Script recommendation method and device, electronic equipment and storage medium

Abstract

The present application relates to artificial intelligence technology, disclose a kind of dialogue recommendation method, comprising: obtaining historical dialogue corpus set, according to the historical dialogue corpus set constructs intent flow;Obtain the first dialogue text corresponding to the previous dialogue node and the second dialogue text corresponding to the current dialogue node, the first dialogue text is carried out intent recognition, obtain the first intent, the second dialogue text is carried out intent recognition, obtain the second intent;In the intent flow, query the intent node matched with the first intent, the matched starting point is obtained by the intent node, according to the matching starting point and the second intent, in the intent flow, obtain the target intent node matched with the second intent;Obtain the recommended dialogue corresponding to the target intent node, and recommend the recommended dialogue to the predetermined business personnel.The present application also proposes a kind of dialogue recommendation device, electronic equipment and storage medium.The present application can improve the accuracy of dialogue recommendation.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method, apparatus, electronic device, and computer-readable storage medium for recommending dialogue. Background Technology

[0002] Script recommendations are commonly used in professional fields such as consultation, Q&A, or sales services, including insurance consultation, transportation consultation, and product promotion. When customers communicate with sales personnel online, the system analyzes the text of customer feedback in real time based on script recommendation technology and recommends relevant scripts to the sales personnel communicating with the customer. This helps the sales personnel understand the customer's intentions and provide more professional services to the customer.

[0003] In the application of script recommendation, the most common approach at present is to use script recommendation based on retrieval technology. This means that based on a pre-set historical dialogue corpus, the text currently fed back by the customer is retrieved, and the scripts or business processes matched by the retrieval are recommended to the sales staff, guiding them to respond according to the recommended scripts or to guide their communication with the customer according to the recommended business processes.

[0004] The aforementioned script recommendations based on retrieval technology are relatively independent for each text retrieval operation, lacking an understanding of the context in which the retrieval object is located. As a result, it is impossible to make an overall judgment on the intent and provide coherent guidance during the online communication process between customers and business personnel, leading to low accuracy of the related recommended scripts. Summary of the Invention

[0005] This invention provides a method, apparatus, electronic device, and computer-readable storage medium for recommending sales scripts, the main purpose of which is to improve the accuracy of sales script recommendations.

[0006] To achieve the above objectives, the present invention provides a script recommendation method, comprising:

[0007] Obtain a set of historical dialogue data, and construct an intent flow based on the set of historical dialogue data;

[0008] Obtain the first dialogue text corresponding to the previous dialogue node and the second dialogue text corresponding to the current dialogue node. Perform intent recognition on the first dialogue text to obtain the first intent, and perform intent recognition on the second dialogue text to obtain the second intent.

[0009] In the intent flow, query the intent node that matches the first intent, take the queried intent node as the matching starting point, and obtain the target intent node that matches the second intent in the intent flow based on the matching starting point and the second intent.

[0010] Obtain the recommended script corresponding to the target intent node, and recommend the recommended script to the preset business personnel.

[0011] Optionally, the process of constructing the intent based on the historical dialogue corpus includes:

[0012] The historical dialogue corpus is split into a normal business scenario corpus and an abnormal business scenario corpus.

[0013] By identifying the customer role and customer intent of each corpus in the normal business scenario corpus, a normal customer intent node is obtained; by identifying the business role and business intent of each corpus in the normal business scenario corpus, a business intent node is obtained.

[0014] Obtain the recommended dialogue corresponding to each business intent node, and use the obtained recommended dialogue as the processing branch of the corresponding business intent node;

[0015] According to the preset business processing rules, the normal customer intent node and the business intent node are associated to obtain the main process;

[0016] Identify the customer roles and customer intentions of each corpus in the abnormal business scenario corpus to obtain abnormal customer intention nodes;

[0017] Obtain the processing script corresponding to each abnormal customer intent node, and use it as the processing branch of the corresponding abnormal customer intent node to obtain a sub-process.

[0018] The main process and the sub-processes are combined to obtain the intended process.

