Control Method, Device, and Medium of a Debate Robot Based on Speech Processing

Abstract

The present disclosure provides a control method, device, medium and speech processing-based debate robot system for a speech processing-based debate robot. The control method for the speech processing-based debate robot includes: obtaining a debate material training set; inputting the debate material training set into a preset training model to control the debate robot to learn the debate materials, where the preset training model includes a low-rank adapter for optimizing the preset training model; inputting a preset debate scenario, the debate duration corresponding to the preset debate scenario, and the debater role of the debate robot into the debate robot; and controlling the debate robot to output target debate language within the debate duration corresponding to the preset debate scenario according to the preset debate scenario and the speech information of the debate scenario. Through the present disclosure, a robot debater can be generated to adapt to various debate scenarios and break through time limitations to accompany humans in debate training or debate competitions.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and specifically, to a control method, device, medium, and a debate robot system based on speech processing for a debate robot. Background Art

[0002] College students can improve their oratory skills by participating in college debate competitions. In a debate competition, debaters can express their viewpoints and arguments clearly, accurately, and systematically, which is of great significance for the future career development and interpersonal communication of students. At the same time, debate competitions can also improve logical thinking ability, problem - analysis ability, and dialectical thinking ability, and contribute to cultivating students' teamwork and leadership skills. The topics of debate competitions cover various academic fields, which can increase students' knowledge reserves, broaden their horizons, and improve their comprehensive qualities.

[0003] However, practicing debate requires multiple students to practice together. Once a student is absent from practice, it will affect the efficiency of debate training. Nowadays, with the progress of artificial intelligence technology, robots have become an indispensable part of modern people's lives. In the field of debate, it has become an urgent need to use debate robots for debate training or debate competitions. Summary of the Invention

[0004] Aiming at the deficiencies in the prior art, the purpose of the present disclosure is to provide a control method, device, medium, and a debate robot system based on speech processing for a debate robot.

[0005] According to a first aspect of the present invention, there is provided a control method for a debate robot based on speech processing, including:

[0006] Obtaining a debate material training set;

[0007] Inputting the debate material training set into a preset training model, and controlling the debate robot to learn debate materials. The preset training model includes a low - rank adapter, and the low - rank adapter is used to optimize the preset training model;

[0008] Inputting a preset debate scenario, the debate duration corresponding to the preset debate scenario, and the debater role of the debate robot into the debate robot;

[0009] Controlling the debate robot to output target debate language within the debate duration corresponding to the preset debate scenario according to the preset debate scenario and the speech information of the debate scenario.

[0010] Optionally, the controlling the debate robot to output target debate language within the debate duration corresponding to the preset debate scenario according to the preset debate scenario and the speech information of the debate scenario includes:

[0011] Obtain the voice information of the debate scenario through the microphone interface connected to the debate robot;

[0012] Determine the voice information of the speaking debater corresponding to the microphone interface in the voice information of the debate scenario according to the number of the microphone interface;

[0013] Input the voice information of the speaking debater in the debate scenario into the speech recognition model to determine the speech text information of the debate scenario;

[0014] Input the preset debate scenario and the speech text information of the debate scenario into the language text generation model to determine the target debate language text;

[0015] Input the target debate language text into the speech broadcast model to determine the target debate language, and control the debate robot to output the target debate language.

[0016] Optionally, the inputting the target debate language text into the speech broadcast model to determine the target debate language and controlling the debate robot to output the target debate language includes:

[0017] Determine the target debate language with the intonation corresponding to the debater role according to the debater role of the debate robot, and control the robot to output the target debate language with the intonation corresponding to the debater role.

[0018] Optionally, the method further includes:

[0019] Determine the debate stance according to the text requirements of the debate topic, where the debate stance includes a positive stance and a negative stance;

[0020] Input the text requirements of the debate topic and the debate stance into the language text generation model to control the debate robot to understand the voice information of the debate scenario.

