Robot verbal skill updating method and system, computer equipment and medium

Abstract

The invention provides a robot verbal skill updating method and system, computer equipment and a medium, and the method comprises the steps: obtaining target data, including a first verbal skill and a second verbal skill; generating a reference verbal skill for the robot based on the first verbal skill, and adding the reference verbal skill and the second verbal skill into a candidate verbal skill library; evaluating the verbal skill in the candidate verbal skill library according to the verbal skill evaluation dimension, and taking the verbal skill in the candidate verbal skill library according with the verbal skill evaluation dimension as a target verbal skill; and automatically performing verbal skill updating on the robot according to the target verbal skill, so that the robot can dialogue with the first person according to the updated verbal skill. According to the method, the target text can be generated by dynamically analyzing a large amount of verbal skill data, and finally the verbal skill updating is performed on the robot by using the target text, so that the whole verbal skill updating process greatly reduces the manual participation degree, improves the robot verbal skill optimization efficiency while reducing the labor cost, and improves the verbal skill optimization efficiency. Therefore, the verbal skill updating iteration period is shortened, and the verbal skill updating optimization time is saved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method and system for updating robot speech, computer equipment and media. Background Art

[0002] For the freight industry, inviting more qualified drivers is a key factor in business expansion. Since different freight companies cover different business scopes, the content of the invitations is also very different, such as inviting drivers to join through certification, inviting drivers to pay membership fees, etc. At the same time, the characteristics of drivers in different cities are also different, so the requirements for the business taken by drivers are often different, so targeted outbound calls are needed.

[0003] However, currently, there are major pain points in the online operation and optimization of targeted outbound calls: (1) High time and labor costs: The entire outbound call script requires a large amount of manpower to cover. If the script is not effective, it is necessary to manually analyze the reasons for the failure of driver invitations in different cities, and then summarize the reasons and optimize the script. This cycle repeats, the operation manpower is exhausted and the outbound call effect cannot be quickly and effectively followed up. Moreover, manually analyzing outbound call data, urban driver group characteristics and driver concerns is time-consuming and labor-intensive at each stage and cannot be carried out on a large scale. Moreover, the analysis is often fragmented and cannot take into account the data of all stages. (2) Poor automation capabilities: Since script optimization requires a lengthy and inefficient process, and as the business expands to other cities, it is impossible for humans to conduct data analysis in a timely manner and complete the confirmation of the optimal script. In addition, the optimization of the entire outbound call technology link is slow, the iteration cycle is often relatively long, and a large number of experiments are required for effect verification and analysis, and it is impossible to achieve full-link automatic script optimization and update. Summary of the invention

[0004] In view of the shortcomings of the prior art described above, the purpose of the present application is to provide a robot speech updating method and system, computer equipment and medium, so as to solve the technical problems existing in the prior art.

[0005] To achieve the above objectives and other related objectives, the present application provides a method for updating robot speech, comprising the following steps:

[0006] Acquire target data, where the target data includes a first speech obtained when a first person and a second person have a conversation, and a second speech obtained when the first person and a robot have a conversation;

[0007] Generate a reference speech for the robot based on the first speech, and add the reference speech and the second speech to a candidate speech library established in advance or in real time;

[0008] Evaluate the speech in the candidate speech library according to the speech evaluation dimension determined in advance or in real time, and use the speech in the candidate speech library that meets the speech evaluation dimension as the target speech;

[0009] The robot's speech is automatically updated according to the target speech, so that the robot can communicate with the first person according to the updated speech.

[0010] In one embodiment of the present application, the process of automatically updating the robot's speech according to the target speech includes:

[0011] Adding the target speech to a pre-established or real-time archive speech library, selecting part of the target speech from the archive speech library, and automatically updating the speech of the robot using the selected part of the target speech;

[0012] Alternatively, the target speech is added to an archive speech library established in advance or in real time, and all the target speech in the archive speech library are used to automatically update the speech of the robot.

[0013] In one embodiment of the present application, the process of generating a reference speech for the robot based on the first speech includes:

[0014] Performing feature marking on the first speech, wherein the feature marking includes at least one of a city, a person level, a dialogue text, and a first person feature;

[0015] The first speech and the feature marking result of the first speech are input into a large model generated in advance or in real time, and the large model is used for learning to output node reference speech and intention reference speech for the robot; wherein the large model is generated based on neural network training.

[0016] In one embodiment of the present application, the process of evaluating the speech in the candidate speech library according to the speech evaluation dimension determined in advance or in real time includes:

[0017] Obtaining pre-determined or real-time evaluation dimensions of speech techniques, wherein the evaluation dimensions of speech techniques include those determined based on speech technique attractiveness, speech technique personification, and speech technique fit;

[0018] According to the attractiveness of the speech, the speech in the candidate speech library is evaluated for speech attractiveness; and according to the anthropomorphism of the speech, the speech in the candidate speech library is evaluated for speech anthropomorphism; and according to the speech fit, the speech in the candidate speech library is evaluated for speech fit;

[0019] When the evaluation result of the attractiveness of a certain speech in the candidate speech library meets the pre-determined or real-time determined attractiveness of the speech, and / or the evaluation result of the personification of the certain speech meets the pre-determined or real-time determined personification of the speech, and / or the evaluation result of the fit of the certain speech meets the pre-determined or real-time determined fit of the speech, the certain speech is marked as meeting the speech update conditions, and the certain speech is used as the target speech.

