matches the consultation intention and the incremental question-and-answer content output by the acquirer. This response corpus matches the acquirer's personalized incremental question-and-answer content, which helps to effectively improve the consultation efficiency.

[0057] An optional technical solution, the above response corpus generation method further includes:

[0058] Sending the response corpus to the doctor's end so that the doctor's end displays the response corpus, and in response to a selection operation input for the displayed response corpus, determining the target corpus selected in the response corpus to send and display the target corpus on the user's end.

[0059] Among them, the doctor's end can be understood as the client used by the provider (i.e., the doctor). The user's end can be understood as the client used by the acquirer (i.e., the user).

[0060] Sending the response corpus to the doctor's end so that the doctor's end receives and displays the response corpus. Exemplarily, Figure 2 A response corpus display example is shown, in which the response corpus available for the doctor to select is shown. On this basis, the doctor can determine whether there is the required response corpus among the displayed response corpora, and if so, can select the response corpus. In this way, the doctor's end can determine the target corpus selected in the response corpus by responding to the selection operation input for the displayed response corpus, and send and display the target corpus on the user's end, so that the user (i.e., the user) of the user's end can obtain the reply given by the doctor for the incremental question-and-answer content output by the user himself.

[0061] The above technical solution, by sending and displaying the response corpus on the doctor's end, can assist the doctor in selecting the required target corpus from the response corpus, and then sending the target corpus to the user's end to reply to the user. Compared with the doctor directly manually inputting the response corpus, this can effectively improve the doctor's consultation efficiency.

[0062] Figure 3It is a flowchart of another response corpus generation method provided in an embodiment of the present invention. This embodiment is optimized based on the above-mentioned technical solutions. In this embodiment, optionally, identifying the inquiry intention of the acquirer of the online consultation service includes: obtaining the main complaint content of the acquirer of the online consultation service; analyzing the main complaint content to identify the inquiry intention of the acquirer. Among them, the explanations of the terms that are the same as or corresponding to the above-mentioned embodiments are not repeated here.

[0063] See also Figure 3 The method of this embodiment may specifically include the following steps:

[0064] S210, obtaining the main complaint of the party receiving the online medical consultation service, analyzing the main complaint, and identifying the medical consultation intention of the party receiving the service.

[0065] The main complaint content can be understood as the content output by the acquirer to characterize its own symptoms or / signs, nature and duration, etc. Since the main complaint content can intuitively reflect the specific demands of the acquirer for initiating this online consultation service, the consultation intention can be identified by analyzing the main complaint content.

[0066] S220. For a first large language model obtained by pre-learning at least one first prompt word and used to generate a response corpus, determine a first target prompt word that matches the intention of the medical consultation from the at least one first prompt word, and input the first target prompt word into the first large language model.

[0067] S230. In response to detecting the incremental question and answer content output by the acquirer, the incremental question and answer content is input into the first large language model to generate a response corpus for responding to the incremental question and answer content based on the first target prompt word and the incremental question and answer content according to the first large language model.

[0068] The technical solution of the embodiment of the present invention realizes accurate identification of the intention of the medical consultation by analyzing the main complaint of the acquirer.

[0069] In an optional technical solution, before inputting the first target prompt word into the first large language model, the response corpus generation method further includes:

[0070] Based on the content of the chief complaint, modify the first target cue word;

[0071] The modified first target prompt word is used as the first target prompt word.

[0072] Among them, based on the content of the chief complaint, the first target prompt is modified so that the first target prompt can not only reflect the intention of the interrogation, but also further reflect the more detailed chief complaint content under this interrogation intention. Such a first target prompt can assist the first large language model in generating response corpora that are more matched to the personalized incremental Q&A content of the acquirer, thereby further improving the interrogation efficiency.

[0073] Exemplarily, assuming the interrogation intention is symptom interrogation, the first target prompt matching symptom interrogation can be "The user's chief complaint is..., you are now a doctor in an Internet hospital, and you need to have multiple rounds of interactive conversations with the patient to consult about the condition. During the condition consultation, you need to ask the patient about the possible symptoms, including the duration of the symptoms, the development of the condition, the intensity and frequency. In addition, you also need to ask about the aggravating and alleviating factors of the disease, related symptoms, whether the patient has received treatment, the specific treatment plan, and the treatment effect, and give a preliminary diagnosis and advice at the end of the conversation."