[0021] Optionally, the inputting the preset debate scenario and the speech text information of the debate scenario into the language text generation model to determine the target debate language text includes:

[0022] If the debate robot is in the argumentation session of the preset debate scenario, input the argumentation session into the language text generation model to determine the first target debate language text;

[0023] If the debate robot is in the confrontation session of the preset debate scenario, input the confrontation session and the speech text information of the debate scenario into the language text generation model to determine the second target debate language text;

[0024] If the debate robot is in the free debate session of a preset debate scenario, input the free debate session and the speech text information of the initial speaking debater in the free debate session into a language text generation model to determine a third target debate language text.

[0025] Optionally, the method further includes:

[0026] Controlling the debate robot to identify the session of the preset debate scenario by means of timing and / or triggering a preset keyword and / or user input.

[0027] Optionally, the method further includes:

[0028] The debate robot stores the speech information of the preset debate scenario.

[0029] According to a second aspect of the present disclosure, there is provided a control device for a debate robot based on speech processing, including:

[0030] An acquisition module, configured to acquire a debate material training set;

[0031] A training module, configured to input the debate material training set into a preset training model, and control the debate robot to learn debate materials. The preset training model includes a low-rank adapter, and the low-rank adapter is used to optimize the preset training model;

[0032] An input module, configured to input a preset debate scenario, the debate duration corresponding to the preset debate scenario, and the role played by the debate robot into the debate robot;

[0033] A control module, configured to control the debate robot to output a target debate language within the debate duration corresponding to the preset debate scenario according to the preset debate scenario and the speech information of the debate scenario.

[0034] According to a third aspect of the present disclosure, there is provided a non-transitory computer-readable storage medium, on which a computer program is stored, characterized in that when the program is executed by a processor, the steps of the method provided in the first aspect of the present disclosure are implemented

[0035] According to a fourth aspect of the present disclosure, there is provided a debate robot system based on speech processing, including:

[0036] A speech recognition module, the speech recognition module is connected to a microphone, and the speech recognition module is configured to recognize the speech information of each speaking debater and convert the speech information of each speaking debater into speech text information;

[0037] A language text generation module, which is connected to the speech recognition module and is used to understand the information of the debate scenario and generate a target debate language text;

[0038] A voice broadcast module, which is connected to the language text generation module and is used to convert the target debate language text into a target debate language.

[0039] Compared with the prior art, the embodiments of the present invention have at least one of the following beneficial effects:

[0040] Through the above technical solutions, by inputting the debate material training set into the preset training model and training the model through the low-rank adapter, the preset training robot obtains the knowledge of the debate materials, thereby controlling the debate robot to learn the debate materials, strengthening the vertical depth of the debate robot in the debate field, and adopting the low-rank adapter to reduce the number of training parameters and improve the training efficiency of the preset training model; inputting the preset debate scenario, the corresponding debate duration of the preset debate scenario, and the debater role of the debate robot into the debate robot, and controlling the debate robot to output the target debate language according to the preset debate scenario and the voice information of the debate scenario, generating a robot debater, and forming a real debate process to provide a sparring function for human debaters so as to improve the debate ability and debate skills of human debaters. Description of the Drawings

[0041] By reading the detailed description of the non-limiting embodiments with reference to the following drawings, other features, objects, and advantages of the present invention will become more obvious:

[0042] Figure 1 is a flowchart of a control method for a debate robot based on speech processing shown according to an exemplary embodiment.

[0043] Figure 2 is a flowchart of a method for controlling a debate robot to output a target debate language shown according to an exemplary embodiment.

[0044] Figure 3 is a block diagram of a control device for a debate robot based on speech processing shown according to an exemplary embodiment.

[0045] Figure 4 is a block diagram of a debate robot system based on speech processing shown according to an exemplary embodiment. Detailed Embodiments

[0046] The present invention will be described in detail below with reference to specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but do not limit the present invention in any form. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can be made. These all belong to the protection scope of the present invention.