[0020] In one embodiment of the present application, the process of obtaining target data includes: receiving conversation data replied by the first person based on the first invitation task, and associating the speech text in the conversation data replied by the first person based on the first invitation task to obtain the first speech; wherein, the first invitation task includes at least one of an authentication invitation, a fee payment invitation, a recall invitation, an attendance invitation and an event invitation issued by the second person to the first person.

[0021] In one embodiment of the present application, the process of acquiring the target data further includes: receiving the conversation data answered by the first person based on the first return question, and associating the speech text in the conversation data answered by the first person based on the first return question to obtain the second speech; wherein the first return question includes the speech generated by the robot based on the first speech;

[0022] Alternatively, the conversation data answered by the first person based on the second follow-up question is received, and the speech text in the conversation data answered by the first person based on the second follow-up question is associated to obtain the second speech; wherein, the second follow-up question includes the robot generated based on the first invitation task, and the first invitation task includes at least one of the authentication invitation, fee payment invitation, recall invitation, attendance invitation and event invitation issued by the second person to the first person.

[0023] In one embodiment of the present application, the first person includes a driver, and the second person includes a manual customer service.

[0024] The present application also provides a robot speech updating system, the system comprising:

[0025] A data acquisition module, used to obtain target data, wherein the target data includes a first speech obtained when a first person and a second person have a conversation, and a second speech obtained when the first person and the robot have a conversation;

[0026] A reference speech module, used to generate a reference speech for the robot based on the first speech, and add the reference speech and the second speech to a candidate speech library established in advance or in real time;

[0027] A speech evaluation module, used to evaluate the speech in the candidate speech library according to the speech evaluation dimension determined in advance or in real time, and use the speech in the candidate speech library that meets the speech evaluation dimension as the target speech;

[0028] The speech updating module automatically updates the speech of the robot according to the target speech, so that the robot can communicate with the first person according to the updated speech.

[0029] The present application also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement any one of the above-mentioned robot speech updating methods.

[0030] The present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any one of the above-mentioned robot speech updating methods.

[0031] As described above, the present application provides a robot speech updating method and system, computer equipment and medium, which have the following beneficial effects: obtaining target data, the target data including a first speech obtained when a first person and a second person have a conversation, and a second speech obtained when the first person and a robot have a conversation; then generating a reference speech for the robot based on the first speech, and adding the reference speech and the second speech to a candidate speech library established in advance or in real time; and evaluating the speech in the candidate speech library according to a speech evaluation dimension determined in advance or in real time, and taking the speech in the candidate speech library that meets the speech evaluation dimension as the target speech; finally, automatically updating the robot's speech according to the target speech, so that the robot can have a conversation with the first person according to the updated speech. It can be seen from this that the present application can generate candidate speech by dynamically analyzing a large amount of speech data obtained in advance or in real time, and then confirm the target text from the candidate speech through the speech evaluation dimension, and finally use the target text to update the robot's speech. The entire speech update process not only greatly reduces the degree of human participation, but also can automatically update the speech, while reducing labor costs and improving the efficiency of robot speech optimization, thereby shortening the speech update iteration cycle and saving speech update optimization time. In addition, the present application generates reference speech for the robot by using the speech obtained during manual dialogue, which can not only perform single-point analysis of question and answer intentions and achieve single-point triggering; but also comprehensively control the final speech effect from different dimensions, eliminate the fragmentation problem that occurs in the speech optimization process, and optimize the whole link to improve the robot's outbound call effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 A schematic diagram of an exemplary system architecture for applying the technical solutions in one or more embodiments of the present application;

[0033] Figure 2 A flowchart of a method for updating robot speech provided in an embodiment of the present application;

[0034] Figure 3 A schematic diagram of the principle of a method for updating robot speech applied to a freight scenario provided by an embodiment of the present application;

[0035] Figure 4 A schematic diagram of forming an archived speech library provided in an embodiment of the present application;

[0036] Figure 5 A schematic diagram of the hardware structure of a robot speech updating method provided in an embodiment of the present application;

[0037] Figure 6 The figure is a schematic diagram of the hardware structure of a computer device suitable for implementing one or more embodiments of the present application. DETAILED DESCRIPTION

[0038] The following is an explanation of the embodiments of the present application by specific specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in this specification. The present application can also be implemented or applied by other different specific embodiments, and the details in this specification can also be based on different viewpoints and applications, without departing from the spirit of the present application. Various modifications or changes are made. It is understood that, in the absence of conflict, the following embodiments and the features in the embodiments can be combined with each other. In addition, it is understood that the illustrations provided in the following embodiments only illustrate the basic concept of the present application in a schematic manner, and only the components related to the present application are shown in the drawings rather than the number, shape and size of the components during the actual implementation. The type, quantity and ratio of each component during its actual implementation can be a random change, and its component layout type may also be more complicated.