[0019] Optionally, splitting the historical dialogue corpus into a normal business scenario corpus and an abnormal business scenario corpus includes:

[0020] The historical dialogue corpus is labeled according to preset business scenario tags;

[0021] Clustering is performed on the annotated historical dialogue corpus to obtain a preset number of cluster centers;

[0022] The historical dialogue data belonging to any of the cluster centers mentioned above will be used as the normal business scenario corpus.

[0023] Historical dialogue data that does not belong to any of the cluster centers mentioned above will be used as a corpus set for abnormal business scenarios.

[0024] Optionally, the step of associating the normal customer intent node and the business intent node according to preset business processing rules to obtain the main process includes:

[0025] In the preset intent library, query the intent ID corresponding to each normal customer intent node and each business intent node;

[0026] Based on the preset business processing rules, identify the sequential relationship of each intent ID;

[0027] The main process is obtained by connecting the corresponding normal customer intent nodes and each of the business intent nodes according to the order relationship.

[0028] Optionally, before querying the intent node matching the first intent in the intent flow, the method further includes:

[0029] Identify whether the second dialogue text is the dialogue text acquired for the first time;

[0030] When the second dialogue text is the dialogue text acquired for the first time, the start node in the intent flow is identified and the start node is used as the matching starting point.

[0031] Optionally, obtaining the target intent node matching the second intent in the intent flow based on the matching starting point and the second intent includes:

[0032] In the intent flow, sub-intent nodes connected to the matching starting point are obtained, and the matching degree between the second intent and each sub-intent node is calculated;

[0033] Select the sub-intent node that meets the preset matching degree condition as the target intent node;

[0034] When there is no sub-intent node that meets the preset matching degree condition, calculate the distance between each intent node in the main process where the matching starting point is located and the second intent.

[0035] Select the intent node that meets the preset distance condition as the target intent node;

[0036] When no intent node satisfies the preset distance condition, in the sub-process of the intent flow, an intent node that satisfies the preset exception handling condition is queried and used as the target intent node.

[0037] Optionally, the step of performing intent recognition on the first dialogue text to obtain a first intent includes:

[0038] The first dialogue text is transformed into a word vector to obtain a text vector.

[0039] Text features of the text vector are extracted using the convolutional layers of a pre-trained intent recognition model;

[0040] The text features are converted into a feature matrix using the fully connected layer of the pre-trained intent recognition model;

[0041] The probability value between the feature matrix and the preset intent label is calculated using a preset activation function, and the intent label corresponding to the probability value greater than the preset probability value threshold is selected as the first intent.

[0042] To address the above problems, the present invention also provides a script recommendation device, the device comprising:

[0043] Intent flow construction module: used to obtain a set of historical dialogue data and construct an intent flow based on the set of historical dialogue data;

[0044] The text intent recognition module is used to obtain the first dialogue text corresponding to the previous dialogue node and the second dialogue text corresponding to the current dialogue node, perform intent recognition on the first dialogue text to obtain the first intent, and perform intent recognition on the second dialogue text to obtain the second intent.

[0045] The text and flow matching module is used to query intent nodes that match the first intent in the intent flow, take the queried intent nodes as the matching starting point, and obtain the target intent node that matches the second intent in the intent flow based on the matching starting point and the second intent.

[0046] The recommended script acquisition module is used to acquire the recommended script corresponding to the target intent node and recommend the recommended script to preset business personnel.

[0047] To address the above problems, the present invention also provides an electronic device, the electronic device comprising:

[0048] Memory, storing at least one computer program; and

[0049] The processor executes the program stored in the memory to implement the aforementioned speech recommendation method.

[0050] To address the aforementioned problems, the present invention also provides a computer-readable storage medium storing at least one computer program, which is executed by a processor in an electronic device to implement the aforementioned speech recommendation method.