[0047] Debate robots can be applied to debate competitions or debate training held in various industries, and can conduct interpersonal debates, act as sparring partners for human debaters, and debates between robots.

[0048] Figure 1 It is a flowchart of a control method for a debate robot based on speech processing shown according to an exemplary embodiment. As Figure 1 shown, a control method for a debate robot based on speech processing includes S11 to S14.

[0049] S11, obtain a debate material training set.

[0050] According to the debate scenario and the debate topic, the user collects professional information materials related to the debate scenario and the debate topic, and organizes these materials into a debate material training set.

[0051] S12, input the debate material training set into a preset training model, and control the debate robot to learn the debate materials. The preset training model includes a low-rank adapter, and the low-rank adapter is used to optimize the preset training model.

[0052] Among them, when optimizing the preset training model through the low-rank adapter, the SVD rank decomposition matrix of the matrix is used to reduce the number of parameters required for model training of the preset training model, and the original weight matrix of the preset training model is frozen, so that the preset training model can reduce the memory and computational amount required when training the model according to the debate material training set, and can reach 1 / 1000 - 1 / 10000 of the original method.

[0053] Exemplarily, the neural network of the preset training model includes multiple dense layers, the weight matrix in the dense layer is full rank, and the dense layer is used to perform matrix multiplication. When optimizing the preset training model through the low-rank adapter, the preset training model has a lower "dimension" and can be projected into a smaller subspace for effective learning.

[0054] Let the weight matrix of the preset training model be W0∈R d×k , and use low-rank decomposition W0+ΔW = W0+B A to represent its update, where, B∈R d×r , A∈R r×k , and the rank r <= min(d,k).

[0055] During the training process of the preset training model, W0 is frozen and no gradient update is performed. A and B contain trainable parameters. Both W0 and ΔW = BA perform multiplication operations with the same input, and then sum the respective output vectors according to the coordinates.

[0056] A is randomly initialized with Gaussian distribution, and B is initialized to zero. Then ΔW = BA is 0 at the beginning of training. During the adaptation process, there is no need to accumulate gradient updates for the weight matrix to obtain full rank. When the number of trainable parameters is increased, training LoRA converges to the original preset training model, converges to MLP based on the low-rank adapter, and converges to a model that cannot accept long input sequences based on the prefix method. During the training process, W = W0 + BA can be explicitly calculated and stored, and inference can be performed. Both W0 and BA are in R d×k so that there is no additional delay during the inference process.

[0057] In the above technical solution, the preset training model is optimized through a low-rank adapter. The preset training model after model training has professional knowledge of debate. The debate robot includes the preset training model, so that the debate robot can master professional knowledge of debate.

[0058] S13, input the preset debate scenario, the debate duration corresponding to the preset debate scenario, and the debater role of the debate robot into the debate robot.

[0059] Among them, the debate robot supports traditional debate scenarios and debate processes, and can also support personalized debate scenarios and debate processes.

[0060] Traditional debate scenarios include: the statement-making session, the cross-examination session, the summary session, the free debate session, and the conclusion session. Another traditional debate scenario includes: the statement-making session, the free debate session, the conclusion session, and the comment session.

[0061] In a possible embodiment, the debate duration corresponding to the preset debate scenario includes the total duration of the preset debate scenario and the duration of each session.

[0062] In the present disclosure, the debate robot can support the setting of no less than four debater roles and the management of no less than four debater roles.

[0063] S14, according to the preset debate scenario and the voice information of the debate scenario, control the debate robot to output target debate language within the debate duration corresponding to the preset debate scenario.

[0064] A language text generation model is internally set in the debate robot. Among them, the language text generation model can adopt models such as chatGPT model, Baidu's "ERNIE Bot" model, gpt4.0 model, and iFlytek model.

[0065] In another possible embodiment, the language text generation model can also adopt a natural language text generation model specifically trained for debates.

