Interactive dialogue control method and system for intelligent robot with body

Abstract

The invention discloses an interactive dialogue control method and system for an intelligent robot with a body, and relates to the field of control of the intelligent robot with the body, and the method comprises the steps: S1, building a dialogue model based on a rule; s2, performing fuzzy matching on the rules; s3, controlling an interactive dialogue process; and S4, dialogue strategies are selected, including an active dialogue strategy, a self-adaptive dialogue strategy and a multi-mode information dialogue strategy, and the dialogue strategies are selected based on rules of a dialogue model. According to the interactive dialogue control method and system for the intelligent robot with the body, the dialogue process and reply can be guided through the pre-defined rules, high interpretability and controllability are achieved, the dialogue process can be controlled, the complete dialogue process can be constructed by combining the rules, and the dialogue efficiency is improved. Based on the interactive dialogue content of the user, the rule matching after combination can be directly carried out, so that the rule meeting the condition can be directly found and the corresponding action can be executed.

Description

Technical Field

[0001] The present invention relates to the field of control of embodied intelligent robots, and in particular to an interactive dialogue control method and system for embodied intelligent robots. Background Art

[0002] Embodied intelligent robot is a new type of robot that integrates multidisciplinary knowledge such as artificial intelligence, robotics and biology. It has body perception, autonomous decision-making and action capabilities, can interact with the environment and complete various tasks.

[0003] When the embodied intelligent robot conducts interactive dialogue, the content units of its interactive dialogue will cause a sense of separation when they are in different application scenarios, which will affect the user's experience of interactive dialogue with the embodied intelligent robot. At the same time, when facing different users, the embodied intelligent robot cannot adopt different interactive dialogue methods according to the current user group, which will affect its practicality to a certain extent.

[0004] Therefore, it is necessary to propose an interactive dialogue control method and system for embodied intelligent robots to solve the above problems. Summary of the invention

[0005] The main purpose of the present invention is to provide an interactive dialogue control method and system for an embodied intelligent robot, which can effectively solve the problems in the background technology.

[0006] To achieve the above object, the technical solution adopted by the present invention is:

[0007] The interactive dialogue control method for an embodied intelligent robot comprises the following steps:

[0008] S1: Establish a rule-based dialogue model, analyze the application scenarios of the embodied intelligent robot, and sort out the needs and common problems faced by users based on the application scenarios, convert the needs and problems into rules, extract the converted rules, and finally store them in the rule library;

[0009] S2: Fuzzy matching of rules;

[0010] S3: Control of interactive dialogue flow;

[0011] S4: Selection of dialogue strategies, including active dialogue strategies, adaptive dialogue strategies and multimodal information dialogue strategies. The selection of dialogue strategies is based on the rules of the dialogue model.

[0012] Preferably, in step S1, the extraction of rules is performed automatically and manually, including collecting conversation data, analyzing the conversation data, and understanding users' common problems, needs and expressions based on text mining technology, wherein when the rules are extracted manually, domain experts in various application scenarios extract manual rules from the collected conversation data; when the rules are extracted automatically, the collected similar conversation data are automatically grouped based on a clustering algorithm, keywords in the group are proposed as rules, and they are sorted based on the grouped rules, including the following operation steps:

[0013] S101: Classify rules according to their functions, application scenarios, question types, and dialogue processes;

[0014] S102: Optimize and organize the extracted rules, check the syntax and logic of the rules; simplify and summarize the rules; test and verify the rules;

[0015] S103: Use visualization tools to display the rules and develop rule documents. The visualization tools include flowcharts, decision trees, tables, and lists.

[0016] Preferably, the expression methods are judged and classified, and the embodied intelligent robot interacts with the user based on the user's habit of conversational expression methods, wherein the expression methods are divided into direct statement, question, example, metaphor and analogy, and the expression rules are established based on this;

[0017] When the user's dialogue is expressed in direct terms, the embodied intelligent robot directly listens to the content of the user's dialogue and makes corresponding dialogue and action responses;

[0018] When the user's dialogue is expressed as asking questions, the embodied intelligent robot interacts with the dialogue based on the questions asked;

[0019] When the user's dialogue is expressed in the form of examples, the embodied intelligent robot learns the dialogue content in combination with the current application scenario, makes judgments based on the examples given by the user during the dialogue, and interacts with the user based on the judgment results;

[0020] When the user conversation is expressed in metaphors and analogies, the conversation content is learned in combination with the current application scenario, and keywords are extracted based on the metaphors and analogies used in the user conversation to understand the specific content of the user conversation and interact with the user based on this.

[0021] Preferably, in step S1, the application scenario is determined based on the functions equipped by the embodied intelligent robot, and the target is determined based on the above content. The targets include children, parents, men, women, couples, children, and the elderly. When conducting an interactive dialogue, the current application scenario is determined based on the dialogue data, and an interactive dialogue is conducted with the corresponding user through the determined application scenario. The embodied intelligent robot adopts different interactive dialogue control methods when it is in different application scenarios.