[0074] On this basis, further, assuming the chief complaint content is "I caught a cold recently and my throat hurts badly", then the chief complaint content can be filled into the first target prompt, and the following first target prompt can be obtained: "The user's chief complaint is I caught a cold recently and my throat hurts badly, you are now a doctor in an Internet hospital, and you need to have multiple rounds of interactive conversations with the patient to consult about the condition. During the condition consultation, you need to ask the patient about the possible symptoms, including the duration of the symptoms, the development of the condition, the intensity and frequency. In addition, you also need to ask about the aggravating and alleviating factors of the disease, related symptoms, whether the patient has received treatment, the specific treatment plan, and the treatment effect, and give a preliminary diagnosis and advice at the end of the conversation."

[0075] The above technical solution sets the first target prompt by combining the interrogation intention and the chief complaint content. This first target prompt can better assist the first large language model in generating response corpora that are more matched to the personalized incremental Q&A content of the acquirer, thereby further improving the interrogation efficiency.

[0076] To better understand the above technical solutions as a whole, the following will give an exemplary description in combination with specific examples. Exemplarily, see Figure 4 , the online interrogation starts, and the user outputs the chief complaint content; analyze the chief complaint content to identify the user's interrogation intention; based on the interrogation intention and the chief complaint content, set the first target prompt and input the first target prompt into the first large language model. Thus, the first large language model can generate and output corresponding response corpora according to the first target prompt and the incremental Q&A content output by the user until the online interrogation ends, thereby realizing targeted response recommendations for the entire interrogation process.

[0077] On this basis, exemplarily, Table 1 shows a doctor-patient conversation record (i.e., the interaction content between the doctor and the patient), and Table 2 shows the response corpus applied in this doctor-patient conversation record.

[0078] Table 1 Doctor-Patient Conversation Record

[0079]

[0080]

[0081] Table 2 Response Corpus Applied in the Doctor-Patient Conversation Record Shown in Table 1

[0082]

[0083]

[0084] In the above example, during the entire process of online consultation services from condition collection, disease diagnosis, medication guidance to life advice, doctors can view the response corpus output by the first large language model, which helps doctors quickly select the target corpus they need from these response corpora to interact with users.

[0085] Figure 5 It is a flowchart of another response corpus generation method provided in an embodiment of the present invention. This embodiment is optimized based on the above technical solutions. In this embodiment, optionally, the first large language model is pre-learned through the following steps: for each of at least one first prompt word, obtain first consultation data that matches the first prompt word; based on the first consultation data and the first prompt word, train a deep learning model so that the deep learning model learns the first consultation data and the first prompt word to obtain the first large language model; where the network structure of the first large language model is the same as that of the deep learning model. Among them, the explanations of the same or corresponding terms as those in the above embodiments are not repeated here.

[0086] See Figure 5 , the method of this embodiment may specifically include the following steps:

[0087] S310. For each of at least one first prompt word, obtain first consultation data that matches the first prompt word.

[0088] Among them, for each of at least one first prompt word, the first consultation data can be understood as the data generated during the historical consultation process that matches the first prompt word, and in particular, can be understood as the data generated due to the consultation intention corresponding to the first prompt word during the historical consultation process. The historical consultation process can be an online consultation process or an offline consultation process that has occurred in the past, and no specific limitation is made here.

[0089] Obtain the first consultation data. In practical applications, optionally, all-department consultation data can be obtained, and then high-quality first consultation data can be screened from the all-department consultation data according to different dimensional indicators. On this basis, as can be seen from the above description, the consultation intention is a dimensional indicator.

[0090] S320. Train a deep learning model based on the first consultation data and the first prompt word, so that the deep learning model learns the first consultation data and the first prompt word to obtain a first large language model for realizing the generation of response corpus; wherein, the network structure of the first large language model is the same as that of the deep learning model.

[0091] Among them, for the deep learning model with the same network structure as the to-be-trained first large language model, train the deep learning model based on the first consultation data and the first prompt word. In this way, after the model training is completed based on different first prompt words and the corresponding first consultation data, the model can learn how to generate response corpus based on different first prompt words, thereby completing the model response training.

