Talk skill process generation method and device, electronic equipment and storage medium

Abstract

The invention provides a verbal skill process generation method and device, electronic equipment and a storage medium. The method comprises the steps that a preset intention recognition model and a dialogue data set are obtained, and the dialogue data set comprises multiple sections of dialogue data; according to an intention recognition model, determining a dialogue intention of each piece of dialogue sub-data in each section of dialogue data; according to the dialogue intention of each piece of dialogue sub-data and the dialogue sequence of each piece of dialogue sub-data in the corresponding dialogue data, a dialogue intention sequence of each piece of dialogue data is generated, and the dialogue intention sequence comprises the dialogue sub-data of the dialogue data arranged according to the dialogue sequence and the dialogue intention of each piece of dialogue sub-data; determining the dialogue intention category of each dialogue intention in each dialogue intention sequence; and according to the dialogue intention category of each dialogue intention in each dialogue intention sequence and the dialogue intention sequence of each section of dialogue data, generating a target dialogue skill process. The target verbal skill process generated by the method is simpler and more efficient, and the practicability of the verbal skill process can be improved.

Description

Technical Field

[0001] Embodiments of the present disclosure relate to the field of artificial intelligence technology, and particularly to a method, device, electronic device, and storage medium for generating a conversation flow chart. Background Art

[0002] In recent years, with the intensification of market competition and the complexity of marketing scenarios, the demand of enterprises for automated marketing technology has increased significantly. Especially in the generation of marketing conversation flow charts, there is an urgent need to provide standardized process guidance for salespersons through efficient and accurate dialogue data analysis, so as to reduce the cost of manual decision-making and improve the conversion efficiency.

[0003] The current technology mainly realizes the generation of marketing conversation flow charts through unsupervised clustering algorithms. The core process is as follows: First, based on shallow features such as the word frequency and semantic similarity of dialogue texts, a large amount of dialogue content is divided into several categories. Then, through manual verification of the clustering results for intent, label correction, and process logic sorting, a conversation flow chart is finally generated.

[0004] However, due to the influence of algorithm complexity and data feature selection in unsupervised clustering algorithms, it is easy to over-segment the dialogue content in the same business stage into multiple categories or confuse the dialogue in different stages, resulting in redundant nodes and chaotic logic in the finally generated conversation flow chart, which requires manual repeated merging and adjustment. Moreover, unsupervised clustering algorithms only rely on surface semantic similarity and cannot recognize the true intent in the dialogue. Subsequently, a large amount of manpower is required to perform intent annotation and verification on the clustering results, which not only increases the time cost but also has a risk of deviation in process design due to differences in manual subjective judgment.

[0005] Therefore, it is necessary to propose a method for generating a conversation flow chart to solve at least one of the above technical problems. Summary of the Invention

[0006] Embodiments of the present disclosure propose a method, device, electronic device, and storage medium for generating a conversation flow chart, which can accurately recognize the intent in dialogue data based on an intent recognition model, thereby significantly reducing manual intervention and improving the overall efficiency. By processing the recognized intent into a conversation intent sequence, the logic is clearer, which is convenient for subsequent analysis and the generation of the target conversation flow chart. And generating the target conversation flow chart based on the conversation intent category can remove the intent that is useless or redundant for the target conversation flow chart, making the generated target conversation flow chart more concise and efficient, thereby improving the practicality of the conversation flow chart.

[0007] In a first aspect, the present disclosure provides a method for generating a conversation flow chart, including:

[0008] Obtaining a preset intent recognition model and a dialogue data set, where the dialogue data set includes multiple segments of dialogue data;

[0009] Determine the conversation intent of each conversation sub - data in each piece of the conversation data according to the said intent recognition model;

[0010] Generate a conversation intent sequence for each piece of the conversation data according to the conversation intent of each conversation sub - data and the conversation order of each conversation sub - data in the corresponding conversation data, where the conversation intent sequence includes the conversation sub - data arranged in the conversation order of the conversation data and the conversation intents of each conversation sub - data;

[0011] Determine the conversation intent category of each conversation intent in each conversation intent sequence;

[0012] Generate a target conversation flow according to the conversation intent category of each conversation intent in each conversation intent sequence and the conversation intent sequence of each piece of the conversation data.

[0013] In some alternative embodiments, the generating a conversation intent sequence for each piece of the conversation data according to the conversation intent of each conversation sub - data and the conversation order of each conversation sub - data in the corresponding conversation data includes:

[0014] Cluster each conversation intent according to the similarity between each conversation intent to obtain at least one conversation intent cluster;

[0015] For each conversation intent cluster, establish a conversation intent cluster label for each conversation intent cluster;

[0016] Generate the conversation intent sequence of each piece of the conversation data according to each conversation intent cluster label and the conversation order of each conversation sub - data in each piece of the conversation data.

[0017] In some alternative embodiments, the generating a target conversation flow according to the conversation intent category of each conversation intent in each conversation intent sequence and the conversation intent sequence of each piece of the conversation data includes:

[0018] Identify the process intent and / or the question - answer intent in the conversation intent sequence of each piece of the conversation data according to the conversation intent category;

[0019] Generate a target conversation flow according to the conversation intent sequence of each piece of the conversation data and the process intent and / or the question - answer intent.

[0020] In some alternative embodiments, the generating a target conversation flow according to the conversation intent sequence of each piece of the conversation data and the question - answer intent includes:

[0021] Generate a Q&A intention knowledge base based on the conversation intention sequence and the Q&A intention of the conversation data described in each paragraph;

[0022] Delete the Q&A intention and the corresponding conversation sub-data in the conversation intention sequence of the conversation data described in each paragraph;

[0023] Generate a process conversation intention sequence for each paragraph of the conversation data based on the conversation intention sequence of the conversation data after deleting the Q&A intention and the corresponding conversation sub-data;

[0024] Generate the target conversation flow according to the process conversation intention sequence of each paragraph of the conversation data.

[0025] In some alternative embodiments, the generating the target conversation flow according to the conversation intention sequence and the process intention of each paragraph of the conversation data includes:

[0026] Generate a process conversation intention sequence for each paragraph of the conversation data according to the front-back order of each process intention in the conversation intention sequence of each paragraph of the conversation data;

[0027] Generate the target conversation flow according to the process conversation intention sequence of each paragraph of the conversation data.

[0028] In some alternative embodiments, the method further includes:

[0029] Delete the process intention and the corresponding conversation sub-data in the conversation intention sequence of each paragraph of the conversation data;

[0030] Generate the Q&A intention knowledge base based on the conversation intention sequence of each paragraph of the conversation data after deleting the process intention and the corresponding conversation sub-data.

[0031] In some alternative embodiments, the generating the target conversation flow according to the conversation intention sequence, the process intention, and the Q&A intention of each paragraph of the conversation data includes:

[0032] Generate a Q&A intention knowledge base based on the conversation intention sequence and the Q&A intention of each paragraph of the conversation data;

[0033] Generate a process conversation intention sequence for each paragraph of the conversation data according to the front-back order of each process intention in the conversation intention sequence of each paragraph of the conversation data;

[0034] Generate the target conversation flow according to the process conversation intention sequence of each paragraph of the conversation data.

[0035] In some alternative embodiments, generating a Q&A intent knowledge base based on the conversation intent sequence and the Q&A intent of each segment of the conversation data includes:

[0036] According to the conversation attributes of each piece of conversation sub-data, dividing the conversation sub-data corresponding to the Q&A intent in each conversation intent sequence into question conversation sub-data and reply conversation sub-data;

[0037] Generating at least one Q&A pair according to the Q&A relationship between the question conversation sub-data and the reply conversation sub-data;

[0038] Generating a Q&A intent knowledge base based on at least one of the Q&A pairs.

[0039] In some alternative embodiments, generating the target conversation flow based on the process conversation intent sequence of each segment of the conversation data includes:

[0040] Based on the process conversation intent sequence of each segment of the conversation data, for each first process intent in the process conversation intent sequence, grouping the first process intent and the second process intents in the process conversation intent sequence, where the second process intents are located after the first process intent in the process conversation intent sequence, and the number of the second process intents is at least one;

[0041] Generating the target conversation flow chart based on the process intent multi-tuples generated from each segment of the conversation data.

[0042] In some alternative embodiments, the position of the second process intent is adjacent to the position of the first process intent.

[0043] In some alternative embodiments, when there are multiple second process intents, the positions of the second process intents are adjacent to each other.

[0044] In a second aspect, the present disclosure provides a conversation flow generation device, including:

[0045] An acquisition unit, configured to acquire a preset intent recognition model and a conversation data set, where the conversation data set includes multiple segments of conversation data;

[0046] An intent determination unit, configured to determine the conversation intent of each piece of conversation sub-data in each segment of the conversation data according to the intent recognition model;

[0047] A sequence generation unit, configured to generate a dialogue intention sequence for each piece of the dialogue data according to the dialogue intention of each piece of the dialogue sub-data and the dialogue order of each piece of the dialogue sub-data in the corresponding dialogue data, where the dialogue intention sequence includes the dialogue sub-data arranged in the dialogue order of the dialogue data and the dialogue intentions of the respective dialogue sub-data;

[0048] An intention category determination unit, configured to determine the dialogue intention category of each dialogue intention in each of the dialogue intention sequences;

[0049] A conversation flow generation unit, configured to generate a target conversation flow according to the dialogue intention category of each dialogue intention in each of the dialogue intention sequences and the dialogue intention sequence of each piece of the dialogue data.