[0022] Preferably, in step S2, the fuzzy matching includes: performing fuzzy matching based on the edit distance, backing up the conversation data, editing the distance of the backed-up content, setting an edit distance threshold for the text data in a rule, and when the edit distance is less than the threshold, judging that the user's conversation input matches the rule, the embodied intelligent robot will conduct a corresponding interactive dialogue with the user, wherein the editing includes inserting, deleting and replacing text data, otherwise it does not match, and a new rule is generated and recorded;

[0023] The fuzzy matching based on edit distance is used to process spelling errors and grammatical errors in the conversation data;

[0024] Fuzzy matching also includes: fuzzy matching based on text data similarity calculation, measuring the similarity between the input of the user dialogue and the text data in the rule by cosine similarity, recording the input of the user dialogue and the content of the text data in the rule as strings, representing them as vectors, and calculating the cosine similarity between the two strings. When the similarity is close, it is determined that the user's dialogue input matches the rule, and the embodied intelligent robot will conduct a corresponding interactive dialogue with the user;

[0025] The fuzzy matching based on text data similarity calculation is used to process expressions in the conversation data that are semantically similar but not completely identical;

[0026] The specific formula for measuring cosine similarity is:

[0027]

[0028] Where A is the input vector of the user conversation, B is the vector of text data in the rule, A*B is the dot product of vector A and vector B, ||A|| and ||B|| represent the moduli of vector A and vector B respectively, cosθ is the calculated cosine similarity, and the value of cosine similarity is between -1-1. When the value is closer to 1, it means that vector A and vector B are more similar, and it is judged that the user's conversation input matches the rule. When the value is closer to -1, it means that vector A and vector B are less similar, and it is judged that the user's conversation input does not match the rule, and a new rule is generated and recorded.

[0029] Preferably, in step S3, the rules are combined to construct a complete interactive dialogue process. When the user has a dialogue with the embodied intelligent robot, the combination is matched based on the corresponding rules and the corresponding operations are performed. The interactive dialogue process is divided into linear and branch based on the application scenario of the embodied intelligent robot, wherein the linear interactive dialogue is carried out in accordance with the pre-combined rule sequence, and the branch interactive dialogue is carried out according to different conditions provided by the user during the interactive dialogue. Different rules are selected for the interactive dialogue.

[0030] Preferably, the active dialogue strategy is a strategy in which the embodied intelligent robot actively initiates dialogue and guides the user to conduct an interactive dialogue, establishes active dialogue rules, and the embodied intelligent robot actively provides relevant information to the user based on the user's behavior;

[0031] The judgment of dangerous and normal user behaviors includes the following judgment steps:

[0032] A: Data collection, which provides visual information data through video monitoring. The visual information data uses target detection algorithms to detect the user's position and posture, and extract the contours and joint positions of the human body;

[0033] Provide environmental information data through sensor data. Environmental information data uses environmental sensors to extract data from the current environment. Environmental sensors include temperature and humidity sensors and hazardous gas sensors.

[0034] Analyze whether there is dangerous behavior data through user conversation data, analyze the conversation data based on rules, and establish a dangerous conversation text database;

[0035] B: Based on the formula and the information data provided, the dangerous behavior and normal behavior are judged. The formula is as follows:

[0036] RI=WF*F+Wi*I+WE*E;

[0037] Where F is the behavior frequency index, I is the behavior intensity index, and E is the environmental factor index. Each index is assigned a weight W, and the sum of the weights is 1. A risk threshold is set based on the application scenario and rules. When RI is less than or equal to the threshold, it is judged as normal behavior. When RI is greater than the threshold, it is judged as dangerous behavior.

[0038] C: Label the information data provided, and clearly mark the information data as normal behavior or dangerous behavior. When the user is in dangerous behavior, the embodied intelligent robot will actively initiate a dialogue with the user.

[0039] Preferably, the adaptive dialogue strategy dynamically adjusts the dialogue strategy of the embodied intelligent robot according to user feedback and the progress of the dialogue.

[0040] Preferably, the multimodal information dialogue strategy is based on the embodied intelligent robot utilizing information in multiple modes, including but not limited to voice, image and gesture, to conduct corresponding interactive dialogue with the user based on the information.

[0041] An interactive dialogue control system for an embodied intelligent robot, and a method for implementing an interactive dialogue control method for an embodied intelligent robot, comprising:

[0042] A rule module, which is used to formulate rules for interactive dialogue control of the embodied intelligent robot, and to establish a dialogue model based on the rule module, to extract the user's dialogue data, and to judge the conditions and actions of the extracted dialogue data, where the conditions are used to judge whether the user's dialogue data meets the working scope of the embodied intelligent robot, and the actions are used for the reply actions made by the embodied intelligent robot when the working scope is met;

[0043] A rule base, which is used to store the rule content formulated for the embodied intelligent robot;

[0044] An application scenario analysis and arrangement module, which is used to analyze the application scenarios used by the embodied intelligent robot, and analyze and adjust the content and method of the interactive dialogue based on the application scenarios obtained after the analysis;