[0092] S330. Identify the consultation intention of the acquirer of the online consultation service, determine the first target prompt word that matches the consultation intention from at least one first prompt word, and input the first target prompt word into the first large language model.

[0093] S340. In response to detecting the incremental Q&A content output by the acquirer, input the incremental Q&A content into the first large language model, so as to generate response corpus for answering the incremental Q&A content based on the first large language model, the first target prompt word, and the incremental Q&A content.

[0094] The technical solution of the embodiment of the present invention trains the model through the first prompt word and the first consultation data that matches the first prompt word, thereby obtaining a first large language model with the function of generating response corpus, and realizing the effective training of the first large language model.

[0095] An optional technical solution is that the first large language model is also pre-learned through the following steps:

[0096] For the provider of the online consultation service, obtain the second consultation data of the provider, and perform feature extraction on the second consultation data to obtain the consultation habit characteristics of the provider;

[0097] Train a deep learning model based on the consultation habit characteristics, so that the deep learning model learns the consultation habit characteristics to obtain the first large language model.

[0098] Among them, the second consultation data can be understood as the data output by the provider during the historical consultation process. This data can intuitively reflect the provider's consultation habits, which can be, for example, at least one of the consultation process, habitual expressions, and medication habits, etc. Obtain the second consultation data.

[0099] Extract features from the second consultation data to obtain the provider's consultation habit features. Exemplarily, Table 3 exemplarily shows the feature dimensions for feature extraction and the consultation habit features extracted under these feature dimensions respectively. Then, a deep learning model can be trained based on the consultation habit features, and thus a first large language model that can output response corpora matching the provider's consultation habits is obtained.

[0100] Table 3 Feature Dimensions and Consultation Habit Features

[0101]

[0102] In the above technical solution, by analyzing the provider's consultation habits and training the model based on the obtained consultation habit features, the response corpora generated by the first large language model thus obtained conform to the consultation habits. Compared with fixed phrases, this anthropomorphic phraseology can improve the consultation experience of both the provider and the acquirer.

[0103] On this basis, optionally, before inputting the first target prompt into the first large language model, the above response corpus generation method further includes:

[0104] Modify the first target prompt based on the consultation habit features;

[0105] Use the modified first target prompt as the first target prompt.

[0106] Among them, modifying the first target prompt based on the consultation habit features means setting the first target prompt by combining the consultation intention and the consultation habit features. Such a first target prompt can not only assist the first large language model in generating response corpora that are more matched with the personalized incremental Q&A content of the acquirer, but also assist the first large language model in generating response corpora that are more in line with the provider's consultation habits, thus improving the provider's consultation experience.

[0107] To better understand each technical solution in the embodiments of the present invention as a whole, the following uses specific examples to exemplarily illustrate it. Exemplarily, see Figure 6, collect the first inquiry data and the second inquiry data; extract features from the second inquiry data based on the feature dimensions representing inquiry habits to obtain inquiry habit features; then, perform model training based on the first inquiry data and the inquiry habit features, and on this basis, perform model evaluation and iteration. The response corpus generated by the obtained first large language model can not only accurately respond to personalized incremental question-and-answer content but also be more in line with inquiry habits.

[0108] Before introducing the following embodiments, an exemplary description of its application scenarios is given to better understand the technical effects that the following embodiments can achieve. Exemplarily, after a doctor outputs target content (such as diagnosis content and medication content, etc.), the user may have questions about the target content and need to further ask the doctor. At this time, the doctor can respond to the user based on the response corpus generated by the first large language model for this.

[0109] On this basis, to further improve the doctor's inquiry efficiency, after the online inquiry platform detects that the doctor outputs target content, it can set a jump link on the target content. For example, set a jump link on the diagnosis content of hypertension, so as to guide the user to click on the jump link to view the popular science content of hypertension. However, such popular science content is relatively conventional and not personalized content provided for the user's actual inquiry process, which will directly affect the user's inquiry experience and inquiry effect and needs to be improved.