[0050] In a third aspect, the present disclosure provides an electronic device, including:

[0051] One or more processors;

[0052] A storage device having one or more programs stored thereon,

[0053] When the above one or more programs are executed by the above one or more processors, the above one or more processors are caused to implement the method described in any one of the embodiments of the first aspect of the present disclosure.

[0054] In a fourth aspect, the present disclosure provides a computer-readable storage medium having a computer program stored thereon, where when the computer program is executed by one or more processors, the method described in any one of the embodiments of the first aspect of the present disclosure is implemented.

[0055] In a fifth aspect, the present disclosure provides a computer program product, including computer programs / instructions, where when the computer programs / instructions are executed by a processor, the method described in any one of the embodiments of the first aspect of the present disclosure is implemented.

[0056] The method, apparatus, electronic device, and storage medium for generating a conversation flow provided by an embodiment of the present disclosure first obtain a preset intent recognition model and a conversation dataset. The conversation dataset includes multiple segments of conversation data. According to the intent recognition model, the conversation intent of each conversation sub-data in each segment of conversation data is determined. Then, according to the conversation intent of each conversation sub-data and the conversation order of each conversation sub-data in the corresponding conversation data, a conversation intent sequence for each segment of conversation data is generated. The conversation intent sequence includes the conversation sub-data arranged in the conversation order and the conversation intent of each conversation sub-data. Next, the conversation intent category of each conversation intent in each conversation intent sequence is determined. Finally, according to the conversation intent category of each conversation intent in each conversation intent sequence and the conversation intent sequence of each segment of conversation data, a target conversation flow is generated. The present disclosure can accurately recognize the intent in the conversation data based on the intent recognition model, thereby significantly reducing manual intervention and improving the overall efficiency. Then, by processing the recognized intent into a conversation intent sequence, the logic is clearer, which is convenient for subsequent analysis and the generation of the target conversation flow. And generating the target conversation flow based on the conversation intent category can remove the intents that are useless or redundant for the target conversation flow, making the generated target conversation flow more concise and efficient, thus enhancing the practicality of the conversation flow. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Other features, objects, and advantages of the present disclosure will become more apparent by reading the following detailed description of non-limiting embodiments with reference to the accompanying drawings. The drawings are only for the purpose of showing the specific embodiments and are not considered as limiting the present invention. In the drawings:

[0058] Figure 1 is a system architecture diagram to which an embodiment of the method for generating a conversation flow according to the present disclosure can be applied;

[0059] Figure 2 is a flowchart of an embodiment of the method for generating a conversation flow according to the present disclosure;

[0060] Figure 3 is a decomposed flowchart of an embodiment of step 203 according to the present disclosure;

[0061] Figure 4 is a schematic diagram of a target conversation flow according to the present disclosure;

[0062] Figure 5 is a schematic structural diagram of an embodiment of the apparatus for generating a conversation flow according to the present disclosure;

[0063] Figure 6 is a schematic structural diagram of a computer system of an electronic device suitable for implementing the embodiments of the present disclosure. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0064] The present disclosure will be further described in detail below in conjunction with the accompanying drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than limiting the invention. Additionally, it should be noted that for the convenience of description, only the parts related to the relevant invention are shown in the accompanying drawings.

[0065] It should be noted that, without conflict, the embodiments in the present disclosure and the features in the embodiments may be combined with each other. The present disclosure will be described in detail below with reference to the accompanying drawings and embodiments.

[0066] Figure 1 An exemplary system architecture 100 showing embodiments of a conversation flow generation method, apparatus, electronic device, and storage medium to which the present disclosure can be applied is illustrated.

[0067] As Figure 1 shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 is used to provide a communication link between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various types of communication connection, such as a wired communication link, a wireless communication link, and so on.

[0068] Users may use the terminal devices 101, 102, 103 to interact with the server 105 through the network 104 to receive or send messages and the like. Various communication client applications may be installed on the terminal devices 101, 102, 103, such as conversation flow generation applications, intent recognition applications, voice interaction applications, video conferencing applications, short video social applications, web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, and so on.

[0069] The terminal devices 101, 102, and 103 can be hardware or software. When the terminal devices 101, 102, and 103 are hardware, they can be various electronic devices with microphones and speakers, including but not limited to smartphones, tablets, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 players (Moving Picture Experts Group Audio Layer IV), portable computers, desktop computers, and so on. When the terminal devices 101, 102, and 103 are software, they can be installed in the above-listed electronic devices. It can be implemented as multiple software or software modules (such as obtaining a preset intent recognition model and a dialogue data set), or it can be implemented as a single software or software module. No specific limitation is made here.

[0070] The server 105 can be a server that provides various services, such as a background server for processing the obtained preset intent recognition model and dialogue data set on the terminal devices 101, 102, and 103. The background server can perform corresponding processing based on the obtained preset intent recognition model and dialogue data set acquired by the terminal device.

[0071] In some cases, the conversation flow generation method provided by the present disclosure can be jointly executed by the terminal devices 101, 102, and 103 and the server 105. For example, the step of "obtaining a preset intent recognition model and a dialogue data set" can be executed by the terminal devices 101, 102, and 103, and the step of "determining the conversation intent of each conversation sub-data in each piece of conversation data according to the intent recognition model" can be executed by the server 105. The present disclosure does not make a limitation in this regard. Correspondingly, the conversation flow generation device can also be respectively arranged in the terminal devices 101, 102, and 103 and the server 105.

[0072] In some cases, the conversation flow generation method provided by the present disclosure can be executed by the server 105. Correspondingly, the conversation flow generation device can also be arranged in the server 105. In this case, the system architecture 100 may not include the terminal devices 101, 102, and 103 either.

[0073] In some cases, the conversation flow generation method provided by the present disclosure can be executed by the terminal devices 101, 102, and 103. Correspondingly, the conversation flow generation device can also be arranged in the terminal devices 101, 102, and 103. In this case, the system architecture 100 may not include the server 105 either.

[0074] It should be noted that the server 105 can be hardware or software. When the server 105 is hardware, it can be implemented as a distributed server cluster composed of multiple servers or as a single server. When the server 105 is software, it can be implemented as multiple software or software modules (for example, used to provide distributed services) or as a single software or software module. No specific limitation is made here.

[0075] It should be understood that Figure 1 the numbers of the terminal devices, the network, and the server in

[0076] The information, data, and signals involved in this disclosure are all authorized by users or fully authorized by all parties, and the collection, use, and processing of relevant data comply with the relevant laws, regulations, and standards of relevant countries and regions.

[0077] Continuing to refer to Figure 2 , Figure 2 FIG. 200 is a flowchart showing an embodiment of the method for generating a conversation flow according to the present disclosure. Figure 2 The shown method for generating a conversation flow can be applied to Figure 1 the terminal device or the server shown in

[0078] Step 201: Obtain a preset intent recognition model and a conversation dataset.

[0079] In this embodiment, the intent recognition model is used to recognize the conversation intent of each conversation, and the intent recognition model can include various known models for recognizing conversation intents.

[0080] For example, the intent recognition model can include: a rule-based model, a machine learning model, a deep learning model, a transfer learning / pre-trained language model, etc.

[0081] The conversation dataset can include multiple segments of conversation data, where one segment of conversation data can include a call record between a customer service staff and a customer.

[0082] Among them, the call record includes the speech content of the customer service staff and the speech content of the customer, and the speech content can include voice, text, images, etc.

[0083] Here, the multiple segments of conversation data in the conversation dataset can be multiple segments of conversation data for a specific scenario.

[0084] Step 202: Determine the conversation intent of each piece of conversation sub-data in each segment of conversation data according to the intent recognition model.

[0085] In this embodiment, the dialogue sub-data is a complete semantic unit with a dialogue intention generated by a single participant (customer or customer service) in a single speech within the dialogue data. The dialogue sub-data can include one or more sentences. A single participant can generate at least one complete semantic unit with a dialogue intention in a single speech, that is, a single participant can include at least one piece of dialogue sub-data in a single speech.

[0086] For example, in a dialogue data containing 9 single speeches (each single speech can be from the customer or the customer service), each complete semantic unit with a dialogue intention can be regarded as an independent piece of dialogue sub-data.

[0087] Example, the following is a piece of dialogue data, specifically:

[0088] Customer service: Hello, is this Mr. Wang?

[0089] Customer: Yes, this is me. Who are you?

[0090] Customer service: Hello, I'm the business manager of xxx. We have specially prepared a personal credit digital asset application plan for you... Do you have a need to apply for digital assets now?

[0091] Customer: How do I operate this?

[0092] Customer service: You log in to xxx... Just follow the prompts and submit the corresponding bill details.

[0093] Customer: Okay, I'll take a look.

[0094] Customer service: The digital asset quota will not incur any fees when not in use. However, we recommend that you perform a quota withdrawal operation immediately after obtaining the quota. This can not only skip the manual review process but also ensure that the digital assets are processed within two minutes at the earliest. In addition, in this way, you can retain the digital asset quota for a long time and achieve cyclic use. If you apply for and withdraw the quota within one hour, you can enjoy the above convenient services. We will send an activity text message to your mobile phone later, detailing the relevant steps and precautions.

[0095] Customer: Okay.

[0096] Customer service: Thank you for your support. Bye.

[0097] In the above dialogue data, among them, "Customer service: Hello, is this Mr. Wang?", "Customer: Yes, this is me. Who are you?", "Customer service: Hello, I'm the business manager of xxx. We have specially prepared a personal credit digital asset application plan for you... Do you have a need to apply for digital assets now?" can be regarded as four pieces of dialogue sub-data respectively.