[0051] This invention constructs an intent flow based on a historical dialogue corpus, providing a process guarantee for script recommendation processing. By querying the target intent node that matches the second intent in the intent flow, the recommended script corresponding to the target intent node is obtained. This combines the textual intent of the dialogue text with the intent nodes in the intent flow, ensuring that the recommended script is within the framework of the intent flow. On the other hand, each time a target intent node is queried, the intent node corresponding to the first intent is used as the matching starting point, which can realize the association of the textual intent of the context of the dialogue text, ensuring the continuity of intent judgment, thereby improving the accuracy of recommended scripts. Attached Figure Description

[0052] Figure 1 This is a flowchart illustrating a speech recommendation method provided in an embodiment of the present invention;

[0053] Figure 2 A detailed implementation flowchart of one step in the script recommendation method provided in an embodiment of the present invention is shown below;

[0054] Figure 3 A detailed implementation flowchart of another step in the speech recommendation method provided in an embodiment of the present invention is shown below;

[0055] Figure 4 A functional block diagram of a speech recommendation device provided in an embodiment of the present invention;

[0056] Figure 5 This is a schematic diagram of the structure of an electronic device for implementing the speech recommendation method according to an embodiment of the present invention.

[0057] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0058] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0059] This application provides a method for recommending dialogue scripts. The execution entity of the dialogue script recommendation method includes, but is not limited to, at least one of the following: a server, a terminal, or other electronic devices that can be configured to execute the method provided in this application. In other words, the dialogue script recommendation method can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0060] Reference Figure 1 The diagram shown is a flowchart illustrating a speech recommendation method provided in an embodiment of the present invention.

[0061] In this embodiment, the script recommendation method includes:

[0062] S1. Obtain a set of historical dialogue data and construct an intent flow based on the set of historical dialogue data;

[0063] In this embodiment of the invention, the historical dialogue corpus refers to a collection of existing communication records generated by online interactions between customers and business personnel in a specific business area, such as the question-and-answer records between business personnel and customers in an insurance consultation scenario.

[0064] In this embodiment of the invention, user-authorized, retrievalable communication records can be obtained from a pre-built storage area using computer statements (Java statements, Python statements, etc.) with data scraping capabilities. The storage area includes, but is not limited to, databases, blockchains, and network caches.

[0065] Understandably, the historical dialogue dataset typically includes two roles: the consultant or the person seeking help, i.e., the customer, and the person providing consulting services or guidance, i.e., the salesperson. Salespeople need to handle different business situations based on the customer's different intentions, and at the same time, they will guide customer personnel based on different business intentions.

[0066] In this embodiment of the invention, the intent flow refers to the online communication and processing flow between customers and business personnel formed by connecting the intents of the customers or business personnel according to preset business logic.

[0067] For details, please refer to Figure 2 As shown, S1 includes:

[0068] S11. The historical dialogue corpus is split into a normal business scenario corpus and an abnormal business scenario corpus.

[0069] S12. Identify the customer role and customer intent of each corpus in the normal business scenario corpus to obtain normal customer intent nodes, and identify the business role and business intent of each corpus in the normal business scenario corpus to obtain business intent nodes.

[0070] S13. Obtain the recommended dialogue corresponding to each business intent node, and use the obtained recommended dialogue as the processing branch of the corresponding business intent node;

[0071] S14. According to the preset business processing rules, associate the normal customer intent node and the business intent node to obtain the main process;

[0072] S15. Identify the customer role and customer intent of each corpus in the abnormal business scenario corpus to obtain the abnormal customer intent node;

[0073] S16. Obtain the processing script corresponding to each abnormal customer intent node, and use it as the processing branch of the corresponding abnormal customer intent node to obtain a sub-process.

[0074] S17. The main process and the sub-processes are combined to obtain the intention process.

[0075] It is understandable that online consultations, services, and product promotions may encounter requests for help or inquiries unrelated to the pre-set business areas. For example, in an online insurance consultation scenario, a customer might ask how to apply for social security. These scenarios unrelated to the pre-set business areas are considered abnormal business scenarios, and pre-defined handling mechanisms can be used to address them.

[0076] In detail, the step of splitting the historical dialogue corpus into a normal business scenario corpus and an abnormal business scenario corpus includes: labeling the historical dialogue corpus according to preset business scenario tags; performing clustering operations on the labeled historical dialogue corpus to obtain a preset number of cluster centers; using historical dialogue corpus belonging to any of the cluster centers as the normal business scenario corpus; and using historical dialogue corpus not belonging to any of the cluster centers as the abnormal business scenario corpus.

[0077] In this embodiment of the invention, the preset business scenario label refers to the scenario label obtained by classifying the types of business provided by the actual online service. For example, in insurance consultation, business scenario labels such as critical illness insurance consultation, education insurance consultation, and financial insurance consultation can be set according to the classification rules of insurance types.