[0066] The language text generation model adopted by the debate robot can adopt one of the above models or a combination of multiple models.

[0067] Through the above technical solution, by inputting the debate material training set into the preset training model and training the model through the low-rank adapter, the preset training robot obtains the knowledge of the debate materials, thereby controlling the debate robot to learn the debate materials, strengthening the vertical depth of the debate robot in the field of debates, and adopting the low-rank adapter to reduce the number of training parameters and improve the training efficiency of the preset training model; inputting the preset debate scenario, the corresponding debate duration of the preset debate scenario, and the debater role of the debate robot into the debate robot, and controlling the debate robot to output the target debate language according to the preset debate scenario and the voice information of the debate scenario, generating a robot debater, and forming a real debate process, providing a sparring function for human debaters to improve the debate ability and skills of human debaters.

[0068] Figure 2 It is a flowchart of a method for controlling a debate robot to output a target debate language shown according to an exemplary embodiment. As Figure 2 shown, in some possible embodiments, controlling the debate robot to output the target debate language within the debate duration corresponding to the preset debate scenario according to the preset debate scenario and the voice information of the debate scenario includes S21 to S25.

[0069] S21, obtain the voice information of the debate scenario through the microphone interface connected to the debate robot.

[0070] S22, determine the voice information of the speaking debater corresponding to the microphone interface in the debate scenario according to the number of the microphone interface.

[0071] In one possible embodiment, connect the debate robot to the microphones of each human debater, number the microphone interfaces to locate each human debater, and obtain the voice information of each human debater through the microphone interfaces.

[0072] S23, input the voice information of the speaking debater in the debate scenario into the speech recognition model to determine the voice text information of the debate scenario.

[0073] The voice information of each human debater is combined with the microphone channel label (microphone interface number) and input into the speech recognition model of the debate robot system. When there are sound inputs from multiple microphone channels at the same time, the microphone channel where the voice information appears earliest is determined to determine the speaking debater, and the voice information of other microphone channels is shielded to avoid conflicts in the debaters' voice information and improve the speech recognition rate.

[0074] An intelligent voice input method is set in the voice recognition model, for example, Baidu input method, iFlytek input method, Huawei input method, and Microsoft input method. The debate robot's system supports various voice inputs and performs automatic voice recognition according to the API interface of the connected microphone to automatically recognize the voice information of each human debater obtained through the microphone interface and convert the voice information of each human debater into voice text information.

[0075] S24, inputting the preset debate scene and the voice text information of the debate scene into the language text generation model to determine the target debate language text.

[0076] Among them, the microphone channel label is marked on the voice text information to determine the speaker of the voice text information. The language text generation model also has a language text understanding function to understand the input voice text information and generate targeted target debate language text.

[0077] S25, inputting the target debate language text into the voice broadcast model, determining the target debate language, and controlling the debate robot to output the target debate language.

[0078] In some possible embodiments, according to the debater role of the debate robot, a target debate language with a tone corresponding to the debater role is determined, and the robot is controlled to output the target debate language with the tone corresponding to the debater role.

[0079] The intonation corresponding to the debater role includes the debater role's timbre and language style.

[0080] Among them, the voice broadcast model can adopt the TTS voice broadcast system, and the timbre and language style of each debater role of the debate robot can be set in the voice broadcast model to generate a target debate language with different intonations.

[0081] For example, the timbre of the robot debater character is set to a male timbre or a female timbre, and the voice styles of the first debater, the second debater, the third debater, and the fourth debater can be set to humorous, eloquent, passionate, friendly, straightforward, and the like.

[0082] Through the above settings, the debate robot outputs the target debate language through the timbre and language style of the set debate character.

[0083] In another possible embodiment, animations of syllables, words, and sentences can also be generated according to the target styles of English consonants and vowels.

[0084] Through the above technical solutions, the human debater making the speech can be located, the accuracy of speech recognition can be improved, and the tone color and language style of the debate robot can be set, which can enhance the debate effect of the debate robot.