[0098] In some alternative embodiments, a single participant may include at least one piece of conversation sub-data in a single speech. The at least one piece of conversation sub-data may be conversation sub-data with the same conversation intention or conversation sub-data with different conversation intentions.

[0099] Example 1, in the above conversation data, for the single speech of the customer "Customer: It's me. Who are you?", it includes two pieces of conversation sub-data "It's me" and "Who are you". Among them, the conversation intention corresponding to the conversation sub-data "It's me" is that the customer confirms it's himself / herself, and the conversation intention corresponding to the conversation sub-data "Who are you" is that the customer asks about the identity of the customer service.

[0100] The conversation intentions of the two pieces of conversation sub-data corresponding to the single speech of the above customer are both process intentions.

[0101] Example 2, in some other conversation data, for the single speech of the customer "Customer: I want to apply for this digital asset. Do I need to provide additional collateral assets during the processing?", it includes two pieces of conversation sub-data "I want to apply for this digital asset" and "Do I need to provide additional collateral assets during the processing?". Among them, the conversation intention corresponding to the conversation sub-data "I want to apply for this digital asset" is that the customer confirms the application, and the conversation intention corresponding to the conversation sub-data "Do I need to provide additional collateral assets during the processing?" is that the customer asks whether to provide collateral assets.

[0102] Among the two pieces of conversation sub-data corresponding to the single speech of the above customer, there is one process intention and one question-and-answer intention.

[0103] In this embodiment, according to the intention recognition model, the conversation intention of each piece of conversation sub-data in each conversation data is determined. Specifically, each conversation data can be input into the intention recognition model, and the conversation intention of each piece of conversation sub-data in each conversation data is output through the intention recognition model.

[0104] Among them, the conversation intention may refer to the purpose or requirement expressed by each conversation.

[0105] For example, in the above example, the conversation intentions of each piece of conversation sub-data may be respectively: the customer service greets the customer, the customer confirms it's himself / herself, the customer asks about the identity of the customer service, the customer service explains his / her identity and purpose, the customer asks how to operate, the customer service tells how to operate, the customer expresses willingness to operate, the customer service makes an activity invitation, the customer expresses affirmation, and the customer service bids farewell to the customer.

[0106] Step 203, generate a conversation intention sequence for each conversation data according to the conversation intention of each piece of conversation sub-data and the conversation order of each piece of conversation sub-data in the corresponding conversation data.

[0107] After obtaining the conversation intent of each conversation sub - data, in this step, the conversation intent sequence of each conversation data can be generated according to the conversation intent of each conversation sub - data and the conversation order of each conversation sub - data in the corresponding conversation data.

[0108] Here, processing each conversation sub - data and its corresponding conversation intent into a conversation intent sequence can make the logic of the entire conversation data clearer, facilitating subsequent analysis and the generation of the target conversation flow.

[0109] Among them, the conversation intent sequence includes the conversation sub - data arranged in the conversation order of the conversation data and the conversation intents of each conversation sub - data.

[0110] For example, the conversation intent sequence can be: {Customer service: Hello, is this Mr. Wang? ------- Customer service greets the customer; Customer: Yes, it is. ------- The customer confirms it's himself; Customer: Who are you? ------- The customer asks about the identity of the customer service; Customer service: Hello, I'm the business manager of xxx. We have specially prepared a personal credit digital asset application plan for you... Do you have a need for digital asset application now? ------- The customer service explains his identity and purpose; Customer: How do I operate this? ------- The customer asks how to operate; Customer service: You log in to the xxx APP... and follow the prompts to submit the corresponding bill details. ------- The customer service tells how to operate; Customer: Okay, I'll take a look. ------- The customer indicates willingness to operate; Customer service: The digital asset quota will not incur any fees when not in use. However, we recommend that you perform a quota withdrawal operation immediately after obtaining the quota. This can not only skip the manual review process and ensure that the digital asset is processed within two minutes at the earliest. In addition, in this way, you can retain the digital asset quota for a long time and achieve cyclic use. If you apply for and withdraw the quota within one hour, you can enjoy the above - mentioned convenient services. We will send an activity text message to your mobile phone later, detailing the relevant steps and precautions. ------- The customer service makes an activity invitation; Customer: Okay. ------- The customer indicates affirmation; Customer service: Thank you for your support. Bye. ------- The customer service bids farewell to the customer}.

[0111] It should be noted that a single participant's single speech can include at least one conversation sub - data. At least one conversation sub - data can be conversation sub - data with the same conversation intent or conversation sub - data with different conversation intents.

[0112] For this situation, the conversation intent sequence can be generated according to the order of each conversation sub - data in the single speech.

[0113] For example, "Customer: It's me. Who are you?" in the dialogue intention sequence can be: {It's me ------ The customer confirms it's themselves; Who are you ------ The customer asks about the identity of the customer service representative}.

[0114] "Customer: I want to apply for this digital asset. Do I need to provide additional collateral assets during the application process?"

[0115] In the dialogue intention sequence, it can be: {Customer: I want to apply for this digital asset ------ The customer confirms the application; Customer: Do I need to provide additional collateral assets during the application process? ------ The customer asks whether to provide collateral assets}.

[0116] In some alternative embodiments, step 203 may include the following steps 2031 - 2033.

[0117] It should be understood that for the dialogue intentions identified by the intention recognition model, there may be multiple similar names for the same dialogue intention. For example, for the dialogue intention of the customer service representative greeting the customer, it may also be recognized as: The customer service representative extends greetings to the customer, The customer service representative greets the customer, The customer service representative says hello to the customer, The customer service representative politely says hello to the customer, etc.

[0118] Here, the diverse naming of dialogue intentions may lead to ambiguity and potential ambiguity of dialogue intentions. To avoid dialogue intention ambiguity and recognition errors caused by different expressions, the following steps 2031 - 2033 can be used to cluster dialogue intentions with the same semantics, and then generate the dialogue intention sequence for each piece of dialogue data based on the clustered dialogue intentions.

[0119] Step 2031, cluster each dialogue intention according to the similarity between each dialogue intention to obtain at least one dialogue intention cluster.

[0120] In this step, each dialogue intention with the same semantics can be clustered according to the similarity between each dialogue intention to obtain at least one dialogue intention cluster.

[0121] In this embodiment, each dialogue intention cluster may include at least one dialogue intention.

[0122] Here, based on various known clustering algorithms, each dialogue intention can be clustered according to the similarity between each dialogue intention.

[0123] Clustering algorithms can include, for example: K - means algorithm, agglomerative hierarchical clustering, divisive hierarchical clustering, Mean Shift clustering, spectral clustering, etc.

[0124] Step 2032, for each dialogue intention cluster, establish a dialogue intention cluster label for each dialogue intention cluster.

[0125] For each dialogue intention cluster, establish a dialogue intention cluster label for each dialogue intention cluster. The dialogue intention label can be the identifier of the dialogue intention cluster, used to uniquely identify the dialogue intention cluster.

[0126] That is to say, the dialogue intention label can be the naming of dialogue intentions with the same semantics of the same type. For example, the semantics of multiple dialogue intentions such as the customer service greets the customer, the customer service sends greetings to the customer, the customer service greets the customer, the customer service says hello to the customer, and the customer service politely says hello to the customer are the same, which is that the customer service greets the customer. Then, these dialogue intentions can be grouped into a dialogue intention cluster. The dialogue intention cluster label corresponding to this dialogue intention cluster can be the name of one of the above dialogue intentions, or a label associated with the semantics of the above dialogue intentions can be set for these dialogue intentions again. In this way, by dividing each dialogue intention into at least one dialogue intention cluster and establishing a dialogue intention cluster label for it, the diversified naming of dialogue intentions can be reduced, the ambiguity and recognition errors caused by different expressions can be avoided, and subsequent processing can be made more convenient.

[0127] Step 2033: Generate a dialogue intention sequence for each segment of dialogue data according to each dialogue intention cluster label and the dialogue order of each dialogue sub-data in each segment of dialogue data.

[0128] In this step, based on the refined dialogue intention cluster label and the dialogue order of each dialogue sub-data in each segment of dialogue data, a dialogue intention sequence for each segment of dialogue data can be generated. In this way, the dialogue intentions in the dialogue data in the dialogue intention sequence can be made clearer, easier to understand and analyze.

[0129] Step 204: Determine the dialogue intention category of each dialogue intention in each dialogue intention sequence.

[0130] In this embodiment, the dialogue intention category can include a process intention and a question-and-answer intention. Determining the dialogue intention category of each dialogue intention in each dialogue intention sequence can refer to determining whether each dialogue intention in each dialogue sequence is a process intention or a question-and-answer intention.

[0131] Among them, the question-and-answer intention can refer to a request for information query or specific problem-solving. Through the dialogue sub-data corresponding to the question-and-answer intention, the customer's demand for factual information (such as product functions, operation steps, rule descriptions, etc.) can be quickly responded to.

[0132] For example, the question-and-answer intention can include: the customer asks how to operate, and the customer service tells how to operate.

[0133] The process intention can be a dialogue intention in the dialogue data used to guide or control the progress of the business process.

[0134] For example, the process intent may include: the customer service greets the customer, the customer confirms it is themselves, the customer asks about the identity of the customer service, the customer service explains their identity and purpose, the customer indicates willingness to operate, the customer service invites the customer to participate in an activity, the customer gives an affirmative response, and the customer service bids farewell to the customer.

[0135] Here, based on various existing dialogue intent category recognition models, the dialogue intent category of each dialogue intent in each dialogue intent sequence can be determined.

[0136] Step 205, generate a target conversation script based on the dialogue intent category of each dialogue intent in each dialogue intent sequence and the dialogue intent sequence of each segment of dialogue data.