[0039] Figure 1 A schematic diagram of an exemplary system architecture to which the technical solution in one or more embodiments of the present application can be applied is shown. Figure 1 As shown, the system architecture 100 may include a terminal device 110, a network 120, and a server 130. The terminal device 110 may include various electronic devices such as a smart phone, a tablet computer, a laptop computer, and a desktop computer. The server 130 may be an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The network 120 may be a communication medium of various connection types that can provide a communication link between the terminal device 110 and the server 130, such as a wired communication link or a wireless communication link.

[0040] According to the implementation requirements, the system architecture in the embodiment of the present application can have any number of terminal devices, networks and servers. For example, the server 130 can be a server group composed of multiple server devices. In addition, the technical solution provided in the embodiment of the present application can be applied to the terminal device 110, can also be applied to the server 130, or can be implemented by the terminal device 110 and the server 130 together, and the present application does not make any special restrictions on this.

[0041] In one embodiment of the present application, the terminal device 110 or the server 130 of the present application can obtain target data, the target data including a first speech obtained when a first person and a second person have a conversation, and a second speech obtained when the first person and a robot have a conversation; then generate a reference speech for the robot based on the first speech, and add the reference speech and the second speech to a candidate speech library established in advance or in real time; and evaluate the speech in the candidate speech library according to a speech evaluation dimension determined in advance or in real time, and use the speech in the candidate speech library that meets the speech evaluation dimension as the target speech; finally, automatically update the robot's speech according to the target speech, so that the robot can have a conversation with the first person according to the updated speech. By using the terminal device 110 or the server 130 to execute the robot's speech updating method, a large amount of speech data obtained in advance or in real time can be dynamically analyzed to generate candidate speech, and then the target text can be confirmed from the candidate speech through the speech evaluation dimension, and finally the robot's speech can be updated using the target text. The entire speech updating process not only greatly reduces the degree of manual participation, but also can automatically update the speech, while reducing labor costs and improving the robot's speech optimization efficiency, thereby shortening the speech update iteration cycle and saving speech update optimization time. In addition, by using the speech obtained during manual conversation to generate reference speech for the robot, not only can the question and answer intention be analyzed at a single point and triggered at a single point; but also the final speech effect can be fully controlled from different dimensions, eliminating the fragmentation problem that occurs during the speech optimization process, and the full-link optimization improves the robot's outbound call effect.

[0042] The above section introduces the contents of an exemplary system architecture that applies the technical solution of the present application. Next, we will continue to introduce the robot speech updating method of the present application.

[0043] In an exemplary embodiment of the present application, Figure 2 As shown, a method for updating robot speech is provided, comprising the following steps:

[0044] S210, obtaining target data, the target data including the first speech obtained when the first person and the second person have a conversation, and the second speech obtained when the first person and the robot have a conversation. In some exemplary embodiments, the first person includes but is not limited to drivers, such as freight drivers, passenger drivers, etc.; the second person includes manual customer service, including but not limited to manual customer service of freight enterprise platforms, manual customer service of passenger enterprise platforms, etc., such as manual customer service for inviting drivers to undergo certification, manual customer service for inviting drivers to pay insurance fees, manual customer service for inviting drivers to pay membership fees, manual customer service for recalling lost drivers, manual customer service for inviting drivers to attend events, etc.

[0045] S220, generating reference speech for the robot based on the first speech, and adding the reference speech and the second speech to a candidate speech library established in advance or in real time.

[0046] S230, evaluate the speech in the candidate speech library according to the speech evaluation dimension determined in advance or in real time, and use the speech in the candidate speech library that meets the speech evaluation dimension as the target speech. In some exemplary embodiments, the speech evaluation dimension includes determining according to the speech attractiveness, speech anthropomorphism and speech fit. For example, the speech evaluation dimension is composed of one or more of speech attractiveness, speech anthropomorphism and speech fit.

[0047] S240, automatically updating the robot's speech according to the target speech, so that the robot can communicate with the first person according to the updated speech.

[0048] In some exemplary embodiments, the process of obtaining target data may include: receiving conversation data replied by the first person based on the first invitation task, and associating the speech text in the conversation data replied by the first person based on the first invitation task to obtain the first speech; wherein the first invitation task includes at least one of an authentication invitation, a fee payment invitation, a recall invitation, an attendance invitation and an event invitation issued by the second person to the first person.

[0049] In some exemplary embodiments, the process of obtaining target data may also include: receiving conversation data answered by the first person based on the first follow-up question, and associating the speech text in the conversation data answered by the first person based on the first follow-up question to obtain a second speech; wherein the first follow-up question includes the speech generated by the robot based on the first speech.

[0050] In some exemplary embodiments, the process of obtaining target data may also include: receiving conversation data answered by the first person based on the second follow-up question, and associating the speech text in the conversation data answered by the first person based on the second follow-up question to obtain the second speech; wherein the second follow-up question includes the robot generated based on the first invitation task, and the first invitation task includes at least one of the authentication invitation, fee payment invitation, recall invitation, attendance invitation and event invitation issued by the second person to the first person.