[0085] In some possible embodiments, the method further includes S15 and S16.

[0086] S15. Determine the debate stance according to the text required by the debate topic. The debate stance includes a positive stance and a negative stance.

[0087] Among them, the positive stance means agreeing with the stance of the debate topic, and the negative stance means refuting the stance of the debate topic.

[0088] S16. Input the text required by the debate topic and the debate stance into the language text generation model to control the debate robot to understand the speech information of the debate scenario.

[0089] Set a Transformer natural language understanding model in the language text generation model. The Transformer natural language understanding model has an understanding function for the debate robot to understand the meaning expressed by the text required by the debate topic, the debate stance, and the meaning expressed by the speech information in the debate scenario.

[0090] If the debate stance of the debate robot is a positive stance, according to the preset debate scenario, the language text generation model understands the speech information of the debate scenario and generates a supportive target debate language text; if the debate stance of the debate robot is a negative stance, according to the preset debate scenario, the language text generation model understands the speech information of the debate scenario and generates a refuting target debate language text.

[0091] Through the above technical solutions, adding a Transformer natural language understanding model to the language text generation model controls the debate robot to have an understanding ability, improves the flexibility of the debate robot, and further enhances the experience of human debaters.

[0092] In some possible embodiments, inputting the preset debate scenario and the speech text information of the debate scenario into the language text generation model to determine the target debate language text includes:

[0093] If the debate robot is in the argumentation session of the preset debate scenario, input the argumentation session into the language text generation model to determine the first target debate language text.

[0094] Among them, the first target debate language text represents the language text of the statement of views. The argumentation session of the debate scenario is input into the debate robot, triggering the language text generation model of the debate robot to understand the debate topic and generate the first target debate language text to express the understanding of the views represented by the current debate topic.

[0095] If the debate robot is in the confrontation session of the preset debate scenario, the voice text information of the confrontation session and the debate scenario is input into the language text generation model to determine the second target debate language text.

[0096] Among them, the second target debate language text represents the different views expressed by each debater role represented by the debate robot according to the requirements of the debate topic. The confrontation session of the debate scenario is input into the debate robot, triggering the language text generation model of the debate robot to understand the debate topic and the voice text information of the debate scenario and generate the second target debate language text to express different views. By way of example, the debate robot can represent the first debater, the second debater, the third debater or the fourth debater, and the debate robot expresses different views according to the debater role.

[0097] If the debate robot is in the free debate session of the preset debate scenario, the voice text information of the free debate session and the initial speaking debater in the free debate session is input into the language text generation model to determine the third target debate language text.

[0098] Among them, the third target debate language text represents the view language text generated by the debate robot according to the voice text information of the opposing debater. The free debate session of the debate scenario is input into the debate robot, triggering the language text generation model of the debate robot to understand the voice text generation information of the opposing debater and generate the third target debate to express the view language text of the position represented by the debate robot.

[0099] In a possible embodiment, in the free debate session, according to the microphone channel label of the microphone and the interface number of the microphone, the voice text information of the initial speaking debater is determined to prevent input conflicts.

[0100] In some possible embodiments, the method further includes:

[0101] Controlling the debate robot to identify the session of the preset debate scenario by means of timing and / or triggering a preset keyword and / or user input.

[0102] As an example, the debate robot can be controlled to judge the session of the current debate scenario according to the timing duration of the debate competition.

[0103] As another example, keywords in the current debate process can also be captured, and based on whether the captured keywords trigger preset keywords, the debate robot is controlled to determine the stage of the current debate scenario.

[0104] As another example, the stage of the debate scenario can also be input into the debate robot through human-machine interaction by means of user input, so as to control the debate robot to determine the stage of the current debate scenario.