[0137] In some alternative embodiments, the process intent and / or Q&A intent in the dialogue intent sequence of each segment of dialogue data can be recognized according to the dialogue intent category, and then, based on the dialogue intent sequence and the process intent and / or Q&A intent of each segment of dialogue data, a target conversation script is generated.

[0138] In some embodiments, for the dialogue intent sequence of each segment of dialogue data, it can first be recognized whether the dialogue intent corresponding to each dialogue sub-data is a process intent or a Q&A intent, and then, based on the process intent and / or Q&A intent and the dialogue intent sequence, a target conversation script is generated.

[0139] The target conversation script may refer to a series of standardized conversation steps or scripts designed to achieve a preset goal in interactive scenarios such as customer service and sales. The target conversation script is used to guide the customer to complete a task or solve a problem through orderly conversation steps, ensuring a smooth and efficient conversation process and achieving the expected result.

[0140] In some alternative embodiments, the above step 205 may include A1 - A4.

[0141] A1, generate a Q&A intent knowledge base based on the dialogue intent sequence and Q&A intent of each segment of dialogue data.

[0142] In this embodiment, for the dialogue intent sequence of each segment of dialogue data, the Q&A intent in each dialogue sub-data can first be recognized, and then, for the Q&A intent, a Q&A knowledge base can be generated.

[0143] In some alternative embodiments, specifically, A1 may include the following A11 - A13:

[0144] A11, according to the dialogue attributes of each dialogue sub-data, divide the dialogue sub-data corresponding to the Q&A intent in each dialogue intent sequence into question dialogue sub-data and reply dialogue sub-data.

[0145] Here, the dialogue attribute may refer to whether the dialogue sub - data belongs to a question or a reply in the dialogue sub - data corresponding to the Q&A intention. According to whether each dialogue sub - data belongs to a question or a reply, the dialogue sub - data corresponding to the reply intention in each dialogue intention can be divided into question dialogue sub - data and reply dialogue sub - data.

[0146] Among them, the question dialogue sub - data corresponds to a question, and the reply dialogue sub - data corresponds to a reply to the question.

[0147] For example, the question dialogue sub - data can be the dialogue sub - data corresponding to a customer asking how to operate.

[0148] The reply dialogue sub - data can be the dialogue sub - data corresponding to the customer service informing how to operate.

[0149] In some alternative embodiments, to determine the dialogue attribute of each dialogue sub - data, the language features of the dialogue sub - data can be analyzed.

[0150] In some alternative embodiments, to determine whether a dialogue sub - data is a question dialogue sub - data:

[0151] (1) It can be detected whether there are interrogative words in the dialogue sub - data: such as "ma", "how", "what", "how", "why", "when", "where", etc.

[0152] (2) Whether it is an interrogative sentence or a rhetorical question: "Can you...?" "Isn't it?"

[0153] (3) Whether it includes keywords with a consultative tone: "please", "excuse me", "want to know", etc.

[0154] In some alternative embodiments, it can be determined whether a dialogue sub - data is a question dialogue sub - data by at least one of (1), (2), and (3).

[0155] For example, "Customer: How do I operate this?" Among them, it includes both the interrogative word "how" and is an interrogative sentence, so it can be determined that the dialogue sub - data "How do I operate this?" is a question dialogue sub - data.

[0156] In some alternative embodiments, to determine whether a dialogue sub - data is a reply dialogue sub - data:

[0157] (1) Detect whether the dialogue sub - data provides information or viewpoints, that is, whether it is an answer, an explanation, or an elaboration of a viewpoint to a certain question.

[0158] (2) There are affirmative or negative expressions: those with clear judgments such as "yes", "no", "have", "haven't", etc. are often replies.

[0159] In some alternative embodiments, it can be determined whether a dialogue sub - data is a reply dialogue sub - data by at least one of (1) and (2).

[0160] For example, "Customer service: Please log in to the xxx APP... Just submit the corresponding bill details according to the prompts." Obviously, it provides information on how to guide the customer to operate. It can be determined that the dialogue sub-data "Please log in to the xxx APP... Just submit the corresponding bill details according to the prompts." is a reply dialogue sub-data.

[0161] In some alternative embodiments, to determine the dialogue attributes of each dialogue sub-data, pre-set Q&A tags can also be established for Q&A intents.

[0162] For example, establish Q&A tags for Q&A intents such as asking, requesting, and doubting as question dialogue sub-data, and establish Q&A tags for Q&A intents such as answering, explaining, and confirming as reply dialogue sub-data.

[0163] For example, for the Q&A intent: the customer asks how to operate, based on the pre-set Q&A tags, it can be determined that the dialogue sub-data "How do I operate this?" corresponding to the customer's question on how to operate is a question dialogue sub-data.

[0164] For the Q&A intent: the customer service tells how to operate, based on the pre-set Q&A tags, it can be determined that the dialogue sub-data "Please log in to the xxx APP... Just submit the corresponding bill details according to the prompts." corresponding to the customer service's instruction on how to operate is a reply dialogue sub-data.

[0165] In some alternative embodiments, to further ensure the accuracy of identifying the dialogue attributes of each dialogue sub-data, the dialogue attributes of each dialogue sub-data can be determined based on at least one of the above methods for determining the dialogue attributes of each dialogue sub-data.

[0166] It should be noted that the above methods for determining the dialogue attributes of each dialogue sub-data are only examples, and various known methods for determining whether a dialogue sub-data is a question or a reply can also be used to determine the dialogue attributes of each dialogue sub-data, which are not specifically limited here.

[0167] A12. Generate at least one Q&A pair according to the Q&A relationship between the question dialogue sub-data and the reply dialogue sub-data.

[0168] After classifying the dialogue sub-data corresponding to the reply intent in each dialogue intent into question dialogue sub-data and reply dialogue sub-data, at least one Q&A pair can be generated according to the question relationship between the question dialogue sub-data and the reply dialogue sub-data.

[0169] Here, the Q&A relationship refers to the logical association and interaction between a question and a reply during a dialogue.

[0170] For example, there is a logical association between the dialogue sub-data corresponding to the above customer's inquiry about how to operate and the dialogue sub-data corresponding to the customer service's response on how to operate. That is, for the question of how to operate asked by the customer, the corresponding answer is the customer service's response on how to operate.

[0171] Then, the dialogue sub-data corresponding to the customer's inquiry about how to operate and the dialogue sub-data corresponding to the customer service's response on how to operate can form a question-and-answer pair.

[0172] A13. Generate a question-and-answer intention knowledge base based on at least one question-and-answer pair.

[0173] Here, generating a question-and-answer intention knowledge base based on at least one question-and-answer pair may mean storing each question-and-answer pair to generate a question-and-answer intention knowledge base. The question-and-answer intention knowledge base can be used to obtain corresponding answers from the question-and-answer intention knowledge base when the intelligent customer service or the human customer service asks questions to answer the customer. By constructing the question-and-answer intention knowledge base, the response speed of the intelligent customer service or the human customer service can be improved, the accuracy of the answer can be ensured, and the customer experience can be enhanced.

[0174] A2. Delete the question-and-answer intention and the dialogue sub-data corresponding to the question-and-answer intention in the dialogue intention sequence of each piece of dialogue data.

[0175] Here, after identifying the question-and-answer intention in the dialogue intention sequence of the dialogue data and generating the question-and-answer intention knowledge base, the question-and-answer intention in the dialogue intention sequence of each piece of dialogue data and the dialogue sub-data corresponding to the question-and-answer intention can be deleted.

[0176] A3. Generate a process dialogue intention sequence for each piece of dialogue data based on the dialogue intention sequence of each piece of dialogue data after deleting the question-and-answer intention and the dialogue sub-data corresponding to the question-and-answer intention.

[0177] It can be understood that the dialogue intention includes the question-and-answer intention and the process intention. After deleting the question-and-answer intention and the corresponding dialogue sub-data, the remaining dialogue intention and dialogue sub-data in the dialogue intention sequence are the process intention and the dialogue sub-data corresponding to the process intention. Then, a process dialogue intention sequence for each piece of dialogue data can be generated based on the dialogue intention sequence of each piece of dialogue data after deleting the question-and-answer intention and the dialogue sub-data corresponding to the question-and-answer intention.

[0178] It should be understood that a single piece of dialogue data may not cover all necessary process intentions, resulting in an incomplete dialogue process or lack of key steps. Therefore, by analyzing multiple pieces of dialogue data and based on the process dialogue intention sequences of multiple pieces of dialogue data, common and complete process intentions can be identified to supplement the missing parts in a single piece of dialogue data, ensuring that each key step is covered and making the final generated target dialogue process more comprehensive and accurate.

[0179] Among them, the process dialogue intention sequence may include the process intentions of the dialogue data arranged in the dialogue order and the corresponding dialogue sub-data of the process intentions, or may include the process intentions arranged in the dialogue order without including the corresponding dialogue sub-data.