[0051] In some exemplary embodiments, the process of generating reference speech for a robot based on a first speech may include: marking the first speech with features, wherein the feature marking includes at least one of a city, a personnel level, a conversation text, and a first personnel feature; inputting the first speech and the feature marking result of the first speech into a pre- or real-time generated large model, learning through the large model, and outputting node reference speech and intention reference speech for the robot; wherein the large model is generated based on neural network training. As an example, a large model may refer to a machine learning model with large-scale parameters and complex computing structures, which can process massive amounts of data and complete various complex tasks, such as natural language processing, computer vision processing, speech recognition, and other tasks.

[0052] In some exemplary embodiments, the process of automatically updating the robot's speech based on the target speech may include: adding the target speech to a pre- or real-time established archive speech library, selecting part of the target speech from the archive speech library, and automatically updating the robot's speech using the selected part of the target speech.

[0053] In some exemplary embodiments, the process of automatically updating the robot's speech based on the target speech may also include: adding the target speech to a pre- or real-time established archive speech library, and automatically updating the robot's speech using all the target speech in the archive speech library.

[0054] In some exemplary embodiments, the process of evaluating the speech in the candidate speech library according to the speech evaluation dimension determined in advance or in real time includes: obtaining the speech evaluation dimension determined in advance or in real time, wherein the speech evaluation dimension includes determining according to the speech attractiveness, speech personification and speech fit; for example, the speech evaluation dimension is composed of one or more of speech attractiveness, speech personification and speech fit. Evaluate the speech in the candidate speech library for speech attractiveness according to the speech attractiveness; and, evaluate the speech personification of the speech in the candidate speech library according to the speech personification; and, evaluate the speech fit of the speech in the candidate speech library according to the speech fit. When the evaluation result of the attractiveness of a certain speech in the candidate speech library meets the pre-determined or real-time determined attractiveness of the speech, and / or the evaluation result of the personification of the speech meets the pre-determined or real-time determined personification of the speech, and / or the evaluation result of the fit of the speech meets the pre-determined or real-time determined fit of the speech, the speech is marked as meeting the speech update conditions and the speech is used as the target speech.

[0055] In another exemplary embodiment of the present application, a method for updating robot speech is provided, which can be applied to freight scenarios, such as Figure 3 As shown, the method comprises the following steps:

[0056] Configure the driver feature big model agent; the driver feature big model agent can be configured to retrieve and analyze manual invitation data in different cities, and the manual invitation data includes the conversation data obtained by the manual customer service for different invitation tasks issued to driver groups in different cities. Among them, the invitation task can be at least one of inviting drivers to authenticate, inviting drivers to pay insurance premiums, inviting drivers to pay membership fees, recalling lost drivers, inviting drivers to attend, and inviting drivers to participate in activities. And input the city, invitation personnel level, conversation text, driver characteristics, etc. into the driver feature big model agent, so that the driver feature big model agent summarizes and outputs the words that can be referenced by each node in the outbound call robot (also known as the outbound call robot bot, robot or bot) (also known as the node reference words) and the words that can be referenced by the intention (also known as the intention reference words).

[0057] Configure the driver return visit model agent; the driver return visit model agent can be configured to formulate an outbound call robot bot for driver return visits. Different questions will be set in the outbound call robot bot for driver return visits, and then the driver's attention characteristics in the city will be collected. The driver return visit model agent can automatically generate return visit questions based on the city's specific information, and fill them into the return call robot bot for outbound calls, so as to collect the driver's reply data, which is recorded as the driver return visit data.

[0058] Configure the candidate speech expert big model agent; the candidate speech expert big model agent can be configured to respectively obtain the characteristics of the node reference speech, the characteristics of the intention reference speech and the characteristics of the driver's return data, and use the characteristics of the node reference speech, the characteristics of the intention reference speech and the characteristics of the driver's return data as the input of the candidate speech expert big model agent, and summarize and organize the candidate speech through the candidate speech expert big model agent, and store the speech output by the candidate speech expert big model agent into the candidate speech library.

[0059] Configure the critic agent (also called the evaluation big model agent). The critic agent can be configured to develop three types of evaluation big model agents, namely critic agent1, critic agent2, and critic agent3. The critic agent makes judgments from three dimensions: the attractiveness of the speech, the personification of the speech, and the fit of the speech. For example, critic agent1 can judge from the dimension of the attractiveness of the speech, critic agent2 can judge from the dimension of the personification of the speech, and critic agent3 can judge from the dimension of the fit of the speech. Only when the three critic agents, critic agent1, criticagent2, and critic agent3, agree that the current speech meets the requirements for the update and replacement of the speech, will the current speech be marked as the target speech and included in the final archived speech library. Figure 4 As shown in the figure, after the candidate speech expert big model agent, a candidate speech library for different nodes or intentions can be obtained. Each node or intention has a variety of different big model agent regular speech for reference. Then, through the three critic agents, critic agent1, critic agent2, and critic agent3, to evaluate this candidate speech library, the final speech of the outbound call robot bot corresponding to the node and intention can be obtained. Finally, the only speech confirmed by the three critic agents, critic agent1, critic agent2, and critic agent3, is used as the optimal speech.

[0060] Update the online outbound call robot bot's scripts, including automatically and quickly updating and confirming the optimal scripts for the online outbound call robot bot based on the scripts in the archived script library.