[0078] In this embodiment of the invention, the historical dialogue corpus can be annotated using a text annotation tool, such as Docano.

[0079] Preferably, the preset number is the same as the preset number of business scenario tags, and can be set according to actual business needs.

[0080] In this embodiment of the invention, a pre-trained intent recognition model can be used to identify the role information and intent information of each corpus. The pre-trained intent recognition model can be a deep learning-based convolutional neural network model, which is a model with text recognition capabilities trained using a large amount of communication texts from customers and business personnel.

[0081] In this embodiment of the invention, a preset mapping relationship between intents and recommended phrases can be used to query recommended phrases corresponding to the business intent node from a preset recommended phrase library. The preset recommended phrase library is indexed by intent, with each intent corresponding to multiple recommended phrases, and each recommended phrase having a different recommendation priority.

[0082] In detail, the step of associating the normal customer intent nodes and the business intent nodes according to preset business processing rules to obtain the main process includes: querying the intent ID corresponding to each normal customer intent node and each business intent node in a preset intent library; identifying the sequential relationship of each intent ID according to the preset business processing rules; and connecting the corresponding normal customer intent nodes and each business intent node in series according to the sequential relationship to obtain the main process.

[0083] In this embodiment of the invention, the preset intent library includes two types: customer intent and business intent. Each type of customer intent and each type of business intent corresponds to a unique intent ID.

[0084] In this embodiment of the invention, the preset business processing rule refers to dividing the business processing logic into processing nodes with different execution relationships according to actual business needs, and each processing node has a different node number.

[0085] In this embodiment of the invention, the main flow in the intent flow is mainly used to solve customer inquiries and requests for help in conventional business scenarios, while the sub-flows are mainly for customer inquiries or requests for help in non-conventional business scenarios.

[0086] S2. Obtain the first dialogue text corresponding to the previous dialogue node and the second dialogue text corresponding to the current dialogue node. Perform intent recognition on the first dialogue text to obtain the first intent. Perform intent recognition on the second dialogue text to obtain the second intent.

[0087] In this embodiment of the invention, taking online insurance consultation as an example, when a customer initiates an online consultation, the dialogue text between the customer and the online business personnel is obtained in real time, and the intent of the dialogue text is identified to accurately grasp the customer's intent or the business intent of the business personnel, thereby promoting communication between the two parties to achieve the expected purpose.

[0088] Understandably, during a conversation between a customer and a salesperson, the conversation can be divided into different dialogue nodes based on the time it takes. For example, a dialogue can be divided into a node every 5 minutes. If the conversation lasts for 15 minutes, it can be divided into 3 nodes. The previous dialogue node is relative to the current dialogue node, that is, the dialogue node that occurred before the current dialogue node and is immediately adjacent to the current dialogue node. The text content and text length of each dialogue node can be different.

[0089] In this embodiment of the invention, a pre-trained intent recognition model can be used to identify the textual intent of the dialogue text.

[0090] Taking the first dialogue text as an example, the detailed steps of performing intent recognition on the first dialogue text to obtain the first intent include: performing word vector conversion on the first dialogue text to obtain text vectors; extracting text features of the text vectors using the convolutional layer of a pre-trained intent recognition model; converting the text features into a feature matrix using the fully connected layer of the pre-trained intent recognition model; calculating the probability value between the feature matrix and a preset intent label using a preset activation function; and selecting the intent label corresponding to the probability value greater than a preset probability value threshold as the first intent.

[0091] It should be noted that the method for performing intent recognition on the second dialogue text is the same as the method for performing intent recognition on the first dialogue text.

[0092] In this embodiment of the invention, the number and size of the convolutional kernels, the number of channels, and the stride of the convolutional and fully connected layers of the pre-trained intent recognition model can be set according to actual business needs.

[0093] In this embodiment of the invention, the preset activation function includes, but is not limited to, the softmax activation function, the sigmoid activation function, and the ReLU activation function. Taking online insurance consultation as an example, the preset intent tags include, but are not limited to, insurance claims consultation, insurance benefit consultation, and insurance renewal consultation.