[0110] Figure 7 It is a flowchart of another response corpus generation method provided in the embodiments of the present invention. This embodiment is optimized based on the above technical solutions. In this embodiment, optionally, the above response corpus generation method may further include: for the provider of the online inquiry service, in response to detecting the target content output by the provider, analyze the interaction content between the acquirer and the provider to obtain key content; where the target content includes at least one of diagnosis content and medication content; obtain a second large language model; where the second large language model is obtained by pre-learning at least one second prompt word and is used to implement the generation of explanatory content; determine a second target prompt word that matches the target content from at least one second prompt word, and modify and update the second target prompt word based on the key content; input the second target prompt word into the second large language model to generate explanatory content for explaining the target content according to the second large language model based on the second target prompt word.

[0111] See Figure 7 , the method of this embodiment may specifically include the following steps:

[0112] S410. Identify the inquiry intention of the acquirer of the online inquiry service.

[0113] S420. For a first large language model obtained by pre-learning at least one first prompt and used to implement response corpus generation, determine a first target prompt that matches the consultation intention from the at least one first prompt, and input the first target prompt into the first large language model.

[0114] S430. In response to detecting the incremental Q&A content output by the acquirer, input the incremental Q&A content into the first large language model, so as to generate a response corpus for answering the incremental Q&A content based on the first large language model, the first target prompt, and the incremental Q&A content.

[0115] S440. For the provider of the online consultation service, in response to detecting the target content output by the provider, analyze the interaction content between the acquirer and the provider to obtain key content; wherein, the target content includes at least one of a diagnosis content and a medication content.

[0116] Wherein, the target content can be understood as the content that needs to be further explained and illustrated output by the provider. Combining with the application scenarios that the embodiments of the present invention may involve, optionally, the target content may be, for example, at least one of a diagnosis content and a medication content, etc. On this basis, further, the diagnosis content can be understood as the content related to disease diagnosis, and the medication content can be understood as the content related to medication advice.

[0117] The interaction content can be understood as the content generated due to the interaction between the acquirer and the provider during the online consultation process. Combining with the application scenarios that the embodiments of the present invention may involve, optionally, the interaction content can be presented in at least one of the ways such as graphics, text, voice, and video, and no specific limitation is made here.

[0118] In response to detecting the target content, in order to facilitate the acquirer's understanding of the target content, it is necessary to further explain and illustrate the target content in combination with the acquirer's actual consultation process. Therefore, the interaction content is analyzed here to obtain key content that can be used to reflect the key information in the actual consultation process.

[0119] In practical applications, optionally, the key content can directly come from the interaction content, or can be obtained by summarizing the interaction content, etc., and no specific limitation is made here.

[0120] Optionally, the key content can be obtained through various methods, such as Named Entity Recognition (NER), content classification, keyword extraction, sentiment analysis, etc., which are not specifically limited herein. Taking NER as an example, the interaction content between the acquirer and the provider can be obtained, and NER is performed on the interaction content to obtain the entity content in the interaction content, and then the entity content is used as the key content. The application of NER realizes the accurate extraction of the key content.

[0121] S450. Obtain a second large language model; wherein, the second large language model is obtained by pre-learning at least one second prompt and is used to generate explanatory content.

[0122] Among them, the second large language model can be understood as a large language model (i.e., a large model) obtained by pre-learning at least one second prompt. The second prompt can be understood as a prompt (i.e., a prompt) that enables the second large language model to learn to have the function of generating explanatory content. The number of the second prompts can be one or more, and various second prompts can be preset respectively based on corresponding target content.

[0123] It should be noted that the first large language model and the second large language model can be the same or different large language models, which can be set according to actual needs and are not specifically limited herein. It can be understood that when the first large language model and the second large language model are the same large language model, it means that this large language model has both the function of generating response corpus and the function of generating explanatory content.

[0124] S460. Determine a second target prompt that matches the target content from at least one second prompt, and modify and update the second target prompt based on the key content.

[0125] Among them, similar to the processing process of the first prompt, a second target prompt that matches the target content can be determined from at least one second prompt. On this basis, optionally, different second prompts can be preset for different departments in advance, so that the second target prompt can be jointly determined in combination with the target content and the consultation department, making the second target prompt more in line with the actual consultation process. Optionally, the content types corresponding to each target content can be analyzed in advance, and then corresponding second prompts can be preset for each content type, so that the second target prompt can be determined according to the detected content type of the target content.