[0045] A fuzzy matching module, which is used to deal with spelling errors, grammatical errors and expression errors made by users during interactive dialogues, and to conduct interactive dialogues based on rules with high content matching similarity provided by users;

[0046] A dialogue strategy selection module, wherein the dialogue strategy selection module adopts different strategies to conduct interactive dialogue with the user based on the information data collected by the embodied intelligent robot, wherein the embodied intelligent robot includes at least one video monitor and an environmental sensor;

[0047] The dialogue strategy selection module also includes a behavior judgment unit, which judges the user's normal behavior and dangerous behavior based on the collected information data, and conducts corresponding interactive dialogue with the user based on the behavior, so as to reduce the user's dangerous behavior;

[0048] An interactive dialogue process control module is used to combine the stored rules, classify the combinations into appropriate application scenarios, and sort the order of the combinations. When conducting an interactive dialogue with a user, an interactive dialogue is conducted with the user based on the combination, application scenario and order.

[0049] Compared with the prior art, the present invention provides an interactive dialogue control method and system for an embodied intelligent robot, which has the following beneficial effects:

[0050] 1. The interactive dialogue control method and system for the embodied intelligent robot, by establishing a rule-based dialogue model, can make corresponding replies based on the rules when conducting interactive dialogues with users, and can extract text data through the rules, can mine and understand the user's problems and needs, and can judge the current application scenario and goal to conduct corresponding interactive dialogues with the user, and can select different interactive dialogue contents according to children, parents, men, women, couples, children, and the elderly, so that the scope of its application is relatively large, and it can be well applied to industrial manufacturing, logistics warehousing, medical care, service industry, public safety, education, and family life.

[0051] 2. The interactive dialogue control method and system for the embodied intelligent robot can determine the user's dialogue goal during the interactive dialogue, and guide the user to provide the required information based on the goal, so as to better complete the interactive dialogue with the user.

[0052] 3. The interactive dialogue control method and system for the embodied intelligent robot can also judge the user's dangerous behavior and normal behavior under normal circumstances, ensure the user's safety based on the collected information data, and mark the calculated behavior data. When it is marked as a dangerous behavior, it can actively communicate with the user to prevent the user from continuing to engage in dangerous behavior.

[0053] 4. The interactive dialogue control method and system for the embodied intelligent robot can guide the dialogue process and responses through pre-defined rules, has high explainability and controllability, and can control the dialogue process. By combining the rules, a complete dialogue process can be constructed. Based on the user's interactive dialogue content, the combined rules can be matched directly, so that the qualified rules can be directly found and the corresponding actions can be executed. When conducting an interactive dialogue, the user can clearly understand the decision-making process and response logic of the embodied intelligent robot, thereby ensuring the reliability and security of the interactive dialogue, and can adapt well to specific fields, including medical, legal and financial, and can ensure the professional knowledge of the embodied intelligent robot during interactive dialogue, and can provide accurate and professional responses.

[0054] 5. The interactive dialogue control method and system for the embodied intelligent robot can effectively improve the naturalness of the dialogue through fuzzy matching, and can allow the user to input interactive dialogues inaccurately to a certain extent, so that the embodied intelligent robot can better understand the user's intentions. At the same time, it can increase the robustness of the interactive dialogue, and can help the embodied intelligent robot give corresponding answers to spelling errors and incomplete expressions in the user's interactive dialogue. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 is a flow chart of the overall method of the present invention;

[0056] Figure 2 is the judgment flowchart of the behavior of the present invention;

[0057] Figure 3 is the system block diagram of the present invention. Detailed implementation manners

[0058] In order to make the technical means, creative features, achieved purposes and functions of the present invention easy to understand, the present invention will be further described below in conjunction with specific implementation manners.

[0059] Embodiment 1:

[0060] As Figure 1 , Figure 2 shown, an interactive dialogue control method for an embodied intelligent robot, including the following operating steps:

[0061] S1: Establish a rule-based dialogue model, analyze the application scenarios of the embodied intelligent robot, sort out the needs and common problems faced by users based on the application scenarios, convert the needs and problems into rules, then extract the converted rules, and finally store them in the rule library;

[0062] The extraction of rules is carried out automatically and manually, including collecting dialogue data, analyzing the dialogue data, and understanding the common problems, needs and expression methods of users based on text mining technology, where the text mining technology includes:

[0063] Convert the dialogue data into text data, extract the problems and needs therein, and select relevant data sources according to specific problems and needs to ensure that the collected text data is targeted and effective. The relevant data sources are collected through previous user feedback, social media, online forums and questionnaires, and web crawlers, API interfaces, and data scraping tools can also be used to collect text data from the selected data sources to ensure that the collected data has a standardized format and complete content;

[0064] Text preprocessing, cleaning the collected text data, removing the noise in the text data, including HTML tags, special characters, punctuation marks, and correcting spelling mistakes and grammar mistakes in the text;

[0065] Word segmentation processing, splitting the text data into individual independent words for subsequent analysis and processing;

[0066] Stop word filtering, removing stop words in the text, such as common meaningless words like "de", "shi", "zai".