[0094] In one embodiment of the present invention, the probability value can be calculated using the following preset activation function:

[0095]

[0096] Where p(a|x) is the probability value between the feature matrix x and the intent label a, w a Let T be the weight vector of intent label a, T be the transpose operator, exp be the expected operator, and A be the preset number of intent labels.

[0097] In this embodiment of the invention, the preset probability value threshold can be set according to the actual situation.

[0098] S3. In the intent flow, query the intent node that matches the first intent, take the queried intent node as the matching starting point, and obtain the target intent node that matches the second intent in the intent flow according to the matching starting point and the second intent.

[0099] In this embodiment of the invention, the matching starting point refers to the intent node in the intent flow that is closely related to the second intent.

[0100] Specifically, before querying the intent node that matches the first intent in the intent flow, the method further includes: identifying whether the second dialogue text is a dialogue text acquired for the first time; when the second dialogue text is a dialogue text acquired for the first time, identifying the start node in the intent flow and using the start node as the matching starting point.

[0101] It is understandable that there is usually a certain logical connection between online conversations between customers and business personnel. During the online conversations between customers and business personnel, the conversation text is continuously obtained. Each time the conversation text is obtained, it is necessary to query and match the target intent node in the intent flow. There is a close logical relationship between the target intent nodes corresponding to adjacent query operations.

[0102] For details, please refer to Figure 3 As shown, the step of obtaining the target intent node that matches the second intent in the intent flow based on the matching starting point and the second intent includes:

[0103] S31. In the intent flow, obtain the sub-intent nodes connected to the matching starting point, and calculate the matching degree between the second intent and each of the sub-intent nodes;

[0104] S32. Select a sub-intent node that meets the preset matching degree condition as the target intent node;

[0105] S33. When there is no sub-intent node that satisfies the preset matching degree condition, calculate the distance between each intent node in the main process where the matching starting point is located and the second intent.

[0106] S34. Select the intent node that meets the preset distance condition as the target intent node;

[0107] S35. When there is no intent node that meets the preset distance condition, in the sub-process of the intent flow, query the intent node that meets the preset exception handling condition and use it as the target intent node.

[0108] In this embodiment of the invention, the matching degree between the second intent and the sub-intent nodes connected to the matching starting point can be calculated using Euclidean distance formula, Mahalanobis distance formula, etc., or the distance between each intent node in the main process where the matching starting point is located and the second intent can be calculated.

[0109] In this embodiment of the invention, the preset matching degree condition may be that the matching degree between the sub-intent node and the second intent is equal to or greater than a preset matching threshold. The preset distance condition may specify a distance threshold, and when the distance between the intent node in the main flow where the matching starting point is located and the second intent is less than or equal to the distance threshold, the corresponding intent node is selected as the target intent node.

[0110] In this embodiment of the invention, if the target intent node is not found in the main process, it indicates that the second intent does not belong to the normal business processing scope, and can then be processed using the processing method provided in the sub-process.

[0111] In this embodiment of the invention, the preset anomaly handling condition may refer to selecting the corresponding intent node as the target intent node when the distance between the second intent and each intent node in the sub-process is less than or equal to a preset anomaly distance threshold.

[0112] S4. Obtain the recommended script corresponding to the target intent node, and recommend the recommended script to the preset business personnel.

[0113] In this embodiment of the invention, the preset business personnel refers to business personnel who communicate with customers via online text or voice.

[0114] In this embodiment of the invention, the target intent node may correspond to multiple recommended scripts. Therefore, it is necessary to perform an optimization operation on the multiple recommended scripts, or to preferably recommend the script as the final recommended script and push the final recommended script to the preset business personnel.

[0115] Preferably, in the intent flow, the recommended phrases under each intent node can be scored or a recommendation priority can be set, and the recommended phrase with the highest score or the highest recommendation priority can be selected as the final recommended phrase. For example, the priority of the recommended phrases can be set according to the frequency of their recommendation, or the distance between each recommended phrase and its corresponding intent node can be calculated, and a score can be set for the corresponding recommended phrase based on the magnitude of the distance.