[0180] For example, the dialogue intention sequence may be: {Customer service: Hello, is this Mr. Wang? ------- The customer service greets the customer; Customer: Yes, this is me. ------- The customer confirms being himself; Who are you? ------- The customer asks about the identity of the customer service; Customer service: Hello, I'm the business manager of xxx. We have prepared a personal credit digital asset application plan especially for you... May I ask if you have a need to apply for digital assets now? ------- The customer service explains his identity and purpose; Customer: How do I operate this? ------- The customer asks how to operate; Customer service: You log in to the xxx APP... and submit the corresponding bill details according to the prompts. ------- The customer service tells how to operate; Customer: Okay, I'll take a look. ------- The customer indicates willingness to operate; Customer service: The digital asset quota will not incur any fees when not in use. However, we recommend that you perform a quota withdrawal operation immediately after obtaining the quota. This can not only skip the manual review process and ensure that the digital assets are processed within two minutes at the fastest. In addition, in this way, you can retain the digital asset quota for a long time and achieve cyclic use. If you apply for and withdraw the quota within one hour, you can enjoy the above convenient services. We will send an activity text message to your mobile phone later, detailing the relevant steps and precautions. ------- The customer service makes an activity invitation; Customer: Okay. ------- The customer indicates affirmation; Customer service: Thank you for your support. Bye. ------- The customer service bids farewell to the customer}.

[0181] Among them, the question-and-answer intention may include: The customer asks how to operate, and the customer service tells how to operate.

[0182] The corresponding dialogue sub-data of the question-and-answer intention may include: "How do I operate this?" "You log in to the xxx APP... and submit the corresponding bill details according to the prompts."

[0183] After deleting the Q&A intent and the corresponding dialogue sub-data in the dialogue intent sequence, the dialogue intent sequence can be: {Customer service: Hello, is this Mr. Wang? ------- The customer service greets the customer; Customer: Yes, this is me. ------- The customer confirms being the person; Who are you? ------- The customer asks about the identity of the customer service; Customer service: Hello, I am the business manager of xxx. We have specially prepared a personal credit digital asset application plan for you... May I ask if you have a need to apply for digital assets now? ------- The customer service explains their identity and purpose; Customer: Okay, let me take a look. ------- The customer indicates willingness to operate; Customer service: The digital asset quota will not incur any fees when not in use. However, we recommend that you perform a quota withdrawal operation immediately after obtaining the quota. This can not only skip the manual review process but also ensure that the digital assets are processed within two minutes at the earliest. In addition, in this way, you can retain the digital asset quota for a long time and achieve cyclic use. If you apply for and withdraw the quota within one hour, you can enjoy the above convenient services. We will send an activity text message to your mobile phone later, detailing the relevant steps and precautions. ------- The customer service makes an activity invitation; Customer: Okay. ------- The customer gives an affirmative response; Customer service: Thank you for your support. Bye. ------- The customer service bids farewell to the customer}.

[0184] In some alternative embodiments, when the process dialogue intent sequence includes process intents arranged in the dialogue order and the corresponding dialogue sub-data of the process intents, the dialogue intent sequence after deleting the Q&A intent and the corresponding dialogue sub-data can be determined as the process dialogue intent sequence.

[0185] In some alternative embodiments, when the process dialogue intent sequence consists of process intents arranged in the dialogue order without including the corresponding dialogue sub-data, the process intents in the dialogue intent sequence after deleting the Q&A intent and the corresponding dialogue sub-data can be extracted, and the process intents can be arranged in the order in the dialogue intent sequence to generate the process dialogue intent sequence. Or, based on the dialogue intent sequence after deleting the Q&A intent and the corresponding dialogue sub-data, the corresponding dialogue sub-data of the process intents can be further deleted while retaining the process intents to generate the process dialogue intent sequence.

[0186] For example, the process dialogue intent sequence can be: {The customer service greets the customer, the customer confirms being the person, the customer service explains their identity and purpose, the customer indicates willingness to operate, the customer service makes an activity invitation, the customer gives an affirmative response, the customer service bids farewell to the customer}.

[0187] In some alternative embodiments, the corresponding process intents in each process intent sequence and the dialogue sub-data corresponding to the process intents may also be stored in a preset process intent knowledge base. In this way, when an intelligent customer service or a human customer service conducts a conversation for a target conversation flow, the corresponding conversation content can be obtained from the process intent knowledge base and output to the customer.

[0188] A4. Generate a target conversation flow according to the process conversation intent sequence of each piece of conversation data.

[0189] After obtaining the process conversation intent sequence of each piece of conversation data, a target conversation flow can be generated according to the process conversation intent sequence of each piece of conversation data. Here, an answer-and-question intent knowledge base can be generated for the reply intent, and for the process intent, a process conversation intent sequence can be generated. Further, a target conversation flow is generated. In this way, by separately processing different categories of conversation intents, the intents that are useless or redundant for the target conversation flow can be removed, making the generated target conversation flow more concise and efficient, thereby improving the practicality of the conversation flow.

[0190] In some alternative embodiments, A4 may include the following A41 - A42.

[0191] A41. Based on the process conversation intent sequence of each piece of conversation data, for each first process intent in the process conversation intent sequence, group the first process intent and the second process intents in the process intent sequence into a process intent multi-tuple.

[0192] Among them, the second process intent is located after the first process intent in the process conversation intent sequence, and the number of second process intents is at least one.

[0193] In some alternative embodiments, the position of the second process intent is adjacent to the position of the first process intent.

[0194] In some alternative embodiments, when there are multiple second process intents, the positions of the second process intents are adjacent.

[0195] In this embodiment, the first process intent may be the first process intent in the process intent multi-tuple, and the second process intents are at least one process intent adjacent to the first process intent and located after the first process intent in the process intent multi-tuple.

[0196] In some alternative embodiments, starting from the first process intent in the process conversation intent sequence, each process intent can be sequentially and orderly determined as the first process intent in turn, at least one process intent after the first process intent can be determined as the second process intent, and the first process intent and the second process intents are grouped into a process intent multi-tuple until it is terminated when a process intent multi-tuple containing X consecutive process intents cannot be generated.

[0197] Among them, X can be the number of process intents in the tuple, where X is greater than or equal to 2.

[0198] In one embodiment, when X is 2, the process intent tuple can be a process intent Figure 2 tuple, where the process intent Figure 2 tuple includes a first process intent and a second process intent.

[0199] In another embodiment, when X is 3, the process intent tuple can be a process intent Figure 3 tuple, where the process intent Figure 3 tuple includes a first process intent and two second process intents.

[0200] In some alternative embodiments, the process intent tuple can be a process intent Figure 2 tuple, a process intent Figure 3 tuple, a process intent Figure 4 tuple, etc., which are not limited herein.

[0201] For example, taking the triple as an example for illustration, the process dialogue intent sequence: {The customer service greets the customer, the customer confirms it is himself / herself, the customer service explains his / her identity and purpose, the customer indicates willingness to operate, the customer service makes an activity invitation, the customer gives an affirmative response, the customer service bids farewell to the customer}.

[0202] The first first process intent can be: The customer service greets the customer, and the second process intent can be: The customer confirms it is himself / herself, and the customer service explains his / her identity and purpose.

[0203] The second first process intent can be: The customer confirms it is himself / herself, and the second process intent can be: The customer service explains his / her identity and purpose, and the customer indicates willingness to operate.

[0204] ……

[0205] The sixth first process intent can be: The customer service makes an activity invitation, and the second process intent can be: The customer gives an affirmative response, and the customer service bids farewell to the customer.

[0206] Since there are no two second process intents following the next first process intent where the customer gives an affirmative response, the generation of the process intent Figure 3 tuple stops.

[0207] Finally, the triples that can be formed by the above process dialogue intention sequences include: [The customer service greets the customer, the customer confirms it is themselves, the customer asks about the identity of the customer service], [The customer confirms it is themselves, the customer asks about the identity of the customer service, the customer service explains their identity and purpose], [The customer asks about the identity of the customer service, the customer service explains their identity and purpose, the customer indicates willingness to operate], [The customer service explains their identity and purpose, the customer indicates willingness to operate, the customer service makes an event invitation], [The customer indicates willingness to operate, the customer service makes an event invitation, the customer gives an affirmative response], [The customer service makes an event invitation, the customer gives an affirmative response, the customer service bids farewell to the customer].

[0208] In some alternative embodiments, the process intention multi-tuples can be process intention Figure 3 multi-tuples.

[0209] It can be understood that the dialogue data is usually a gradually advancing process, and each dialogue depends on the result of the previous dialogue. Through the process intention Figure 3 multi-tuples, the gradually progressive relationship between dialogues can be well captured. Each triple usually contains an initial intention, a refined intention, and a response or further refined intention, and can describe a complete dialogue segment. While using binary tuples can represent the relationship between two dialogues and generate a target conversation flow based on the relationship between the two dialogues, in many cases, this is not sufficient to capture all the necessary details in the dialogue. For example, in the dialogue between the customer service and the customer, using binary tuples such as "The customer service greets the customer" and "The customer confirms it is themselves" cannot fully describe the subsequent interactions. Although binary tuples can also generate a target conversation flow, it will make it difficult to understand and analyze the dialogue flow. While four-tuples or more than four-tuples can also generate a target conversation flow, they will introduce too many details and complexities, resulting in unnecessary redundant information and increasing the difficulty of processing and maintenance.

[0210] Therefore, using the process intention Figure 3 multi-tuples can well balance complexity and clarity, can clearly describe the progress of the dialogue, and will not introduce too much complexity, facilitating understanding and analysis. Those skilled in the art should understand that using the process intention Figure 2 multi-tuples and four-tuples can also generate a target conversation flow.

[0211] A42, generate a target conversation flow based on the process intention multi-tuples generated from each piece of dialogue data.

[0212] In this embodiment, for example, the various triples can be connected in sequence to generate a target conversation flow. Here, generating a target conversation flow through triples can ensure the logical coherence between each dialogue step.

[0213] Here, the target conversation flow can include multiple process intentions arranged according to a specific process.

[0214] The target conversation flow can be presented in the form of a target conversation flow tree to more intuitively display different branches and paths of the conversation, helping intelligent customer service or human customer service to flexibly handle various situations in actual operations.