[0067] Feature extraction, which counts the frequency of occurrence of each word in the text data to understand the theme and keywords of the text;

[0068] Use latent Dirichlet allocation techniques for subject modeling to extract latent themes and topics from text data;

[0069] Sentiment analysis is performed on text data to understand the emotional tendencies expressed in the text, such as positive, negative, or neutral.

[0070] Problem and demand analysis: classify problems in text data based on extracted features, and use machine learning algorithms to classify and predict problems in text data;

[0071] Demand mining, mining user needs and expectations from text data based on the above content.

[0072] The embodied intelligent robot can judge and classify the expression methods based on the user's conversational expression habits. The expression methods are divided into direct statement, question, example, metaphor and analogy, and the rules of expression methods are established based on this.

[0073] When the user's dialogue is expressed in direct terms, the embodied intelligent robot directly listens to the content of the user's dialogue and makes corresponding dialogue and action responses;

[0074] When the user's dialogue is expressed as asking questions, the embodied intelligent robot interacts with the dialogue based on the questions asked;

[0075] When the user dialogue is expressed in the form of examples, the embodied intelligent robot learns the dialogue content in combination with the current application scenario, and makes judgments based on the examples given by the user during the dialogue, interacts with the user according to the judgment results, and when the example given is judged to be wrong, the embodied intelligent robot interacts with the correct example, which can be used to teach the user through the embodied intelligent robot;

[0076] When the user's conversation is expressed in metaphors and analogies, the conversation content is learned in combination with the current application scenario, and keywords are extracted based on the metaphors and analogies used in the user's conversation to understand the specific content of the user's conversation and interact with the user based on this;

[0077] When manually extracting rules, domain experts in various application scenarios manually extract rules from the collected conversation data; when automatically extracting rules, similar conversation data collected are automatically grouped based on a clustering algorithm, keywords in the groups are proposed as rules, and they are sorted based on the grouped rules, including the following steps:

[0078] S101: Classify the rules according to their functions, application scenarios, question types, and dialogue processes. When one of the classified contents is in the same field as another classified content, they are integrated and classified into the same group;

[0079] S102: Optimize and organize the extracted rules, check the syntax and logic of the rules; simplify and summarize the rules; test and verify the rules;

[0080] Remove duplicate rules. Check whether there are duplicate rules in the rules. If there are, keep one of them and delete the other duplicate rules.

[0081] Simplify rules. For complex rules, simplify them, merge multiple similar conditions into a simpler condition, and merge multiple actions into a more general action;

[0082] Optimize the order of rules: Optimize the order of rules according to their importance and frequency of use, and put commonly used rules in front, so that the embodied intelligent robot can respond faster to the content of the rules in front;

[0083] S103: Use visualization tools to display the rules and develop rule documents. The visualization tools include flowcharts, decision trees, tables, and lists.

[0084] Create documents for individual rules, record the purpose of creation, applicable scenarios and specific content of the rules, use clear language to describe the rules in the documents, and avoid using vague and ambiguous expressions.

[0085] The application scenarios are determined based on the functions equipped by the embodied intelligent robot. The application scenarios include industrial manufacturing, logistics warehousing, medical care, service industry, public safety, education, and family life. The targets are determined based on the above content. The targets include children, parents, men, women, couples, children, the elderly, and patients. When conducting interactive dialogues, the current application scenario is determined based on the dialogue data, and interactive dialogues are conducted with the corresponding users through the determined application scenarios. The embodied intelligent robot adopts different interactive dialogue control methods when it is in different application scenarios.

[0086] By establishing a rule-based dialogue model, when interacting with users, corresponding replies can be made based on the rules, and text data can be extracted through the rules. It can explore and understand the user's problems and needs. At the same time, it can judge the current application scenario and goal and have corresponding interactive dialogues with the user. Different interactive dialogue content can be selected according to children, parents, men, women, couples, children, and the elderly, which makes its application scope larger and can be well applied to industrial manufacturing, logistics warehousing, medical care, service industry, public safety, education, and family life.

[0087] S2: Perform fuzzy matching on the rules. The fuzzy matching includes: performing fuzzy matching based on the edit distance, backing up the conversation data, editing the distance of the backed-up content, setting an edit distance threshold for the text data in the rule, and when the edit distance is less than the threshold, it is determined that the user's conversation input matches the rule. The embodied intelligent robot will conduct a corresponding interactive dialogue with the user, where editing includes inserting, deleting and replacing text data. Otherwise, it does not match, and a new rule is generated and recorded;

[0088] Fuzzy matching based on edit distance is used to handle spelling errors and grammatical errors in conversation data;