[0116] This invention constructs an intent flow based on a historical dialogue corpus, providing process assurance for script recommendation processing. By querying the target intent node matching the second intent in the intent flow in real time, the recommended script corresponding to the target intent node is obtained. This achieves the combination of the textual intent of the dialogue text with the intent nodes in the intent flow, ensuring that the recommended script is within the framework of the intent flow. On the other hand, each time a target intent node is queried, the intent node corresponding to the first intent is used as the matching starting point, which can realize the association of the textual intent of the context of the dialogue text, ensuring the continuity of intent judgment, thereby improving the accuracy of recommended scripts.

[0117] like Figure 4 The diagram shown is a functional block diagram of a speech recommendation device provided in an embodiment of the present invention.

[0118] The script recommendation device 100 of this invention can be installed in an electronic device. Depending on the functions implemented, the script recommendation device 100 may include an intent flow construction module 101, a text intent recognition module 102, a text and flow matching module 103, and a recommended script acquisition module 104. The module described in this invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and which are stored in the memory of the electronic device.

[0119] In this embodiment, the functions of each module / unit are as follows:

[0120] The intent flow construction module 101 is used to obtain a historical dialogue corpus set and construct an intent flow based on the historical dialogue corpus set.

[0121] The text intent recognition module 102 is used to obtain the first dialogue text corresponding to the previous dialogue node and the second dialogue text corresponding to the current dialogue node, perform intent recognition on the first dialogue text to obtain the first intent, and perform intent recognition on the second dialogue text to obtain the second intent.

[0122] The text and flow matching module 103 is used to query intent nodes that match the first intent in the intent flow, take the queried intent nodes as the matching starting point, and obtain the target intent node that matches the second intent in the intent flow according to the matching starting point and the second intent.

[0123] The recommended script acquisition module 104 is used to acquire the recommended script corresponding to the target intent node and recommend the recommended script to preset business personnel.

[0124] In detail, each module in the script recommendation device 100 described in this embodiment of the invention employs the same methods as described above during use. Figures 1 to 3The same technical means are used to recommend the same sales pitches as described above, and can produce the same technical effects, so they will not be repeated here.

[0125] like Figure 5 The diagram shown is a structural schematic of an electronic device for implementing a speech recommendation method according to an embodiment of the present invention.

[0126] The electronic device 1 may include a processor 10, a memory 11 and a bus, and may also include a computer program stored in the memory 11 and capable of running on the processor 10, such as a script recommendation program.

[0127] The memory 11 includes at least one type of readable storage medium, such as flash memory, portable hard drive, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 11 can be an internal storage unit of the electronic device 1, such as a portable hard drive. In other embodiments, the memory 11 can be an external storage device of the electronic device 1, such as a plug-in portable hard drive, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the electronic device 1. Furthermore, the memory 11 can include both internal and external storage units of the electronic device 1. The memory 11 can be used not only to store application software and various types of data installed on the electronic device 1, such as the code of a script recommendation program, but also to temporarily store data that has been output or will be output.

[0128] In some embodiments, the processor 10 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 10 is the control unit of the electronic device, connecting various components of the entire electronic device through various interfaces and lines. It executes programs or modules (such as script recommendation programs) stored in the memory 11, and calls data stored in the memory 11 to perform various functions of the electronic device 1 and process data.

[0129] The bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into an address bus, a data bus, a control bus, etc. The bus is configured to enable communication between the memory 11 and at least one processor 10, etc.

[0130] Figure 5 Only electronic devices with components are shown; it will be understood by those skilled in the art that... Figure 5 The structure shown does not constitute a limitation on the electronic device 1, and may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0131] For example, although not shown, the electronic device 1 may also include a power supply (such as a battery) to power various components. Preferably, the power supply can be logically connected to the at least one processor 10 through a power management device, thereby enabling functions such as charging management, discharging management, and power consumption management. The power supply may also include one or more DC or AC power supplies, recharging devices, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components. The electronic device 1 may also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be described in detail here.

[0132] Furthermore, the electronic device 1 may also include a network interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a Wi-Fi interface, a Bluetooth interface, etc.), which is typically used to establish communication connections between the electronic device 1 and other electronic devices.