[0215] It can be understood that the above example is just the process meaning Figure 3 tuple of a piece of conversation data. Here, the conversation dataset can include multiple pieces of conversation data. By merging the triples of multiple pieces of conversation data, a target conversation flow is generated, and the common conversation paths and variants that can achieve specific goals in the conversation data between the customer service and the customer can be found to construct a complete and flexible conversation flow to adapt to different user needs and scenario changes.

[0216] Reference Figure 4 , Figure 4 is a schematic diagram of a target conversation flow according to the present disclosure.

[0217] As Figure 4 shown, the target conversation flow can be generated based on multiple pieces of conversation data. First, the customer service can greet the customer. When the customer confirms that it is himself / herself, the customer service can introduce the activity. When the customer confirms that it is not himself / herself, the customer service ends the conversation and hangs up.

[0218] After the customer service introduces the activity, if the customer expresses affirmation / default, the customer service further invites the customer to participate in the activity. If the customer expresses negation, the customer service ends the conversation and hangs up.

[0219] After the customer service invites the customer to participate in the activity, if the customer expresses affirmation / default, the customer service successfully invites the customer and then ends the conversation and hangs up. If the customer expresses negation, the customer service ends the conversation and hangs up.

[0220] Here, after generating the target conversation flow based on the process meaning Figure 3 tuple of each piece of conversation data, the word frequency of each process intention in the target conversation flow among all the process intentions in the entire target conversation flow can also be calculated. Then, based on a preset word frequency threshold, the process intentions with a word frequency less than the word frequency threshold are deleted, and the process intentions greater than or equal to the word frequency threshold are retained. In this way, by deleting the low-frequency process intentions, unnecessary conversation steps can be reduced, making the conversation more concise and efficient, and the conversation efficiency can be improved.

[0221] In some alternative embodiments, step 205 above may include B1-B4.

[0222] B1. Generate a process conversation intention sequence for each piece of conversation data according to the front-back order of each process intention in the conversation intention sequence of each piece of conversation data.

[0223] In this embodiment, for the dialogue intention sequence of each piece of dialogue data, the process intentions in each piece of dialogue sub-data can be identified, and then for the process intentions, according to the order before and after of each process intention in the dialogue intention sequence of each piece of dialogue data, a process dialogue intention sequence of each piece of dialogue data can be generated.

[0224] Similarly, the process dialogue intention sequence can include the process intentions arranged in the dialogue order of the dialogue data and the dialogue sub-data corresponding to the process intentions, or can include the process intentions arranged in the dialogue order without including the corresponding dialogue sub-data.

[0225] For example, the dialogue intention sequence can be: {Customer service: Hello, is this Mr. Wang? ------- Customer service greets the customer; Customer: Yes, this is me. ------- Customer confirms it is himself; Who are you? ------- Customer asks about the identity of the customer service; Customer service: Hello, I am the business manager of xxx. We have prepared a personal credit digital asset application plan especially for you... May I ask if you have a need for digital asset application now? ------- Customer service explains his identity and purpose; Customer: How do I operate this? ------- Customer asks how to operate; Customer service: You log in to the xxx APP... Just follow the prompts to submit the corresponding bill details. ------- Customer service tells how to operate; Customer: Okay, I'll take a look. ------- Customer indicates willingness to operate; Customer service: The digital asset quota will not incur any fees when not in use. However, we recommend that you perform a quota withdrawal operation immediately after obtaining the quota. This can not only skip the manual review process and ensure that the digital asset is processed within the fastest two minutes. In addition, in this way, you can retain the digital asset quota for a long time and achieve cyclic use. If you apply for and withdraw the quota within one hour, you can enjoy the above convenient services. We will send an activity text message to your mobile phone later, detailing the relevant steps and precautions. ------- Customer service makes an activity invitation; Customer: Okay. ------- Customer indicates affirmation; Customer service: Thank you for your support. Bye. ------- Customer service bids farewell to the customer}.

[0226] The process intentions in the dialogue intention sequence can include: customer service greets the customer, customer confirms it is himself, customer asks about the identity of the customer service, customer service explains his identity and purpose, customer indicates willingness to operate, customer service makes an activity invitation, customer indicates affirmation, customer service bids farewell to the customer.

[0227] In some alternative embodiments, when the process dialogue intention sequence includes the process intentions arranged in the dialogue order of the dialogue data and the dialogue sub-data corresponding to the process intentions, the corresponding process intentions and the dialogue sub-data corresponding to the process intentions can be extracted from the dialogue intention sequence according to the order before and after of each process intention in the dialogue intention sequence of each piece of dialogue data to generate the process dialogue intention sequence.

[0228] The process dialogue intention sequence can be: {Customer service: Hello, is this Mr. Wang? ------- Customer service greets the customer; Customer: Yes, this is me. ------- The customer confirms being the person; Who are you? ------- The customer asks about the identity of the customer service; Customer service: Hello, I am the business manager of xxx. We have specially prepared a personal credit digital asset application plan for you... May I ask if you have a need to apply for digital assets now? ------- The customer service explains their identity and purpose; Customer: Okay, let me take a look. ------- The customer indicates willingness to operate; Customer service: The digital asset quota will not incur any fees when not in use. However, we recommend that you perform a quota withdrawal operation immediately after obtaining the quota. This can not only skip the manual review process and ensure that the digital assets are processed within the fastest two minutes. In addition, in this way, you can retain the digital asset quota for a long time and achieve cyclic use. If you apply for and withdraw the quota within one hour, you can enjoy the above convenient services. We will send an activity text message to your mobile phone later, detailing the relevant steps and precautions. ------- The customer service makes an activity invitation; Customer: Okay. ------- The customer indicates affirmation; Customer service: Thank you for your support. Bye. ------- The customer service bids farewell to the customer}.

[0229] When the process dialogue intention sequence only includes the process intentions arranged in the order of the dialogue and does not include the corresponding dialogue sub-data, the corresponding process intentions can be extracted from the dialogue intention sequence according to the front-back order of each process intention in the dialogue intention sequence of each piece of dialogue data to generate the process dialogue intention sequence.

[0230] The process dialogue intention sequence can be: {The customer service greets the customer, the customer confirms being the person, the customer service explains their identity and purpose, the customer indicates willingness to operate, the customer service makes an activity invitation, the customer indicates affirmation, the customer service bids farewell to the customer}.

[0231] In some optional implementation manners, the corresponding process intentions in each process intention sequence and the dialogue sub-data corresponding to the process intentions can also be stored in a preset process intention knowledge base. In this way, when the intelligent customer service or the manual customer service conducts a dialogue for the target conversation process, the corresponding dialogue content can be obtained from the process intention knowledge base and output to the customer.

[0232] B2. Generate a target conversation process according to the process dialogue intention sequence of each piece of dialogue data.

[0233] After obtaining the process dialogue intention sequence of each piece of dialogue data, a target conversation process can be generated according to the process dialogue intention sequence of each piece of dialogue data.

[0234] Similarly, for the sequence of process dialogue intents in each segment of dialogue data, starting from the first process intent, the current starting position and the next two process intents are sequentially associated as a process intent Figure 3 tuple, and the starting position is moved one position backward until it is impossible to generate a process intent Figure 3 tuple containing three consecutive process intents. Based on the process intent Figure 3 tuples generated from each segment of dialogue data, a target conversation flow is generated.

[0235] B3. Delete the process intents in the dialogue intent sequence of each segment of dialogue data and the corresponding dialogue sub-data of the process intents.

[0236] Here, after identifying the process intents in the dialogue intent sequence of the dialogue data and generating the target conversation flow, the process intents in the dialogue intent sequence of each segment of dialogue data and the corresponding dialogue sub-data of the process intents can be deleted.

[0237] B4. Based on the dialogue intent sequence of each segment of dialogue data after deleting the process intents and the corresponding dialogue sub-data of the process intents, generate a question-and-answer intent knowledge base.

[0238] It can be understood that the dialogue intents include question-and-answer intents and process intents. After deleting the process intents and the corresponding dialogue sub-data, the remaining dialogue intents and dialogue sub-data in the dialogue intent sequence are the question-and-answer intents and the corresponding dialogue sub-data of the question-and-answer intents. Then, based on the dialogue intent sequence of each segment of dialogue data after deleting the process intents and the corresponding dialogue sub-data of the process intents, a question-and-answer intent knowledge base can be generated.

[0239] Deleting the process intents in the dialogue intent sequence of each segment of dialogue data and the corresponding dialogue sub-data of the process intents can obtain the dialogue intent sequence of each segment of dialogue data after deleting the process intents and the corresponding dialogue sub-data of the process intents.

[0240] The dialogue intent sequence of each segment of dialogue data after deleting the process intents and the corresponding dialogue sub-data of the process intents can be: {Customer: How do I operate this? -------- The customer asks how to operate; Customer service: You log in to the xxx APP... and submit the corresponding bill details according to the prompts. -------- The customer service tells how to operate}.

[0241] Similarly, first, according to the dialogue attributes of each dialogue sub-data, the dialogue sub-data corresponding to the question-and-answer intents in each dialogue intent sequence can be divided into question dialogue sub-data and answer dialogue sub-data. Then, based on the question-and-answer relationship between the question dialogue sub-data and the answer dialogue sub-data, at least one question-and-answer pair can be generated. Finally, based on at least one question-and-answer pair, a question-and-answer intent knowledge base can be generated.

[0242] In some alternative embodiments, step 205 described above may include C1-C3.