[0089] Fuzzy matching also includes: fuzzy matching based on text data similarity calculation, measuring the similarity between the input of the user dialogue and the text data in the rule by cosine similarity, recording the input of the user dialogue and the content of the text data in the rule as strings, representing them as vectors, and calculating the cosine similarity between the two strings. When the similarity is close, it is determined that the user's dialogue input matches the rule, and the embodied intelligent robot will conduct a corresponding interactive dialogue with the user;

[0090] Fuzzy matching based on text data similarity calculation is used to process semantically similar but not completely identical expressions in conversation data;

[0091] The specific formula for measuring cosine similarity is:

[0092]

[0093] Where A is the input vector of the user conversation, B is the vector of text data in the rule, A*B is the dot product of vector A and vector B, ||A|| and ||B|| represent the moduli of vector A and vector B respectively, cosθ is the calculated cosine similarity, and the value of cosine similarity is between -1-1. When the value is closer to 1, it means that vector A and vector B are more similar, and it is judged that the user's conversation input matches the rule. When the value is closer to -1, it means that vector A and vector B are less similar, and it is judged that the user's conversation input does not match the rule, and a new rule is generated and recorded.

[0094] Fuzzy matching can effectively improve the naturalness of conversations, allowing users to input inaccurate interactive conversations to a certain extent, so that the embodied intelligent robot can better understand the user's intentions. At the same time, it can increase the robustness of interactive conversations. When faced with spelling errors and incomplete expressions in user interactive conversations, the embodied intelligent robot can give corresponding answers.

[0095] S3: Control of the interactive dialogue process. The rules are combined to construct a complete interactive dialogue process. When the user has a dialogue with the embodied intelligent robot, the combination is matched based on the corresponding rules and the corresponding operations are performed. The interactive dialogue process is divided into linear and branch based on the application scenario of the embodied intelligent robot. The linear interactive dialogue is carried out in the order of pre-combined rules, and the branch interactive dialogue is carried out according to different conditions provided by the user during the interactive dialogue.

[0096] Before controlling the interactive dialogue process, the user's dialogue goals are determined. The dialogue goals include providing information, solving problems, completing tasks, and social activities. Based on this, the topic, scope, and process of the interactive dialogue are determined;

[0097] Explain the determined dialogue goals to the user, let the user understand the purpose and process of the dialogue, and ask guiding questions to guide the user into the dialogue topic. Based on the user's interactive dialogue content, provide multiple rules, and proceed with the subsequent process after the user confirms.

[0098] Design appropriate questions based on the goals and rules of the conversation, gradually guide users to provide the required information, use open-ended questions, closed-ended questions, and multiple-choice questions to obtain different types of information, and provide feedback to users to confirm whether their understanding is correct, and ask follow-up questions based on their needs;

[0099] It also includes the management of dialogue status, which locates the status of the current interactive dialogue, including the initial status, question answering status and task execution status. Users can switch states freely and make corresponding decisions based on the status. When in the question answering status, they can choose the appropriate answer to reply according to the user's question. When in the task execution status, they can perform corresponding interactive dialogue operations according to the progress of the task.

[0100] Predefined rules can be used to guide the flow and responses of conversations. It has high explainability and controllability, and can control the flow of conversations. By combining rules, a complete conversation flow can be constructed. Based on the content of the user's interactive conversation, the combined rules can be matched directly to directly find qualified rules and perform corresponding actions. When conducting interactive conversations, users can clearly understand the decision-making process and response logic of the embodied intelligent robot, thereby ensuring the reliability and security of interactive conversations. It can also adapt well to specific fields, including medical, legal and financial, and can ensure the professional knowledge of the embodied intelligent robot during interactive conversations, and can provide accurate and professional responses.

[0101] S4: Selection of dialogue strategies, including active dialogue strategies, adaptive dialogue strategies and multi-modal information dialogue strategies. The selection of dialogue strategies is based on the rules of the dialogue model.

[0102] The active dialogue strategy is a strategy for the embodied intelligent robot to actively initiate dialogue and guide users to conduct interactive dialogues. The active dialogue rules are established, and the embodied intelligent robot actively provides relevant information to users based on the user's behavior;

[0103] The judgment of dangerous and normal user behaviors includes the following judgment steps:

[0104] A: Data collection, which provides visual information data through video monitoring. The visual information data uses target detection algorithms to detect the user's position and posture, and extract the contours and joint positions of the human body;

[0105] Provide environmental information data through sensor data. Environmental information data uses environmental sensors to extract data from the current environment. Environmental sensors include temperature and humidity sensors and hazardous gas sensors.

[0106] Analyze whether there is dangerous behavior data through user conversation data, analyze the conversation data based on rules, and establish a dangerous conversation text database;

[0107] B: Based on the formula and the information data provided, the dangerous behavior and normal behavior are judged. The formula is as follows:

[0108] RI=WF*F+Wi*I+WE*E;

[0109] Where F is the behavior frequency index, I is the behavior intensity index, and E is the environmental factor index. Each index is assigned a weight W, and the sum of the weights is 1. A risk threshold is set based on the application scenario and rules. When RI is less than or equal to the threshold, it is judged as normal behavior. When RI is greater than the threshold, it is judged as dangerous behavior.