[0105] Exemplarily, inputting 1 represents the theoretical stage; inputting 2 represents the confrontation stage; inputting 3 represents the summary stage; inputting 4 represents the free debate stage; inputting 5 represents the conclusion stage.

[0106] In some possible embodiments, the refuting debate language can be determined in the following manner:

[0107] Input the preset debate scenario into the debate robot;

[0108] Use a language text generation model to identify the current debater role of the debate robot;

[0109] Determine the stage of the current debate scenario where the debate robot is located;

[0110] According to the preset debate scenario and the stage of the current debate scenario where the debate robot is located, control the language text model to generate targeted refuting debate text content, and output the refuting debate language according to the current debater role of the debate robot to achieve the debate.

[0111] In some possible embodiments, the method further includes:

[0112] The debate robot stores the voice information of the preset debate scenario.

[0113] Exemplarily, the debate robot can store the voice information output by the opposing debaters in a list, that is, store the voice information of the first debater, second debater, third debater, and fourth debater of the opposing side in the system of the debate robot in a list.

[0114] The debate robot can also store the voice information output by the debaters on the side where the debate robot is located in a list, that is, store the voice information of the first debater, second debater, third debater, and fourth debater on the side of the debate robot in the system of the debate robot in a list.

[0115] In a possible embodiment, the debate robot takes the stored voice information of the opposing debaters and the debaters on the side of the debate robot as input, and controls the language text generation model of the debate robot to accurately generate the target debate language text to achieve a good debate effect.

[0116] In some possible embodiments, the language text generation model may be controlled to generate the target debate language text at a preset output speed, so as to meet the requirement of the debate duration corresponding to the preset debate scenario, and determine the number of words to be generated according to the requirements of the debate scenario.

[0117] Exemplarily, the preset output speed may be 150 words per minute.

[0118] Through the above technical solutions, the adversarial debate between the debate robot and the human debater is realized, which promotes the improvement of the human debater's ability. Moreover, the debate robot can adapt to different debate scenarios, break through the time limit and accompany the human debater to conduct debate training. The debate robot can also be used for debate teaching, which has far-reaching significance for the field of debate.

[0119] Based on the same concept, the present disclosure also provides a control device for a debate robot based on speech processing. Figure 3 It is a block diagram of a control device for a debate robot based on speech processing shown according to an exemplary embodiment. As Figure 3 shown, the control device 100 for the debate robot based on speech processing includes an acquisition module 110, a training module 120, an input module 130, and a control module 140.

[0120] The acquisition module 110 is configured to acquire a debate material training set;

[0121] The training module 120 is configured to input the debate material training set into a preset training model, and control the debate robot to learn the debate materials. The preset training model includes a low-rank adapter, and the low-rank adapter is used to optimize the preset training model;

[0122] The input module 130 is configured to input a preset debate scenario, the debate duration corresponding to the preset debate scenario, and the role played by the debate robot into the debate robot;

[0123] The control module 140 is configured to control the debate robot to output the target debate language within the debate duration corresponding to the preset debate scenario according to the preset debate scenario and the speech information of the debate scenario.

[0124] Through the above technical solution, by inputting the debate material training set into a preset training model and training the model through a low-rank adapter, the preset training robot obtains debate material knowledge, thereby controlling the debate robot to learn the debate materials, strengthening the vertical depth of the debate robot in the field of debate, and adopting a low-rank adapter to reduce the number of training parameters and improve the training efficiency of the preset training model; inputting the preset debate scenario, the corresponding debate duration of the preset debate scenario, and the debater role of the debate robot into the debate robot, and controlling the debate robot to output the target debate language according to the preset debate scenario and the voice information of the debate scenario, generating a robot debater, and forming a real debate process, providing a sparring function for human debaters to improve the debate ability and debate skills of human debaters.