[0061] In addition, the process control big model agent can also be configured, including configuring the process control big model agent as the control brain to control the entry and exit conditions of other big model agents. For example, the process control big model agent can determine other big model agents that need to be transferred for the next step of processing based on the results obtained at the current moment. That is, through a master control big model agent, the functional entry and exit of each sub-big model agent can be completed, thereby completing the automated processing capabilities of different links.

[0062] In summary, the present application provides a robot speech updating method, by acquiring target data, the target data includes the first speech obtained when the first person and the second person have a conversation, and the second speech obtained when the first person and the robot have a conversation; then based on the first speech, a reference speech for the robot is generated, and the reference speech and the second speech are added to the candidate speech library established in advance or in real time; and the speech in the candidate speech library is evaluated according to the speech evaluation dimension determined in advance or in real time, and the speech in the candidate speech library that meets the speech evaluation dimension is used as the target speech; finally, the robot's speech is automatically updated according to the target speech, so that the robot can talk with the first person according to the updated speech. It can be seen from this that the method can generate candidate speech by dynamically analyzing a large amount of speech data obtained in advance or in real time, and then confirm the target text from the candidate speech through the speech evaluation dimension, and finally use the target text to update the robot's speech. The entire speech update process not only greatly reduces the degree of manual participation, but also can automatically update the speech, while reducing labor costs and improving the robot's speech optimization efficiency, thereby shortening the speech update iteration cycle and saving speech update optimization time. In addition, this method generates reference words for the robot by utilizing the words obtained during manual conversations. It can not only perform single-point analysis of question-answering intentions and achieve single-point triggering, but also comprehensively control the final words effect from different dimensions, eliminating the fragmentation problem that occurs during the words optimization process, and optimizing the entire chain to improve the robot's outbound call effect.

[0063] In addition, by utilizing the rich natural semantic understanding capabilities of the big model, this method can analyze the feature data required for different links, and automatically summarize the words to complete the storage update operation. At the same time, by using the big model agent for automatic control, the fragmentation problem that occurred in the previous targeted outbound call word optimization process is completely eliminated, thereby forming a full-link optimization effect. Although the relevant technology can compare and screen out words from the preset word library for reply, this reply method is a static reply and belongs to word screening; while this method uses the big model agent to dynamically analyze the existing data and generate candidate words, and then finally confirms the words through dimensional evaluation and applies them to the online targeted outbound call robot, so that from analysis to optimization to confirmation of the final words, all are controlled by the big model agent, and no manpower is required to intervene in the middle, and it also replaces the static reply process of pre-set words in the original word optimization confirmation process. In addition, with the accelerated expansion of online invitations to cities and the increasing outbound call and analysis costs required for different scenarios, by using large model agents for automated control, a set of scripts suitable for invitation scenarios in different cities can be quickly formed with good scalability. Online outbound calls can be quickly reused without complex operations, which can reduce manpower costs while doubling the effect.

[0064] like Figure 5 As shown, in another exemplary embodiment of the present application, a robot speech updating system is provided, comprising:

[0065] The data acquisition module 510 is used to obtain target data, and the target data includes the first speech obtained when the first person and the second person have a conversation, and the second speech obtained when the first person and the robot have a conversation. In some exemplary embodiments, the first person includes but is not limited to drivers, such as freight drivers, passenger drivers, etc.; the second person includes manual customer service, including but not limited to manual customer service of freight enterprise platforms, manual customer service of passenger enterprise platforms, etc., such as manual customer service for inviting drivers to authenticate, manual customer service for inviting drivers to pay insurance fees, manual customer service for inviting drivers to pay membership fees, manual customer service for recalling lost drivers, manual customer service for inviting drivers to attend, and manual customer service for inviting drivers to participate in activities, etc.

[0066] A reference speech module 520, for generating a reference speech for the robot based on the first speech, and adding the reference speech and the second speech to a candidate speech library established in advance or in real time;

[0067] The speech evaluation module 530 is used to evaluate the speech in the candidate speech library according to the speech evaluation dimension determined in advance or in real time, and use the speech in the candidate speech library that meets the speech evaluation dimension as the target speech. In some exemplary embodiments, the speech evaluation dimension includes determining according to the speech attractiveness, speech anthropomorphism and speech fit. For example, the speech evaluation dimension is composed of one or more of the speech attractiveness, speech anthropomorphism and speech fit.

[0068] The speech updating module 540 automatically updates the speech of the robot according to the target speech, so that the robot can communicate with the first person according to the updated speech.

[0069] In some exemplary embodiments, the process of obtaining target data may include: receiving conversation data replied by the first person based on the first invitation task, and associating the speech text in the conversation data replied by the first person based on the first invitation task to obtain the first speech; wherein the first invitation task includes at least one of an authentication invitation, a fee payment invitation, a recall invitation, an attendance invitation and an event invitation issued by the second person to the first person.

[0070] In some exemplary embodiments, the process of obtaining target data may also include: receiving conversation data answered by the first person based on the first follow-up question, and associating the speech text in the conversation data answered by the first person based on the first follow-up question to obtain a second speech; wherein the first follow-up question includes the speech generated by the robot based on the first speech.