[0243] C1. Generate a Q&A intention knowledge base according to the conversation intention sequence and Q&A intention of each segment of conversation data.

[0244] In this embodiment, for the conversation intention sequence of each segment of conversation data, the Q&A intention in each conversation sub-data can be identified, and then a Q&A knowledge base can be generated for the Q&A intention.

[0245] Similarly, first, according to the conversation attributes of each conversation sub-data, the conversation sub-data corresponding to the Q&A intention in each conversation intention sequence can be divided into question conversation sub-data and reply conversation sub-data. Then, at least one Q&A pair can be generated according to the Q&A relationship between the question conversation sub-data and the reply conversation sub-data. Finally, a Q&A intention knowledge base can be generated according to at least one Q&A pair.

[0246] C2. Generate a process conversation intention sequence for each segment of conversation data according to the sequence of the front and back of each process intention in the conversation intention sequence of each segment of conversation data.

[0247] In this embodiment, for the conversation intention sequence of each segment of conversation data, the process intention in each conversation sub-data can be identified, and then for the process intention, a process conversation intention sequence for each segment of conversation data can be generated according to the sequence of the front and back of each process intention in the conversation intention sequence of each segment of conversation data.

[0248] Similarly, the process conversation intention sequence may include the process intention arranged in the conversation order of the conversation data and the conversation sub-data corresponding to the process intention, or may include the process intention arranged in the conversation order without including the corresponding conversation sub-data.

[0249] In some alternative embodiments, the corresponding process intention in each process intention sequence and the conversation sub-data corresponding to the process intention may also be stored in a preset process intention knowledge base. In this way, when the intelligent customer service or the artificial customer service conducts a conversation for the target conversation flow, the corresponding conversation content can be obtained from the process intention knowledge base and output to the customer.

[0250] C3. Generate a target conversation flow according to the process conversation intention sequence of each segment of conversation data.

[0251] After obtaining the process conversation intention sequence of each segment of conversation data, a target conversation flow can be generated according to the process conversation intention sequence of each segment of conversation data.

[0252] Similarly, for the process conversation intention sequence of each segment of conversation data, starting from the first process intention, the current starting position and the two subsequent process intentions are associated as the process intention in turn.Figure 3 Tuple, and move the starting position backward by one bit until it is impossible to generate a process intention that contains three consecutive process intentions Figure 3 Tuple, and then terminate. Generate the target conversation flow based on the process intention Figure 3 Tuple generated from each piece of conversation data.

[0253] Here, C1 can be executed first and then C2 and C3, or C2 and C3 can be executed first and then C1, without specific restrictions.

[0254] For the specific implementation, reference can be made to A1 - A4 and B1 - B4, which will not be elaborated here.

[0255] The conversation flow generation method provided by the embodiments of the present disclosure, first, obtains a preset intention recognition model and a conversation data set, where the conversation data set includes multiple pieces of conversation data. According to the intention recognition model, determine the conversation intention of each conversation sub - data in each piece of conversation data. Then, according to the conversation intention of each conversation sub - data and the conversation order of each conversation sub - data in the corresponding conversation data, generate a conversation intention sequence for each piece of conversation data, where the conversation intention sequence includes the conversation sub - data arranged in the conversation order of the conversation data and the conversation intention of each conversation sub - data. Next, determine the conversation intention category of each conversation intention in each conversation intention sequence. Finally, generate the target conversation flow according to the conversation intention category of each conversation intention in each conversation intention sequence and the conversation intention sequence of each piece of conversation data. The present disclosure can accurately identify the intention in the conversation data based on the intention recognition model, thus significantly reducing manual intervention and improving the overall efficiency. Then, by processing the recognized intention into a conversation intention sequence, its logic is clearer, which is convenient for subsequent analysis and the generation of the target conversation flow. And generating the target conversation flow based on the conversation intention category can remove the useless or redundant intentions for the target conversation flow, making the generated target conversation flow more concise and efficient, thereby enhancing the practicality of the conversation flow.

[0256] For further reference Figure 5 As an implementation of the methods shown in the above figures, an embodiment of a conversation flow generation device is provided by the present disclosure. This device embodiment corresponds to Figure 2 the method embodiment shown, and this device can be specifically applied to various terminal devices.

[0257] Such as Figure 5As shown in the figure, the speech process generation device of this embodiment, device 500 includes: an acquisition unit 501, an intention determination unit 502, a sequence generation unit 503, an intention category determination unit 504, and a speech process generation unit 505. Among them, the acquisition unit 501 is used to acquire a preset intention recognition model and a dialogue data set, where the dialogue data set includes multiple segments of dialogue data; the intention determination unit 502 is used to determine the dialogue intention of each dialogue sub-data in each segment of dialogue data according to the intention recognition model; the sequence generation unit 503 is used to generate a dialogue intention sequence for each segment of dialogue data according to the dialogue intention of each dialogue sub-data and the dialogue order of each dialogue sub-data in the corresponding dialogue data, where the dialogue intention sequence includes the dialogue sub-data arranged in the dialogue order and the dialogue intention of each dialogue sub-data; the intention category determination unit 504 is used to determine the dialogue intention category of each dialogue intention in each dialogue intention sequence; the speech process generation unit 505 is used to generate a target speech process according to the dialogue intention category of each dialogue intention in each dialogue intention sequence and the dialogue intention sequence of each segment of dialogue data.

[0258] In this embodiment, the specific processing of the acquisition unit 501, the intention determination unit 502, the sequence generation unit 503, the intention category determination unit 504, and the speech process generation unit 505 and the technical effects brought by them can be respectively referred to Figure 2 the relevant descriptions of steps 201 to 205 in the corresponding embodiment, which will not be elaborated here.

[0259] In some alternative embodiments, the sequence generation unit 503 may be further configured to:

[0260] Cluster each dialogue intention according to the similarity between each dialogue intention to obtain at least one dialogue intention cluster;

[0261] For each dialogue intention cluster, establish a dialogue intention cluster label for each dialogue intention cluster;

[0262] Generate a dialogue intention sequence for each segment of dialogue data according to each dialogue intention cluster label and the dialogue order of each dialogue sub-data in each segment of dialogue data.

[0263] In some alternative embodiments, the speech process generation unit 505 may be further configured to:

[0264] Identify the process intention and / or question-and-answer intention in the dialogue intention sequence of each segment of dialogue data according to the dialogue intention category;

[0265] Generate a target speech process according to the dialogue intention sequence of each segment of dialogue data and the process intention and / or question-and-answer intention.

[0266] In some alternative embodiments, the conversation flow generation unit 505 may be further configured to:

[0267] Generate a question-and-answer intent knowledge base based on the conversation intent sequence and question-and-answer intent of each segment of conversation data;

[0268] Delete the question-and-answer intent and the corresponding conversation sub-data in the conversation intent sequence of each segment of conversation data;

[0269] Generate a flow conversation intent sequence for each segment of conversation data based on the conversation intent sequence of each segment of conversation data after deleting the question-and-answer intent and the corresponding conversation sub-data;

[0270] Generate a target conversation flow according to the flow conversation intent sequence of each segment of conversation data.

[0271] In some alternative embodiments, the conversation flow generation unit 505 may be further configured to:

[0272] Generate a flow conversation intent sequence for each segment of conversation data according to the front-to-back order of each flow intent in the conversation intent sequence of each segment of conversation data;

[0273] Generate a target conversation flow according to the flow conversation intent sequence of each segment of conversation data.

[0274] In some alternative embodiments, it further includes:

[0275] Delete the flow intent and the corresponding conversation sub-data in the conversation intent sequence of each segment of conversation data;

[0276] Generate a question-and-answer intent knowledge base based on the conversation intent sequence of each segment of conversation data after deleting the flow intent and the corresponding conversation sub-data.

[0277] In some alternative embodiments, the conversation flow generation unit 505 may be further configured to:

[0278] Generate a question-and-answer intent knowledge base based on the conversation intent sequence and question-and-answer intent of each segment of conversation data;

[0279] Generate a flow conversation intent sequence for each segment of conversation data according to the front-to-back order of each flow intent in the conversation intent sequence of each segment of conversation data;

[0280] Generate a target conversation flow according to the flow conversation intent sequence of each segment of conversation data.

[0281] In some alternative embodiments, the conversation flow generation unit 505 may be further configured to:

[0282] According to the dialogue attributes of each dialogue sub - data, divide the dialogue sub - data corresponding to the question - answering intention in each dialogue intention sequence into question dialogue sub - data and answer dialogue sub - data;

[0283] Generate at least one question - answer pair according to the question - answering relationship between the question dialogue sub - data and the answer dialogue sub - data;

[0284] Generate a question - answering intention knowledge base according to at least one question - answer pair.

[0285] In some alternative embodiments, the conversation flow generation unit 505 may be further configured to:

[0286] Based on the process dialogue intention sequence of each segment of dialogue data, for each first process intention in the process dialogue intention sequence, group the first process intention and the second process intentions in the process dialogue intention sequence into a process intention multi - tuple, where the second process intentions are located after the first process intention in the process dialogue intention sequence and the number of the second process intentions is at least one;

[0287] Generate a target conversation flow chart based on the process intention multi - tuples generated from each segment of dialogue data.

[0288] In some alternative embodiments, the position of the second process intention is adjacent to the position of the first process intention.

[0289] In some alternative embodiments, when there are multiple second process intentions, the positions of the second process intentions are adjacent.

[0290] It should be noted that the implementation details and technical effects of each unit in the conversation flow generation device provided by the embodiments of the present disclosure can be referred to the descriptions of other embodiments in the present disclosure, and will not be elaborated here.