[0126] Furthermore, in order to enable the second large language model to generate not only explanatory content that matches the target content, but also explanatory content that matches the actual consultation process, the second target prompt can be modified based on the key content, and the second target prompt can be updated based on the modification result.

[0127] S470. Input the second target prompt into the second large language model, so as to generate, according to the second large language model and based on the second target prompt, an explanatory content for explaining the target content.

[0128] Among them, by inputting the second target prompt into the second large language model, it is possible to utilize the second large language model to generate, based on the second target prompt, an explanatory content for explaining the target content and matching the actual consultation process (i.e., personalized). Combining with the application scenarios that the embodiments of the present invention may involve, optionally, when the target content is a diagnosis content, the explanatory content may be disease explanations and precautions, etc.; further optionally, when the target content is a medication content, the explanatory content may be medication guidance and medication instructions, etc.

[0129] On this basis, optionally, after generating the explanatory content, the explanatory content may be sent to the user terminal so that the user terminal displays the explanatory content, thereby enabling the user to understand the target content in a timely manner. Exemplarily, as Figure 8 shown, after detecting the diagnosis content of indigestion, the online consultation platform generates the corresponding explanatory content and directly sends the explanatory content to the user terminal for display in the doctor-patient conversation.

[0130] The technical solution of the embodiments of the present invention, in response to detecting the target content output by the provider, can analyze the interaction content for reflecting the actual consultation process to obtain the key content, and then set the second target prompt based on the key content and the target content, so that the second large language model can be utilized to generate an explanatory content that can personalize the explanation of the target content in combination with the actual consultation process, thereby ensuring the consultation experience and consultation effect of the acquirer.

[0131] In order to better understand each technical solution in the embodiments of the present invention as a whole, the following uses specific examples to conduct an exemplary description. Exemplarily, as Figure 9 shown, after detecting the diagnosis content and medication content output by the doctor, the online consultation platform identifies the interaction content between the doctor and the patient based on the NER technology to obtain the entity content (i.e., the key content). Exemplarily, Table 4 exemplarily shows the interaction content between the doctor and the patient, and Table 5 exemplarily shows the entity content obtained based on this interaction content.

[0132] Further, determine the second target prompt words corresponding to the diagnosis content and medication content, and modify the second target prompt words based on the key content. Thus, the setting of the second target prompt words is completed. Exemplarily, assume the second target prompt words corresponding to the diagnosis content and medication content are as follows: "You are now a dermatologist. Based on the patient's chief complaint symptoms (...), location (...), time (...), give the diagnosis as (...), recommend the medication (...). You need to give the diagnosis reason and rehabilitation precautions for the disease, and give the efficacy description and precautions for the medication, so that the user can better understand and follow your advice."

[0133] Then, modify the above second target prompt words based on the entity content shown in Table 5 to obtain the finally set second target prompt words: "You are now a dermatologist. Based on the patient's chief complaint symptoms (red pimples, bumps, itching), location (on the face, on the skin), time (half a month), give the diagnosis as (allergic dermatitis), recommend the medication (Drug A). You need to give the diagnosis reason and rehabilitation precautions for the disease, and give the efficacy description and precautions for the medication, so that the user can better understand and follow your advice."

[0134] Input the second target prompt words into the second large language model, so that the second large language model can be used to generate personalized description content based on the second target prompt words. This description content includes the disease description for explaining the disease content and the medication description for explaining the medication content shown in Table 6 for the interaction content.

[0135] Then, display the description content in the doctor-patient conversation and prompt the user to read it, so that the user can better understand the disease and medication information, which helps the user to receive treatment in the correct way for recovery.

[0136] In the above example, by using the second large language model and combining the interaction content between doctors and patients, targeted disease descriptions and medication descriptions can be given to the user, which helps the user to better understand the disease and medication plan, and thus contributes to the effective treatment and recovery of the disease.