[0133] Optionally, the electronic device 1 may further include a user interface, which may be a display, an input unit (such as a keyboard), or a standard wired or wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display may also be appropriately referred to as a screen or display unit, used to display information processed in the electronic device 1 and to display a visual user interface.

[0134] It should be understood that the embodiments described are for illustrative purposes only and are not limited to this structure in the scope of the patent application.

[0135] The speech recommendation program stored in the memory 11 of the electronic device 1 is a combination of multiple instructions, which, when run in the processor 10, can achieve the following:

[0136] Obtain a set of historical dialogue data, and construct an intent flow based on the set of historical dialogue data;

[0137] Obtain the first dialogue text corresponding to the previous dialogue node and the second dialogue text corresponding to the current dialogue node. Perform intent recognition on the first dialogue text to obtain the first intent, and perform intent recognition on the second dialogue text to obtain the second intent.

[0138] In the intent flow, query the intent node that matches the first intent, take the queried intent node as the matching starting point, and obtain the target intent node that matches the second intent in the intent flow based on the matching starting point and the second intent.

[0139] Obtain the recommended script corresponding to the target intent node, and recommend the recommended script to the preset business personnel.

[0140] Furthermore, if the modules / units integrated in the electronic device 1 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. The computer-readable storage medium can be volatile or non-volatile. For example, the computer-readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, or a read-only memory (ROM).

[0141] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor of an electronic device, can perform the following:

[0142] Obtain a set of historical dialogue data, and construct an intent flow based on the set of historical dialogue data;

[0143] Obtain the first dialogue text corresponding to the previous dialogue node and the second dialogue text corresponding to the current dialogue node. Perform intent recognition on the first dialogue text to obtain the first intent, and perform intent recognition on the second dialogue text to obtain the second intent.

[0144] In the intent flow, query the intent node that matches the first intent, take the queried intent node as the matching starting point, and obtain the target intent node that matches the second intent in the intent flow based on the matching starting point and the second intent.

[0145] Obtain the recommended script corresponding to the target intent node, and recommend the recommended script to the preset business personnel.

[0146] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0147] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be embraced within the invention. No appended diagram markings in the claims should be construed as limiting the scope of the claims.

[0148] The blockchain referred to in this invention is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked together using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include an underlying blockchain platform, a platform product service layer, and an application service layer.

[0149] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0150] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in a system claim may also be implemented by a single unit or device through software or hardware. The term "second class" is used to indicate names and does not indicate any specific order.

[0151] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for recommending sales scripts, characterized in that, The method includes: Obtain a historical dialogue corpus set, split the historical dialogue corpus set into a normal business scenario corpus set and an abnormal business scenario corpus set, based on the normal business scenario corpus set, identify and associate normal customer intent nodes with business intent nodes to obtain the main process, based on the abnormal business scenario corpus set, identify abnormal customer intent nodes and obtain the corresponding processing scripts to obtain the sub-process, and combine the main process and the sub-process to obtain the intent process. Obtain the first dialogue text corresponding to the previous dialogue node and the second dialogue text corresponding to the current dialogue node. Perform intent recognition on the first dialogue text to obtain the first intent, and perform intent recognition on the second dialogue text to obtain the second intent. In the intent flow, query intent nodes that match the first intent, and use the queried intent nodes as the matching starting point. In the intent flow, obtain sub-intent nodes connected to the matching starting point, calculate the matching degree between the second intent and each sub-intent node, and select sub-intent nodes that meet the preset matching degree conditions as target intent nodes. Alternatively, when there are no sub-intent nodes that meet the preset matching degree conditions, select intent nodes in the main flow that meet the preset distance conditions from the second intent as target intent nodes. If there are no intent nodes in the main flow that meet the preset distance conditions, query intent nodes that meet the preset exception handling conditions in the sub-flow as target intent nodes. Obtain the recommended script corresponding to the target intent node, and recommend the recommended script to the preset business personnel.