[0071] In some exemplary embodiments, the process of obtaining target data may also include: receiving conversation data answered by the first person based on the second follow-up question, and associating the speech text in the conversation data answered by the first person based on the second follow-up question to obtain the second speech; wherein the second follow-up question includes the robot generated based on the first invitation task, and the first invitation task includes at least one of the authentication invitation, fee payment invitation, recall invitation, attendance invitation and event invitation issued by the second person to the first person.

[0072] In some exemplary embodiments, the process of generating reference speech for a robot based on a first speech may include: marking the first speech with features, wherein the feature marking includes at least one of a city, a personnel level, a conversation text, and a first personnel feature; inputting the first speech and the feature marking result of the first speech into a pre- or real-time generated large model, learning through the large model, and outputting node reference speech and intention reference speech for the robot; wherein the large model is generated based on neural network training. As an example, a large model may refer to a machine learning model with large-scale parameters and complex computing structures, which can process massive amounts of data and complete various complex tasks, such as natural language processing, computer vision processing, speech recognition, and other tasks.

[0073] In some exemplary embodiments, the process of automatically updating the robot's speech based on the target speech may include: adding the target speech to a pre- or real-time established archive speech library, selecting part of the target speech from the archive speech library, and automatically updating the robot's speech using the selected part of the target speech.

[0074] In some exemplary embodiments, the process of automatically updating the robot's speech based on the target speech may also include: adding the target speech to a pre- or real-time established archive speech library, and automatically updating the robot's speech using all the target speech in the archive speech library.

[0075] In some exemplary embodiments, the process of evaluating the speech in the candidate speech library according to the speech evaluation dimension determined in advance or in real time includes: obtaining the speech evaluation dimension determined in advance or in real time, wherein the speech evaluation dimension includes determining according to the speech attractiveness, speech personification and speech fit; for example, the speech evaluation dimension is composed of one or more of speech attractiveness, speech personification and speech fit. Evaluate the speech in the candidate speech library for speech attractiveness according to the speech attractiveness; and, evaluate the speech personification of the speech in the candidate speech library according to the speech personification; and, evaluate the speech fit of the speech in the candidate speech library according to the speech fit. When the evaluation result of the attractiveness of a certain speech in the candidate speech library meets the pre-determined or real-time determined attractiveness of the speech, and / or the evaluation result of the personification of the speech meets the pre-determined or real-time determined personification of the speech, and / or the evaluation result of the fit of the speech meets the pre-determined or real-time determined fit of the speech, the speech is marked as meeting the speech update conditions and the speech is used as the target speech.

[0076] In another exemplary embodiment of the present application, a robot speech updating system is provided. The robot speech updating subsystem can be applied to freight scenarios. For specific application process, see Figure 3 , Figure 4 As well as the other embodiments mentioned above, they will not be described in detail here.

[0077] In summary, the present application provides a robot speech updating system, which obtains target data, including the first speech obtained when the first person and the second person have a conversation, and the second speech obtained when the first person and the robot have a conversation; then generates a reference speech for the robot based on the first speech, and adds the reference speech and the second speech to the candidate speech library established in advance or in real time; and evaluates the speech in the candidate speech library according to the speech evaluation dimension determined in advance or in real time, and takes the speech in the candidate speech library that meets the speech evaluation dimension as the target speech; finally, the robot's speech is automatically updated according to the target speech, so that the robot can talk to the first person according to the updated speech. It can be seen from this that the system can generate candidate speech by dynamically analyzing a large amount of speech data obtained in advance or in real time, and then confirm the target text from the candidate speech through the speech evaluation dimension, and finally use the target text to update the robot's speech. The entire speech update process not only greatly reduces the degree of manual participation, but also can automatically update the speech, while reducing labor costs and improving the optimization efficiency of the robot speech, thereby shortening the speech update iteration cycle and saving speech update optimization time. In addition, this system generates reference words for robots by utilizing the words obtained during manual conversations. It can not only perform single-point analysis of question-answering intentions and achieve single-point triggering, but also comprehensively control the final words effect from different dimensions, eliminating the fragmentation problems that occur during the words optimization process, and improving the robot's outbound call effect through full-link optimization.

[0078] In addition, by utilizing the rich natural semantic understanding capabilities of the big model, this system can analyze the feature data required for different links, and automatically summarize the words to complete the update operation. At the same time, by using the big model agent for automatic control, the fragmentation problem that occurred in the previous targeted outbound call word optimization process is completely eliminated, thus forming a full-link optimization effect. Although the relevant technology can filter out words from the preset word library through comparison for reply, this reply method is a static reply and belongs to word screening; this system uses the big model agent to dynamically analyze the existing data and generate candidate words, and then finally confirms the words through dimensional evaluation and applies them to the online targeted outbound call robot, so that from analysis to optimization to confirmation of the final words, all are controlled by the big model agent, without the need for manpower intervention in the middle, and it also replaces the static reply process of pre-set words in the original word optimization confirmation process. In addition, with the accelerated expansion of online invitations to cities and the increasing outbound call and analysis costs required for different scenarios, by using large model agents for automated control, a set of scripts suitable for invitation scenarios in different cities can be quickly formed with good scalability. Online outbound calls can be quickly reused without complex operations, which can reduce manpower costs while doubling the effect.