[0137] Table 4 Example of the interaction content between doctors and patients

[0138]

[0139]

[0140] Table 5 Entity content corresponding to the interaction content shown in Table 4

[0141] Department Dermatology Department Disease Pityriasis rosea, eczema, allergic dermatitis Symptom Red pimples, bumps, itching Time Half a month Location On the face, on the skin Medicine Drug A

[0142] Table 6 Description content corresponding to the entity content shown in Table 5

[0143]

[0144]

[0145] Figure 10 This is the structural block diagram of the response corpus generation device provided in the embodiments of the present invention. This device is used to execute the response corpus generation method provided in any of the above embodiments. This device and the response corpus generation methods of the above embodiments belong to the same inventive concept. For the details not described in detail in the embodiments of the response corpus generation device, reference can be made to the embodiments of the above response corpus generation method. Refer to Figure 10 , specifically, this device may include: an inquiry intention recognition module 510, a first target prompt word input module 520, and a response corpus generation module 530. Among them,

[0146] The inquiry intention recognition module 510 is used to recognize the inquiry intention of the acquirer of the online inquiry service;

[0147] The first target prompt word input module 520 is used to determine, for a first large language model obtained by pre-learning at least one first prompt word and used to implement response corpus generation, a first target prompt word that matches the inquiry intention from at least one first prompt word, and input the first target prompt word into the first large language model;

[0148] The response corpus generation module 530 is used to, in response to detecting the incremental question-and-answer content output by the acquirer, input the incremental question-and-answer content into the first large language model, and generate a response corpus for responding to the incremental question-and-answer content based on the first large language model, the first target prompt word, and the incremental question-and-answer content.

[0149] Optionally, the inquiry intention recognition module 510 may include:

[0150] The main complaint content acquisition unit is used to acquire the main complaint content of the acquirer of the online inquiry service;

[0151] The inquiry intention recognition unit is used to analyze the main complaint content and recognize the inquiry intention of the acquirer.

[0152] On this basis, optionally, the above response corpus generation device may further include:

[0153] The first modification module for the first target prompt word is used to modify the first target prompt word based on the main complaint content before inputting the first target prompt word into the first large language model;

[0154] The first update module for the first target prompt word is used to use the modified first target prompt word as the first target prompt word.

[0155] Optionally, the first large language model is pre-trained through the following modules:

[0156] The first medical interview data acquisition module is used to obtain, for each of at least one first prompt word, first medical interview data that matches the first prompt word;

[0157] The first large language model first acquisition module is used to train a deep learning model based on the first medical interview data and the first prompt word, so that the deep learning model learns the first medical interview data and the first prompt word to obtain the first large language model; wherein, the network structure of the first large language model is the same as that of the deep learning model.

[0158] On this basis, optionally, the first large language model is also pre-trained through the following modules:

[0159] The medical interview habit feature acquisition module is used to obtain, for the provider of the online medical interview service, second medical interview data of the provider, and extract features from the second medical interview data to obtain the medical interview habit features of the provider;

[0160] The first large language model second acquisition module is used to train a deep learning model based on the medical interview habit features, so that the deep learning model learns the medical interview habit features to obtain the first large language model.

[0161] On this basis, optionally, the above response corpus generation device may further include:

[0162] The first target prompt word second modification module is used to modify the first target prompt word based on the medical interview habit features before inputting the first target prompt word into the first large language model;

[0163] The first target prompt word second update module is used to use the modified first target prompt word as the first target prompt word.

[0164] Based on any of the above devices, optionally, the device may further include:

[0165] The response corpus sending module is used to send the response corpus to the doctor side, so that the doctor side displays the response corpus, and in response to a selection operation input for the displayed response corpus, determines the target corpus selected in the response corpus, and sends and displays the target corpus on the user side.

[0166] Optionally, the above response corpus generation device may further include:

[0167] A key content obtaining module, which can be used for the provider of online consultation services to analyze the interaction content between the acquirer and the provider in response to detecting the target content output by the provider, and obtain the key content; wherein, the target content includes at least one of diagnostic content and medication content;

[0168] A second large language model obtaining module, which is used to obtain a second large language model; wherein, the second large language model is obtained by pre-learning at least one second prompt word and is used to implement the generation of explanatory content;

[0169] A second target prompt word updating module, which is used to determine a second target prompt word that matches the target content from at least one second prompt word, and modify and update the second target prompt word based on the key content;

[0170] An explanatory content generating module, which is used to input the second target prompt word into the second large language model, and generate explanatory content for explaining the target content based on the second large language model and the second target prompt word.