2. The script recommendation method as described in claim 1, characterized in that, The process of identifying and associating normal customer intent nodes with business intent nodes based on the normal business scenario corpus to obtain the main process, and identifying abnormal customer intent nodes and obtaining corresponding processing scripts based on the abnormal business scenario corpus to obtain the sub-process, includes: By identifying the customer role and customer intent of each corpus in the normal business scenario corpus, a normal customer intent node is obtained; by identifying the business role and business intent of each corpus in the normal business scenario corpus, a business intent node is obtained. Obtain the recommended dialogue corresponding to each business intent node, and use the obtained recommended dialogue as the processing branch of the corresponding business intent node; According to the preset business processing rules, the normal customer intent node and the business intent node are associated to obtain the main process; Identify the customer roles and customer intentions of each corpus in the abnormal business scenario corpus to obtain abnormal customer intention nodes; Obtain the processing script corresponding to each abnormal customer intent node, and use it as the processing branch of the corresponding abnormal customer intent node to obtain a sub-process.

3. The script recommendation method as described in claim 2, characterized in that, The step of splitting the historical dialogue corpus into a normal business scenario corpus and an abnormal business scenario corpus includes: The historical dialogue corpus is labeled according to preset business scenario tags; Clustering is performed on the annotated historical dialogue corpus to obtain a preset number of cluster centers; The historical dialogue data belonging to any of the cluster centers mentioned above will be used as the normal business scenario corpus. Historical dialogue data that does not belong to any of the cluster centers mentioned above will be used as a corpus set for abnormal business scenarios.

4. The script recommendation method as described in claim 2, characterized in that, The process of associating the normal customer intent node and the business intent node according to preset business processing rules to obtain the main flow includes: In the preset intent library, query the intent ID corresponding to each normal customer intent node and each business intent node; Based on the preset business processing rules, identify the sequential relationship of each intent ID; The main process is obtained by connecting the corresponding normal customer intent nodes and each of the business intent nodes according to the order relationship.

5. The script recommendation method as described in claim 1, characterized in that, Before querying the intent node that matches the first intent in the intent flow, the method further includes: Identify whether the second dialogue text is the dialogue text acquired for the first time; When the second dialogue text is the dialogue text acquired for the first time, the start node in the intent flow is identified and the start node is used as the matching starting point.

6. The script recommendation method as described in claim 1, characterized in that, The step of performing intent recognition on the first dialogue text to obtain the first intent includes: The first dialogue text is transformed into a word vector to obtain a text vector. Text features of the text vector are extracted using the convolutional layers of a pre-trained intent recognition model; The text features are converted into a feature matrix using the fully connected layer of the pre-trained intent recognition model; The probability value between the feature matrix and the preset intent label is calculated using a preset activation function, and the intent label corresponding to the probability value greater than the preset probability value threshold is selected as the first intent.

7. A script recommendation device, characterized in that, The device includes: Intent flow construction module: used to obtain a historical dialogue corpus set, split the historical dialogue corpus set into a normal business scenario corpus set and an abnormal business scenario corpus set, based on the normal business scenario corpus set, identify and associate normal customer intent nodes and business intent nodes to obtain the main flow, based on the abnormal business scenario corpus set, identify abnormal customer intent nodes and obtain the corresponding processing scripts to obtain sub-flows, and combine the main flow and the sub-flows to obtain the intent flow. The text intent recognition module is used to obtain the first dialogue text corresponding to the previous dialogue node and the second dialogue text corresponding to the current dialogue node, perform intent recognition on the first dialogue text to obtain the first intent, and perform intent recognition on the second dialogue text to obtain the second intent. The text and flow matching module is used to query intent nodes that match the first intent in the intent flow, take the queried intent nodes as the matching starting point, obtain sub-intent nodes connected to the matching starting point in the intent flow, calculate the matching degree between the second intent and each sub-intent node, select sub-intent nodes that meet the preset matching degree conditions as target intent nodes, or, when there are no sub-intent nodes that meet the preset matching degree conditions, select intent nodes that meet the preset distance conditions from the second intent in the main flow as target intent nodes; if there are no intent nodes that meet the preset distance conditions in the main flow, query intent nodes that meet the preset exception handling conditions in the sub-flow as target intent nodes. The recommended script acquisition module is used to acquire the recommended script corresponding to the target intent node and recommend the recommended script to preset business personnel.

8. An electronic device, characterized in that, The electronic device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor to enable the at least one processor to perform the speech recommendation method as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the speech recommendation method as described in any one of claims 1 to 6.

Images