[0079] It can be understood that the robot speech updating system provided in the above embodiment and the robot speech updating method provided in the above embodiment belong to the same concept, wherein the specific manner in which the robot speech updating method performs operations has been described in detail in the above method embodiment, and will not be repeated here. In actual applications, the robot speech updating system provided in the above embodiment can distribute the above functions to different functional modules as needed, that is, divide the internal structure of the robot speech updating system into different functional modules, and then implement all or part of the functions of the corresponding functional modules through the robot speech updating method described in the above embodiment, and this is not specifically limited here.

[0080] The present application also provides a computer device, which may include a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program so that the computer device executes Figure 2 The steps of the robot speech updating method. Figure 6 FIG. 1 is a schematic diagram showing a structure of a computer device 1000. Figure 6 As shown, the computer device 1000 includes: a processor 1010 , a memory 1020 , a power supply 1030 , a display unit 1040 , and an input unit 1060 .

[0081] The processor 1010 is the control center of the computer device 1000. It uses various interfaces and lines to connect various components, and executes various functions of the computer device 1000 by running or executing computer programs / instructions stored in the memory 1020, thereby monitoring the computer device 1000 as a whole. In the embodiment of the present application, when the processor 1010 calls the computer program stored in the memory 1020, the following is executed: Figure 2 The steps of the robot speech updating method. Optionally, the processor 1010 may include one or more processing units; preferably, the processor 1010 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface and application, etc., and the modem processor mainly processes wireless communication. In some embodiments, the processor and memory may be implemented on a single chip, and in some embodiments, they may also be implemented separately on separate chips.

[0082] The memory 1020 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, various applications, etc.; the data storage area may store instruction data, etc. created according to the use of the computer device 1000. In addition, the memory 1020 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices, etc.

[0083] The computer device 1000 also includes a power source 1030 (such as a battery) for supplying power to various components. The power source can be logically connected to the processor 1010 through a power management system, so that the power management system can manage functions such as charging, discharging, and power consumption.

[0084] The display unit 1040 may be used to display information input by the user or information provided to the user and various menus of the computer device 1000, and in the embodiment of the present application, is mainly used to display the display interface of each application in the computer device 1000 and objects such as text and pictures displayed in the display interface. The display unit 1040 may include a display panel 1050. The display panel 1050 may be configured in the form of a liquid crystal display (LCD), an organic light-emitting diode (OLED), or the like.

[0085] The input unit 1060 may be used to receive information such as numbers or characters input by the user. The input unit 1060 may include a touch panel 1070 and other input devices 1080. The touch panel 1070, also known as a touch screen, may collect touch operations of the user on or near it (such as operations of the user using a finger, a stylus, or any other suitable object or accessory on or near the touch panel 1070).

[0086] Specifically, the touch panel 1070 can detect the user's touch operation, detect the signal brought by the touch operation, convert these signals into touch point coordinates, send them to the processor 1010, and receive and execute the command sent by the processor 1010. In addition, the touch panel 1070 can be implemented in various types such as resistive, capacitive, infrared and surface acoustic wave. Other input devices 1080 can include but are not limited to one or more of a physical keyboard, function keys (such as volume control keys, power on / off keys, etc.), a trackball, a mouse, a joystick, etc.

[0087] Of course, the touch panel 1070 can cover the display panel 1050. When the touch panel 1070 detects a touch operation on or near it, it is transmitted to the processor 1010 to determine the type of touch event, and then the processor 1010 provides a corresponding visual output on the display panel 1050 according to the type of touch event. Figure 6 In the embodiment, the touch panel 1070 and the display panel 1050 are two independent components to implement the input and output functions of the computer device 1000, but in some embodiments, the touch panel 1070 and the display panel 1050 can be integrated to implement the input and output functions of the computer device 1000.

[0088] The computer device 1000 may further include one or more sensors, such as a pressure sensor, a gravity acceleration sensor, a proximity light sensor, etc. Of course, according to the needs of specific applications, the computer device 1000 may also include other components such as a camera.

[0089] The embodiment of the present application also provides a computer-readable storage medium, in which a computer program / instruction is stored. When the computer program / instruction is executed by a processor, the above-mentioned device can perform the above-mentioned Figure 2 The steps of the robot speech updating method.

[0090] It can be understood by those skilled in the art that Figure 6 This is only an example of a computer device and does not constitute a limitation on the device. The device may include more or fewer components than shown in the figure, or combine certain components, or different components. For the convenience of description, the above parts are divided into modules (or units) according to their functions and described separately. Of course, when implementing this application, the functions of each module (or unit) can be implemented in the same or multiple software or hardware.

[0091] Those skilled in the art should understand that the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be applied to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the functions in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the functions specified in the process. Figure 1 A process or multiple processes and / or boxes Figure 1These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0092] It is understandable that the above embodiments collect, store, use, process, transmit, provide, disclose, delete, etc. related data (such as target data, etc.) with or with the consent of the user. For example, the target data is authorized with the knowledge and consent of the user; or it is actively provided by the user after reading the relevant instructions, or it is actively authorized / provided / uploaded by the user when using some or all of the functions described in the above embodiments, or it is obtained by other means / channels with or with the consent of the user.