[0171] On this basis, optionally, the key content obtaining module may include:

[0172] An entity content obtaining unit, which is used to obtain the interaction content between the acquirer and the provider, and perform named entity recognition on the interaction content to obtain the entity content in the interaction content;

[0173] A key content obtaining unit, which is used to use the entity content as the key content.

[0174] Another option is that the above response corpus generating device may further include:

[0175] An explanatory content sending module, which is used to send the explanatory content to the user terminal so that the user terminal displays the explanatory content.

[0176] The response corpus generation device provided by the embodiment of the present invention identifies the consultation intention of the acquirer of the online consultation service through the consultation intention recognition module; then, through the first target prompt word input module, for the first large language model obtained by pre-learning at least one first prompt word and used to generate the response corpus, the first target prompt word matching the consultation intention can be determined from at least one first prompt word, and the first target prompt word is input into the first large language model; on this basis, through the response corpus generation module, in response to detecting the incremental Q&A content output by the acquirer, the incremental Q&A content can be input into the first large language model, so that the first large language model can be used to generate the response corpus based on the first target prompt word and the incremental Q&A content. The above device generates a response corpus for answering the incremental Q&A content that matches the consultation intention by using the first large language model based on the first target prompt word that matches the consultation intention and the incremental Q&A content output by the acquirer. This response corpus matches the acquirer's personalized incremental Q&A content, which helps to effectively improve the consultation efficiency.

[0177] The response corpus generation device provided by the embodiment of the present invention can execute the response corpus generation method provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method.

[0178] It should be noted that in the embodiments of the above response corpus generation device, the included units and modules are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific names of the functional units are only for the convenience of mutual distinction and do not limit the protection scope of the present invention.

[0179] Figure 11 FIG. shows a schematic structural diagram of an electronic device 10 that can be used to implement the embodiments of the present invention. The electronic device is intended to represent various forms of digital computers, such as, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, personal digital processors, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are only examples and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0180] As Figure 11As shown, the electronic device 10 includes at least one processor 11 and a memory communicatively connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc. The memory stores a computer program executable by the at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 into the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0181] Multiple components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disc, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0182] The processor 11 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 11 include but are not limited to a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as the response corpus generation method.

[0183] In some embodiments, the response corpus generation method can be implemented as a computer program tangibly embodied in a computer-readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the response corpus generation method described above can be executed. Alternatively, in other embodiments, the processor 11 can be configured to execute the response corpus generation method by any other appropriate means (e.g., by means of firmware).

[0184] The various embodiments of the systems and techniques described above in this specification can be implemented in digital electronic circuitry, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems-on-chip (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which may be a special-purpose or general-purpose programmable processor that receives data and instructions from, and transmits data and instructions to, a storage system, at least one input device, and at least one output device.

[0185] The computer program for implementing the methods of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the computer programs, when executed by the processor, cause the functions / operations specified in the flowchart and / or block diagram to be implemented. The computer program can be executed entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or server.

[0186] In the context of the present invention, a computer-readable storage medium can be a tangible medium that can contain or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. The computer-readable storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, the computer-readable storage medium can be a machine-readable signal medium. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0187] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and a pointing device (e.g., a mouse or a trackball) through which the user can provide input to the electronic device. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0188] The systems and techniques described herein can be implemented in a computing system including backend components (e.g., as a data server), or a computing system including middleware components (e.g., an application server), or a computing system including frontend components (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system including any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected to each other by digital data communication in any form or medium (e.g., a communication network). Examples of communication networks include: local area network (LAN), wide area network (WAN), blockchain network, and the Internet.

[0189] The computing system can include a client and a server. The client and the server are generally far from each other and usually interact through a communication network. The relationship between the client and the server is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system and solves the defects of difficult management and weak business scalability existing in traditional physical hosts and VPS services.

[0190] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps recited in the present invention can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved, and no limitation is made herein.