[0125] Optionally, the control module 140 includes:

[0126] An acquisition sub-module, configured to acquire the voice information of the debate scenario through a microphone interface connected to the debate robot;

[0127] A first determination sub-module, configured to determine the speaking debater corresponding to the microphone interface in the voice information of the debate scenario according to the number of the microphone interface;

[0128] A second determination sub-module, configured to input the voice information of the speaking debater in the debate scenario into a speech recognition model to determine the voice text information of the debate scenario;

[0129] A third determination sub-module, configured to input the preset debate scenario and the voice text information of the debate scenario into a language text generation model to determine the target debate language text;

[0130] An output sub-module, configured to input the target debate language text into a voice broadcast model to determine the target debate language, and control the debate robot to output the target debate language.

[0131] Optionally, the output sub-module includes:

[0132] A fourth determination sub-module, configured to determine the target debate language with the intonation corresponding to the debater role according to the debater role of the debate robot, and control the robot to output the target debate language with the intonation corresponding to the debater role.

[0133] Optionally, the device 100 further includes:

[0134] A first determination module, configured to determine a debate stance according to the text requirements of the debate topic, where the debate stance includes a positive stance and a negative stance;

[0135] An understanding module for inputting the text of the debate topic requirements and the input language text of the debate position into a language text generation model, and controlling the debate robot to understand the voice information of the debate scenario.

[0136] Optionally, the third determination sub-module includes:

[0137] A fifth determination sub-module for, if the debate robot is in the argumentation session of a preset debate scenario, inputting the argumentation session into the language text generation model to determine a first target debate language text;

[0138] A sixth determination sub-module for, if the debate robot is in the confrontation session of a preset debate scenario, inputting the confrontation session and the voice text information of the debate scenario into the language text generation model to determine a second target debate language text;

[0139] A seventh determination sub-module for, if the debate robot is in the free debate session of a preset debate scenario, inputting the free debate session and the voice text information of the initial speaking debater in the free debate session into the language text generation model to determine a third target debate language text.

[0140] Optionally, the device 100 further includes:

[0141] An identification module for controlling the debate robot to identify the session of a preset debate scenario through timing and / or triggering a preset keyword and / or user input.

[0142] Optionally, the device 100 further includes:

[0143] A storage module for the debate robot to store the voice information of the preset debate scenario.

[0144] Regarding the system in the above embodiments, the specific manners in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0145] In another exemplary embodiment, a non-transitory computer-readable storage medium including program instructions is further provided. When the program instructions are executed by a processor, the steps of the trachea segmentation method based on connectivity analysis in the first aspect above are implemented. For example, the computer-readable storage medium may be the memory including the program instructions above, and the above program instructions may be executed by the processor of the electronic device to complete the trachea segmentation method based on connectivity analysis.

[0146] In another exemplary embodiment, a computer program product is further provided. The computer program product includes a computer program executable by a programmable device, and the computer program has a code portion for performing the above-described trachea segmentation method based on connectivity analysis when executed by the programmable device.

[0147] Figure 4 is a block diagram of a speech processing-based debate robot system shown according to an exemplary embodiment. As Figure 4 shown, in some possible embodiments, a speech processing-based debate robot system is further provided, including:

[0148] A speech recognition module, the speech recognition module is connected to a microphone, and the speech recognition module is configured to recognize the speech information of each speaking debater and convert the speech information of each speaking debater into speech text information;

[0149] A language text generation module, the language text generation module is connected to the speech recognition module, and the language text generation module is configured to understand the information of the debate scenario and generate a target debate language text;

[0150] A speech broadcast module, the speech broadcast module is connected to the language text generation module, and the speech broadcast module is configured to convert the target debate language text into a target debate language.

[0151] The specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various deformations or modifications within the scope of the claims, which do not affect the essence of the present invention. The above preferred features can be combined arbitrarily without conflict.