[0093] It is understood that, although the terms first, second, etc. may be used to describe persons, etc. in the embodiments of the present application, these terms are only used to distinguish persons from each other. For example, without departing from the scope of the embodiments of the present application, the first person may also be referred to as the second person, and similarly, the second person may also be referred to as the first person.

[0094] The above embodiments are merely illustrative of the principles and effects of the present application and are not intended to limit the present application. Anyone familiar with the technology may modify or change the above embodiments without violating the spirit and scope of the present application. Therefore, all equivalent modifications or changes made by a person of ordinary skill in the art without departing from the spirit and technical ideas disclosed in the present application shall still be covered by the claims of the present application.

Claims

1. A method for updating robot speech, characterized in that: The method comprises: Acquire target data, where the target data includes a first speech obtained when a first person and a second person have a conversation, and a second speech obtained when the first person and a robot have a conversation; Generate a reference speech for the robot based on the first speech, and add the reference speech and the second speech to a candidate speech library established in advance or in real time; Evaluate the speech in the candidate speech library according to the speech evaluation dimension determined in advance or in real time, and use the speech in the candidate speech library that meets the speech evaluation dimension as the target speech; The robot's speech is automatically updated according to the target speech, so that the robot can communicate with the first person according to the updated speech.

2. The robot speech updating method according to claim 1, characterized in that: The process of automatically updating the robot's speech according to the target speech includes: Adding the target speech to a pre-established or real-time archive speech library, selecting part of the target speech from the archive speech library, and automatically updating the speech of the robot using the selected part of the target speech; Alternatively, the target speech is added to an archive speech library established in advance or in real time, and all the target speech in the archive speech library are used to automatically update the speech of the robot.

3. The robot speech updating method according to claim 1, characterized in that: The process of generating a reference speech for the robot based on the first speech includes: Performing feature marking on the first speech, wherein the feature marking includes at least one of a city, a person level, a dialogue text, and a first person feature; The first speech and the feature marking result of the first speech are input into a large model generated in advance or in real time, and the large model is used for learning to output node reference speech and intention reference speech for the robot; wherein the large model is generated based on neural network training.

4. The method for updating robot speech according to any one of claims 1 to 3, characterized in that: The process of evaluating the speech in the candidate speech library according to the speech evaluation dimension determined in advance or in real time includes: Obtaining pre-determined or real-time evaluation dimensions of speech techniques, wherein the evaluation dimensions of speech techniques include those determined based on speech technique attractiveness, speech technique personification, and speech technique fit; According to the attractiveness of the speech, the speech in the candidate speech library is evaluated for speech attractiveness; and according to the anthropomorphism of the speech, the speech in the candidate speech library is evaluated for speech anthropomorphism; and according to the speech fit, the speech in the candidate speech library is evaluated for speech fit; When the evaluation result of the attractiveness of a certain speech in the candidate speech library meets the pre-determined or real-time determined attractiveness of the speech, and / or the evaluation result of the personification of the certain speech meets the pre-determined or real-time determined personification of the speech, and / or the evaluation result of the fit of the certain speech meets the pre-determined or real-time determined fit of the speech, the certain speech is marked as meeting the speech update conditions, and the certain speech is used as the target speech.

5. The robot speech updating method according to claim 1 or 3, characterized in that: The process of obtaining target data includes: receiving the conversation data replied by the first person based on the first invitation task, and associating the speech text in the conversation data replied by the first person based on the first invitation task to obtain the first speech; wherein, the first invitation task includes at least one of the authentication invitation, fee payment invitation, recall invitation, attendance invitation and event invitation issued by the second person to the first person.

6. The robot speech updating method according to claim 5, characterized in that: The process of acquiring the target data further includes: receiving the dialogue data answered by the first person based on the first return question, and associating the speech text in the dialogue data answered by the first person based on the first return question to obtain the second speech; wherein the first return question includes the speech text generated by the robot based on the first speech text; Alternatively, the conversation data answered by the first person based on the second follow-up question is received, and the speech text in the conversation data answered by the first person based on the second follow-up question is associated to obtain the second speech; wherein, the second follow-up question includes the robot generated based on the first invitation task, and the first invitation task includes at least one of the authentication invitation, fee payment invitation, recall invitation, attendance invitation and event invitation issued by the second person to the first person.

7. The robot speech updating method according to claim 1, characterized in that: The first personnel include drivers, and the second personnel include manual customer service.

8. A robot speech updating system, characterized in that: The system comprises: A data acquisition module, used to obtain target data, wherein the target data includes a first speech obtained when a first person and a second person have a conversation, and a second speech obtained when the first person and the robot have a conversation; A reference speech module, used to generate a reference speech for the robot based on the first speech, and add the reference speech and the second speech to a candidate speech library established in advance or in real time; A speech evaluation module, used to evaluate the speech in the candidate speech library according to the speech evaluation dimension determined in advance or in real time, and use the speech in the candidate speech library that meets the speech evaluation dimension as the target speech; The speech updating module automatically updates the speech of the robot according to the target speech, so that the robot can communicate with the first person according to the updated speech.

9. A computer device, characterized in that: It includes a memory, a processor and a computer program stored in the memory, and the processor executes the computer program to implement the robot speech updating method described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the robot speech updating method described in any one of claims 1 to 7 is implemented.

Images