[0191] The above specific embodiments do not constitute a limitation on the protection scope of the present invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for generating response corpus, characterized in that, Including: Identifying the consultation intention of the acquirer of the online consultation service; For a first large language model obtained by pre-learning at least one first prompt word and used to implement response corpus generation, determining a first target prompt word that matches the consultation intention from the at least one first prompt word, and inputting the first target prompt word into the first large language model; In response to detecting the incremental Q&A content output by the acquirer, inputting the incremental Q&A content into the first large language model, so as to generate a response corpus for answering the incremental Q&A content based on the first large language model, the first target prompt word, and the incremental Q&A content.

2. The method according to claim 1, wherein The identifying the consultation intention of the acquirer of the online consultation service includes: Obtaining the main complaint content of the acquirer of the online consultation service; Analyzing the main complaint content to identify the consultation intention of the acquirer.

3. The method according to claim 2, characterized in that, Before inputting the first target prompt word into the first large language model, it further includes: Modifying the first target prompt word based on the main complaint content; Taking the modified first target prompt word as the first target prompt word.

4. The method according to claim 1, wherein The first large language model is pre-learned through the following steps: For each first prompt word in the at least one first prompt word, obtaining first consultation data that matches the first prompt word; Training a deep learning model based on the first consultation data and the first prompt word, so that the deep learning model learns the first consultation data and the first prompt word to obtain the first large language model; wherein, the network structure of the first large language model is the same as that of the deep learning model.

5. The method according to claim 4, characterized in that, The first large language model is also pre-learned through the following steps: For the provider of the online consultation service, obtaining second consultation data of the provider and extracting features from the second consultation data to obtain the consultation habit features of the provider; Training the deep learning model based on the consultation habit features, so that the deep learning model learns the consultation habit features to obtain the first large language model.

6. The method according to claim 5, characterized in that, Before inputting the first target prompt word into the first large language model, it further includes: Modifying the first target prompt word based on the consultation habit features; Taking the modified first target prompt word as the first target prompt word.

7. The method according to any one of claims 1-6, characterized in that, It further includes: Sending the response corpus to the doctor side, so that the doctor side displays the response corpus, and in response to a selection operation input for the displayed response corpus, determining the target corpus selected in the response corpus, and sending and displaying the target corpus on the user side.

8. The method according to claim 1, characterized in that, It further includes: For the provider of the online consultation service, in response to detecting the target content output by the provider, analyzing the interaction content between the acquirer and the provider to obtain key content; wherein, the target content includes at least one of diagnostic content and medication content. Obtain the second largest language model; wherein, the second largest language model is obtained by pre-learning at least one second prompt word and is used to implement the generation of explanatory content; Determine a second target prompt word that matches the target content from the at least one second prompt word, and modify and update the second target prompt word based on the key content; Input the second target prompt word into the second largest language model to generate explanatory content for explaining the target content based on the second target prompt word according to the second largest language model.

9. The method according to claim 8, characterized in that The analysis of the interaction content between the acquirer and the provider to obtain key content includes: Obtain the interaction content between the acquirer and the provider, and perform named entity recognition on the interaction content to obtain the entity content in the interaction content; Use the entity content as the key content.

10. The method according to claim 8 or 9, characterized in that, Further includes: Send the explanatory content to the user terminal so that the user terminal displays the explanatory content.

11. A response corpus generation device, characterized in that, Includes: An inquiry intention recognition module for recognizing the inquiry intention of the acquirer of the online inquiry service; A first target prompt word input module for determining, from the at least one first prompt word, a first target prompt word that matches the inquiry intention for a first large language model obtained by pre-learning at least one first prompt word and used to implement the generation of response corpus, and inputting the first target prompt word into the first large language model; A response corpus generation module for, in response to detecting incremental question-and-answer content output by the acquirer, inputting the incremental question-and-answer content into the first large language model to generate response corpus for answering the incremental question-and-answer content based on the first large language model, the first target prompt word, and the incremental question-and-answer content.

12. An electronic device, characterized in that, Includes: At least one processor; And A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to cause the at least one processor to execute the response corpus generation method according to any one of claims 1-10.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a processor to implement the response corpus generation method according to any one of claims 1-10 when executed.

Images