[0110] C: Label the provided information data and clearly mark the information data as normal behavior or dangerous behavior. When the user is in dangerous behavior, the embodied intelligent robot will actively initiate a dialogue with the user.

[0111] The adaptive dialogue strategy dynamically adjusts the dialogue strategy of the embodied intelligent robot according to the user's feedback and the progress of the dialogue. When the user is not satisfied with the answer of the embodied intelligent robot, the embodied intelligent robot adjusts the content of the subsequent interactive dialogue according to the corresponding content of the user's feedback to meet the user's needs.

[0112] The multimodal information dialogue strategy is based on the embodied intelligent robot using information in multiple modes, including but not limited to voice, images and gestures, to conduct corresponding interactive dialogues with users based on the information.

[0113] Embodiment 2:

[0114] like Figure 3 As shown, the interactive dialogue control system for an embodied intelligent robot includes:

[0115] The rule module is used to formulate rules for the interactive dialogue control of the embodied intelligent robot, and to establish a dialogue model based on the rule module, to extract the user's dialogue data, and to judge the conditions and actions of the extracted dialogue data. The conditions are used to judge whether the user's dialogue data meets the working scope of the embodied intelligent robot, and the actions are used for the reply actions made by the embodied intelligent robot when the working scope is met;

[0116] The rule module includes a rule self-setting unit, a reward unit, and a priority setting unit;

[0117] In the rule self-making unit, users make rules for the embodied intelligent robot according to their own needs. After making the rules, the embodied intelligent robot compares and judges the rules based on its own functions. If the rules are judged to be in line with the functions, the rule making is completed. If it is judged to be not in line with the functions, the embodied intelligent robot interacts with the user and guides the user to make rules that meet the functions.

[0118] Reward unit: Based on the guidance of the embodied intelligent robot and the interactive dialogue content with the user, when the user performs specific behaviors, the embodied intelligent robot conducts a praise-type interactive dialogue to improve the user's experience and enthusiasm for using the embodied intelligent robot. The specific behaviors include but are not limited to when the embodied intelligent robot guides the user to make a corresponding answer, the embodied intelligent robot teaches a child user, and the user performs certain sports movements;

[0119] The priority setting unit sets the priority of rules based on the content, application scenarios and user goals of the rules. When multiple rules are applicable at the same time, the embodied intelligent robot automatically determines the priority of the rules, where the content, application scenarios and user goals of the rules can be combined with each other. For example, based on a single application scenario, the priority is determined according to the content of the rules and the user goals.

[0120] The rule base is used to store the rules formulated for the embodied intelligent robot.

[0121] Application scenario analysis and arrangement module: The application scenario analysis and arrangement module is used to analyze the application scenarios used by the embodied intelligent robot, and analyze and adjust the content and method of the interactive dialogue based on the application scenarios obtained after analysis;

[0122] The application scenario analysis and arrangement module includes demand analysis unit, behavior analysis unit, and arrangement unit;

[0123] The demand analysis unit mines the user's needs based on the current application scenario obtained through analysis. The user's needs include but are not limited to search, answering questions, and communication and chatting. The subsequent interactive dialogue rules are predicted and sorted for the analyzed current user needs. When the rule is food search, the user's next interactive dialogue content is predicted, and food information, food location, and food recommendations provided by the network are provided to the user, so as to improve the fluency of the user's dialogue with the embodied intelligent robot.

[0124] The behavior analysis unit analyzes the user's interactive dialogue behavior according to the rules set by the user and the content of the interactive dialogue with the embodied intelligent robot, cleans the raw data in the interactive dialogue behavior, removes noise data and erroneous data, and standardizes the data to make data from different sources comparable. Based on the application scenario, different behavior patterns are identified from the processed data, including information sharing behavior patterns, social interaction behavior patterns, and transportation travel behavior sharing patterns. The identified behavior patterns are converted into quantifiable feature forms, and rules similar to user behaviors are selected based on the feature forms.

[0125] The sorting unit is used to integrate the information obtained by the demand analysis unit and the behavior analysis unit, merge the user's needs and the user's habitual behaviors, and select the most appropriate rules to interact with the user based on the integrated information.

[0126] Fuzzy matching module: The fuzzy matching module is used to deal with spelling errors, grammatical errors and expression errors made by users during interactive dialogues, and conduct interactive dialogues based on rules with high similarity matching of content provided by users;

[0127] The fuzzy matching module includes a data preprocessing unit, a matching rule setting unit, and a result output and sorting unit;

[0128] The data preprocessing unit is used to clean the data collected by the embodied intelligent robot and perform standardization processing on the data, and can convert the data into text data and audio data in a unified manner, wherein the formats of the text data and the audio data are the same;

[0129] The matching rule setting unit sets matching rules according to specific application scenarios, and is used to set the looseness of key rule data matching, including allowing partial keyword matching, synonym matching, and similar behavior matching;

[0130] The result output and sorting unit can output the matching results after completing the matching based on the algorithm, sort the results according to the similarity of the matching results, put the most similar results in the front, and interact with the current user based on the results in the front. If the user is not satisfied with the interactive dialogue, the subsequent results will be selected to continue the interactive dialogue.