[0291] Next, refer to Figure 6 , which shows a schematic structural diagram of a computer system 600 of a terminal device suitable for implementing the present disclosure. Figure 6 The shown computer system 600 is only an example, and should not bring any limitation to the functions and usage scopes of the embodiments of the present disclosure.

[0292] As Figure 6As shown, the computer system 600 may include a processing device (such as a central processing unit, a graphics processing unit, etc.) 601, which may perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the computer system 600 are also stored. The processing device 601, the ROM 602, and the RAM 603 are connected to each other through a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0293] Generally, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the computer system 600 to communicate with other devices wirelessly or wiredly to exchange data. Although Figure 6 the computer system 600 of an electronic device with various devices is shown, it should be understood that it is not required to implement or have all the shown devices. Instead, more or fewer devices may be implemented or had.

[0294] Specifically, according to an embodiment of the present disclosure, the process described above with reference to the flowchart may be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes program codes for performing the method shown in the flowchart. In such an embodiment, the computer program may be downloaded and installed from a network through the communication device 609, or installed from the storage device 608, or installed from the ROM 602. When the computer program is executed by the processing device 601, the above functions defined in the method of the embodiment of the present disclosure are executed.

[0295] It should be noted that the above-mentioned computer-readable medium in the present disclosure can be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of the computer-readable storage medium can include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, 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 disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, the computer-readable storage medium can be any tangible medium that contains or stores a program, and this program can be used by or in combination with an instruction execution system, apparatus, or device. In the present disclosure, the computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer-readable signal medium can also be any computer-readable medium other than the computer-readable storage medium, and this computer-readable signal medium can send, propagate, or transmit a program for use by or in combination with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted by any appropriate medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0296] The above-mentioned computer-readable medium can be included in the above-mentioned electronic device; or it can exist separately without being assembled into the electronic device.

[0297] The above-mentioned computer-readable medium carries one or more programs, and when the above-mentioned one or more programs are executed by the electronic device, the electronic device is enabled to implement Figures 2 - 4 the speech flow generation method shown in any of the illustrated embodiments and their optional embodiments.

[0298] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages or combinations thereof. The programming languages include object-oriented programming languages such as Java, Smalltalk, C++, Python, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, executed as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or it may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0299] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code that contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than marked in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, may be implemented by a dedicated hardware-based system for performing the specified functions or operations, or may be implemented by a combination of dedicated hardware and computer instructions.

[0300] The units involved in the embodiments described in the present disclosure may be implemented in software or in hardware. Among them, the name of the unit does not constitute a limitation to the unit itself in some cases. For example, the acquisition unit may also be described as "the unit for acquiring the intention recognition model and the dialogue data set".

[0301] The above description is only the preferred embodiments of the present disclosure and an explanation of the applied technical principles. Those skilled in the art should understand that the scope of disclosure involved in the present disclosure is not limited to the technical solutions formed by the specific combination of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosure concept. For example, the technical solutions formed by mutually replacing the above features with the (but not limited to) technical features having similar functions disclosed in the present disclosure.

Claims

1. A method for generating a conversation flow, characterized in that, The method includes: Obtaining a preset intention recognition model and a dialogue data set, where the dialogue data set includes multiple segments of dialogue data; Determining the dialogue intention of each piece of dialogue sub-data in each segment of the dialogue data according to the intention recognition model; Generating a dialogue intention sequence for each segment of the dialogue data according to the dialogue intention of each piece of dialogue sub-data and the dialogue order of each piece of dialogue sub-data in the corresponding dialogue data, where the dialogue intention sequence includes the dialogue sub-data arranged in the dialogue order of the dialogue data and the dialogue intentions of each piece of dialogue sub-data; Determining the dialogue intention category of each dialogue intention in each of the dialogue intention sequences; Generating a target conversation flow according to the dialogue intention category of each dialogue intention in each of the dialogue intention sequences and the dialogue intention sequence of each segment of the dialogue data.

2. The method according to claim 1, wherein The step of generating a dialogue intention sequence for each segment of the dialogue data according to the dialogue intention of each piece of dialogue sub-data and the dialogue order of each piece of dialogue sub-data in the corresponding dialogue data includes: Clustering each of the dialogue intentions according to the similarity between the dialogue intentions to obtain at least one dialogue intention cluster; Establishing a dialogue intention cluster label for each of the dialogue intention clusters; Generating the dialogue intention sequence of each segment of the dialogue data according to the dialogue intention cluster labels and the dialogue order of each piece of dialogue sub-data in each segment of the dialogue data.

3. The method according to claim 1, characterized in that, The step of generating a target conversation flow according to the dialogue intention category of each dialogue intention in each of the dialogue intention sequences and the dialogue intention sequence of each segment of the dialogue data includes: Identifying the process intention and / or the question-and-answer intention in the dialogue intention sequence of each segment of the dialogue data according to the dialogue intention category; Generating a target conversation flow according to the dialogue intention sequence of each segment of the dialogue data and the process intention and / or the question-and-answer intention.

4. The method according to claim 3, wherein The step of generating a target conversation flow according to the dialogue intention sequence of each segment of the dialogue data and the question-and-answer intention includes: Generating a question-and-answer intention knowledge base according to the dialogue intention sequence of each segment of the dialogue data and the question-and-answer intention; Deleting the question-and-answer intention and the corresponding dialogue sub-data in the dialogue intention sequence of each segment of the dialogue data; Generating a process dialogue intention sequence for each segment of the dialogue data based on the dialogue intention sequence of each segment of the dialogue data after deleting the question-and-answer intention and the corresponding dialogue sub-data; Generating the target conversation flow according to the process dialogue intention sequence of each segment of the dialogue data.

5. The method according to claim 3, wherein The step of generating a target conversation flow according to the dialogue intention sequence of each segment of the dialogue data and the process intention includes: Generating a process dialogue intention sequence for each segment of the dialogue data according to the front-back order of each of the process intentions in the dialogue intention sequence of each segment of the dialogue data; Generating the target conversation flow according to the process dialogue intention sequence of each segment of the dialogue data.

6. The method according to claim 5, wherein The method further includes: Delete the process intent in the dialogue intent sequence of each piece of the dialogue data and the corresponding dialogue sub-data of the process intent; Generate the Q&A intent knowledge base based on the dialogue intent sequence of each piece of the dialogue data after deleting the process intent and the corresponding dialogue sub-data of the process intent.

7. The method according to claim 3, wherein The generation of the target dialogue flow according to the dialogue intent sequence, the process intent, and the Q&A intent of each piece of the dialogue data includes: Generate a Q&A intent knowledge base according to the dialogue intent sequence and the Q&A intent of each piece of the dialogue data; Generate a process dialogue intent sequence for each piece of the dialogue data according to the front-back order of each process intent in the dialogue intent sequence of each piece of the dialogue data; Generate the target dialogue flow according to the process dialogue intent sequence of each piece of the dialogue data.

8. The method according to any one of claims 4 or 7, characterized in that, The generation of the Q&A intent knowledge base according to the dialogue intent sequence and the Q&A intent of each piece of the dialogue data includes: According to the dialogue attributes of each dialogue sub-data, divide the dialogue sub-data corresponding to the Q&A intent in each dialogue intent sequence into question dialogue sub-data and answer dialogue sub-data; Generate at least one Q&A pair according to the Q&A relationship between the question dialogue sub-data and the answer dialogue sub-data; Generate a Q&A intent knowledge base according to at least one Q&A pair.

9. The method according to any one of claims 4 or 5, characterized in that The generation of the target dialogue flow according to the process dialogue intent sequence of each piece of the dialogue data includes: Based on the process dialogue intent sequence of each piece of the dialogue data, for each first process intent in the process dialogue intent sequence, group the first process intent and the second process intents in the process dialogue intent sequence, where the second process intents are located after the first process intent in the process dialogue intent sequence, and the number of the second process intents is at least one; Generate the target dialogue flow chart based on the process intent multi-tuples generated from each piece of the dialogue data.

10. The method according to claim 9, wherein The position of the second process intent is adjacent to the position of the first process intent.

11. The method according to any one of claims 9 or 10, characterized in that When there are multiple second process intents, the positions of the second process intents are adjacent to each other.

12. A dialogue flow generation device, comprising: An acquisition unit, configured to acquire a preset intent recognition model and a dialogue data set, where the dialogue data set includes multiple pieces of dialogue data; An intent determination unit, configured to determine the dialogue intent of each dialogue sub-data in each piece of the dialogue data according to the intent recognition model; A sequence generation unit, configured to generate a dialogue intent sequence for each piece of the dialogue data according to the dialogue intent of each dialogue sub-data and the dialogue order of each dialogue sub-data in the corresponding dialogue data, where the dialogue intent sequence includes the dialogue sub-data arranged in dialogue order of the dialogue data and the dialogue intents of each dialogue sub-data; An intent category determination unit, configured to determine the dialogue intent category of each dialogue intent in each dialogue intent sequence; A conversation script generation unit, configured to generate a target conversation script according to the conversation intention categories of the respective conversation intentions in each of the conversation intention sequences and the conversation intention sequences of each segment of the conversation data.

13. An electronic device, comprising: One or more processors; A storage device having stored thereon one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the method according to any one of claims 1-11.

14. A computer-readable storage medium having a computer program stored thereon, wherein, When a computer program is executed by one or more processors, the method according to any one of claims 1-11 is implemented.

15. A computer program product, comprising a computer program / instructions, where the computer program / instructions, when executed by a processor, implement the method according to any one of claims 1-11.

Images