Claims

1. A control method for a debate robot based on speech processing, characterized in that, Including: Obtain a debate material training set; Input the debate material training set into a preset training model, and control the debate robot to learn the debate materials. The preset training model includes a low-rank adapter, and the low-rank adapter is used to optimize the preset training model; Input a preset debate scenario, the debate duration corresponding to the preset debate scenario, and the debater role of the debate robot into the debate robot; According to the preset debate scenario and the speech information of the debate scenario, control the debate robot to output target debate language within the debate duration corresponding to the preset debate scenario; Among them, the controlling the debate robot to output target debate language within the debate duration corresponding to the preset debate scenario according to the preset debate scenario and the speech information of the debate scenario includes: Obtain the speech information of the debate scenario through a microphone interface connected to the debate robot; Determine the speech information of the speaking debater corresponding to the microphone interface in the debate scenario according to the number of the microphone interface; Input the speech information of the speaking debater in the debate scenario into a speech recognition model to determine the speech text information of the debate scenario; Input the preset debate scenario and the speech text information of the debate scenario into a language text generation model to determine the target debate language text; Input the target debate language text into a speech broadcast model to determine the target debate language, and control the debate robot to output the target debate language.

2. The control method of a debate robot based on speech processing according to claim 1, characterized in that The inputting the target debate language text into a speech broadcast model to determine the target debate language and controlling the debate robot to output the target debate language includes: Determine the target debate language with the intonation corresponding to the debater role according to the debater role of the debate robot, and control the robot to output the target debate language with the intonation corresponding to the debater role.

3. The control method of a debate robot based on speech processing according to claim 1, characterized in that, The method further includes: Determine a debate stance according to the debate topic requirement text, and the debate stance includes a positive stance and a negative stance; Input the debate topic requirement text and the debate stance into a language text generation model to control the debate robot to understand the speech information of the debate scenario.

4. The control method of a debate robot based on speech processing according to claim 1, characterized in that, The inputting the preset debate scenario and the speech text information of the debate scenario into a language text generation model to determine the target debate language text includes: If the debate robot is in the argumentation session of the preset debate scenario, input the argumentation session into the language text generation model to determine the first target debate language text; If the debate robot is in the confrontation session of the preset debate scenario, input the confrontation session and the speech text information of the debate scenario into the language text generation model to determine the second target debate language text; If the debate robot is in the free debate session of the preset debate scenario, input the free debate session and the speech text information of the initial speaking debater in the free debate session into the language text generation model to determine the third target debate language text.

5. The control method of a debate robot based on speech processing according to claim 1, characterized in that, The method further includes: Control the debate robot to identify the links of a preset debate scenario by means of timing and / or triggering preset keywords and / or user input.

6. The control method of a debate robot based on speech processing according to claim 1, wherein, The method further includes: The debate robot stores the voice information of the preset debate scenario.

7. A control device for a debate robot based on speech processing, characterized in that, It includes: An acquisition module for acquiring a debate material training set; A training module for inputting the debate material training set into a preset training model to control the debate robot to learn debate materials. The preset training model includes a low-rank adapter for optimizing the preset training model; An input module for inputting a preset debate scenario, the debate duration corresponding to the preset debate scenario, and the roles played by the debate robot into the debate robot; A control module for controlling the debate robot to output target debate language within the debate duration corresponding to the preset debate scenario according to the preset debate scenario and the voice information of the debate scenario; Among them, the control module includes: An acquisition sub-module for acquiring the voice information of the debate scenario through a microphone interface connected to the debate robot; A first determination sub-module for determining the speaker debater corresponding to the microphone interface in the voice information of the debate scenario according to the number of the microphone interface; A second determination sub-module for inputting the voice information of the speaker debater in the debate scenario into a speech recognition model to determine the speech text information of the debate scenario; A third determination sub-module for inputting the preset debate scenario and the speech text information of the debate scenario into a language text generation model to determine the target debate language text; An output sub-module for inputting the target debate language text into a speech broadcast model to determine the target debate language and controlling the debate robot to output the target debate language.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by a processor, it implements the steps of the control method of a speech processing-based debate robot according to any one of claims 1-6.

Images