[0131] A dialogue strategy selection module, wherein the dialogue strategy selection module adopts different strategies to conduct interactive dialogues with users based on information data collected by the embodied intelligent robot, and the embodied intelligent robot includes at least one video monitor and an environmental sensor;

[0132] The dialogue strategy selection module also includes a behavior judgment unit, which judges the user's normal behavior and dangerous behavior based on the collected information data, and conducts corresponding interactive dialogues with the user based on the behavior, so as to reduce the user's dangerous behavior;

[0133] The interactive dialogue process control module is used to combine the stored rules, classify the combinations into appropriate application scenarios, and sort the order of the combinations. When interacting with the user, the interactive dialogue is conducted based on the combination, application scenario and order.

[0134] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. An interactive dialogue control method for an embodied intelligent robot, characterized in that: The steps include: S1: Establish a rule-based dialogue model, analyze the application scenarios of the embodied intelligent robot, and sort out the needs and common problems faced by users based on the application scenarios, convert the needs and problems into rules, extract the converted rules, and finally store them in the rule library; S2: Fuzzy matching of rules; S3: Control of interactive dialogue flow; S4: Selection of dialogue strategies, including active dialogue strategies, adaptive dialogue strategies and multimodal information dialogue strategies. The selection of dialogue strategies is based on the rules of the dialogue model.

2. The interactive dialogue control method for an embodied intelligent robot according to claim 1, characterized in that: In step S1, the extraction of rules is performed automatically and manually, including collecting conversation data, analyzing the conversation data, and understanding the common problems, needs and expressions of users based on text mining technology. When the rules are extracted manually, experts in the fields of various application scenarios extract manual rules from the collected conversation data; when the rules are extracted automatically, the collected similar conversation data are automatically grouped based on a clustering algorithm, and the keywords in the group are proposed as rules, and they are sorted based on the grouped rules, including the following operation steps: S101: Classify rules according to their functions, application scenarios, question types, and dialogue processes; S102: Optimize and organize the extracted rules, check the syntax and logic of the rules; simplify and summarize the rules; test and verify the rules; S103: Use visualization tools to display the rules and develop rule documents. The visualization tools include flowcharts, decision trees, tables, and lists.

3. The interactive dialogue control method for an embodied intelligent robot according to claim 2, characterized in that: The expression methods are judged and classified. The embodied intelligent robot interacts with the user based on the user's habit of conversational expression methods. The expression methods are divided into direct statement, question, example, metaphor and analogy, and the expression rules are established based on this; When the user's dialogue is expressed in direct terms, the embodied intelligent robot directly listens to the content of the user's dialogue and makes corresponding dialogue and action responses; When the user's dialogue is expressed as asking questions, the embodied intelligent robot interacts with the dialogue based on the questions asked; When the user's dialogue is expressed in the form of examples, the embodied intelligent robot learns the dialogue content in combination with the current application scenario, makes judgments based on the examples given by the user during the dialogue, and interacts with the user based on the judgment results; When the user conversation is expressed in metaphors and analogies, the conversation content is learned in combination with the current application scenario, and keywords are extracted based on the metaphors and analogies used in the user conversation to understand the specific content of the user conversation and interact with the user based on this.

4. The interactive dialogue control method for an embodied intelligent robot according to claim 1, characterized in that: In the step S1, the application scenario is determined based on the functions equipped by the embodied intelligent robot, and the target is determined based on the above content. The targets include children, parents, men, women, couples, children, and the elderly. When conducting interactive dialogues, the current application scenario is determined based on the dialogue data, and an interactive dialogue is conducted with the corresponding user through the determined application scenario. The embodied intelligent robot adopts different interactive dialogue control methods when it is in different application scenarios.

5. The interactive dialogue control method for an embodied intelligent robot according to claim 1, characterized in that: In step S2, the fuzzy matching includes: performing fuzzy matching based on the edit distance, backing up the conversation data, editing the distance of the backed-up content, setting an edit distance threshold for the text data in a rule, and when the edit distance is less than the threshold, it is determined that the user's conversation input matches the rule, and the embodied intelligent robot will conduct a corresponding interactive dialogue with the user, wherein the editing includes inserting, deleting and replacing text data, otherwise it does not match, and a new rule is generated and recorded; The fuzzy matching based on edit distance is used to process spelling errors and grammatical errors in the conversation data; Fuzzy matching also includes: fuzzy matching based on text data similarity calculation, measuring the similarity between the input of the user dialogue and the text data in the rule by cosine similarity, recording the input of the user dialogue and the content of the text data in the rule as strings, representing them as vectors, and calculating the cosine similarity between the two strings. When the similarity is close, it is determined that the user's dialogue input matches the rule, and the embodied intelligent robot will conduct a corresponding interactive dialogue with the user; The fuzzy matching based on text data similarity calculation is used to process expressions in the conversation data that are semantically similar but not completely identical; The specific formula for measuring cosine similarity is: Where A is the input vector of the user conversation, B is the vector of text data in the rule, A*B is the dot product of vector A and vector B, ||A|| and ||B|| represent the moduli of vector A and vector B respectively, cosθ is the calculated cosine similarity, and the value of cosine similarity is between -1-1. When the value is closer to 1, it means that vector A and vector B are more similar, and it is judged that the user's conversation input matches the rule. When the value is closer to -1, it means that vector A and vector B are less similar, and it is judged that the user's conversation input does not match the rule, and a new rule is generated and recorded.

6. The interactive dialogue control method for an embodied intelligent robot according to claim 1, characterized in that: In step S3, the rules are combined to construct a complete interactive dialogue process. When the user has a dialogue with the embodied intelligent robot, the combination is matched based on the corresponding rules and the corresponding operations are performed. The interactive dialogue process is divided into linear and branch based on the application scenario of the embodied intelligent robot. The linear interactive dialogue is carried out in accordance with the pre-combined rule sequence, and the branch interactive dialogue is carried out according to different conditions provided by the user during the interactive dialogue.

7. The interactive dialogue control method for an embodied intelligent robot according to claim 1, characterized in that: The active dialogue strategy is a strategy for the embodied intelligent robot to actively initiate dialogue and guide users to conduct interactive dialogues. Active dialogue rules are established, and the embodied intelligent robot actively provides relevant information to users based on the user's behavior; The judgment of dangerous and normal user behaviors includes the following judgment steps: A: Data collection, which provides visual information data through video monitoring. The visual information data uses target detection algorithms to detect the user's position and posture, and extract the contours and joint positions of the human body; Provide environmental information data through sensor data. Environmental information data uses environmental sensors to extract data from the current environment. Environmental sensors include temperature and humidity sensors and hazardous gas sensors. Analyze whether there is dangerous behavior data through user conversation data, analyze the conversation data based on rules, and establish a dangerous conversation text database; B: Based on the formula and the information data provided, the dangerous behavior and normal behavior are judged. The formula is as follows: RI=WF*F+Wi*I+WE*E; Where F is the behavior frequency index, I is the behavior intensity index, and E is the environmental factor index. Each index is assigned a weight W, and the sum of the weights is 1. A risk threshold is set based on the application scenario and rules. When RI is less than or equal to the threshold, it is judged as normal behavior. When RI is greater than the threshold, it is judged as dangerous behavior. C: Label the information data provided, and clearly mark the information data as normal behavior or dangerous behavior. When the user is in dangerous behavior, the embodied intelligent robot will actively initiate a dialogue with the user.

8. The interactive dialogue control method for an embodied intelligent robot according to claim 1, characterized in that: The adaptive dialogue strategy dynamically adjusts the dialogue strategy of the embodied intelligent robot according to user feedback and the progress of the dialogue.

9. The interactive dialogue control method for an embodied intelligent robot according to claim 1, characterized in that: The multimodal information dialogue strategy is based on the embodied intelligent robot using information in multiple modes, including but not limited to voice, image and gesture, to conduct corresponding interactive dialogue with the user based on the information.

10. An interactive dialogue control system for an embodied intelligent robot, used to implement the interactive dialogue control method for an embodied intelligent robot as claimed in any one of claims 1 to 9, characterized in that: include: A rule module, which is used to formulate rules for interactive dialogue control of the embodied intelligent robot, and to establish a dialogue model based on the rule module, to extract the user's dialogue data, and to judge the conditions and actions of the extracted dialogue data, where the conditions are used to judge whether the user's dialogue data meets the working scope of the embodied intelligent robot, and the actions are used for the reply actions made by the embodied intelligent robot when the working scope is met; A rule base, which is used to store the rule content formulated for the embodied intelligent robot; An application scenario analysis and arrangement module, which is used to analyze the application scenarios used by the embodied intelligent robot, and analyze and adjust the content and method of the interactive dialogue based on the application scenarios obtained after the analysis; A fuzzy matching module, which is used to deal with spelling errors, grammatical errors and expression errors made by users during interactive dialogues, and to conduct interactive dialogues based on rules with high content matching similarity provided by users; A dialogue strategy selection module, wherein the dialogue strategy selection module adopts different strategies to conduct interactive dialogue with the user based on the information data collected by the embodied intelligent robot, wherein the embodied intelligent robot includes at least one video monitor and an environmental sensor; The dialogue strategy selection module also includes a behavior judgment unit, which judges the user's normal behavior and dangerous behavior based on the collected information data, and conducts corresponding interactive dialogue with the user based on the behavior, so as to reduce the user's dangerous behavior; An interactive dialogue process control module is used to combine the stored rules, classify the combinations into appropriate application scenarios, and sort the order of the combinations. When conducting an interactive dialogue with a user, an interactive dialogue is conducted with the user based on the combination, application scenario and order.

Images