Man-machine system optimization interaction method based on feedback enthusiasm intelligent evaluation

Abstract

The invention relates to the technical field of man-machine interaction methods and systems (G06F3), in particular to a man-machine system optimization interaction method based on feedback enthusiasm intelligent evaluation, which comprises the following steps: S1, collecting user interaction data; s2, performing quantitative evaluation on the emotion of the user; s3, classifying, identifying and identifying the user; s4, based on the feedback enthusiasm model, performing intelligent evaluation on interaction feedback of different types of users; s5, generating a user emotion evaluation matrix, matching the user emotion evaluation matrix with a historical matrix, and optimizing an interaction strategy; s6, calculating the reference degree of the interactive content through matching analysis; s7, guiding the user to complete the target task in the interaction process; and S8, iteratively optimizing the system performance based on the user feedback data. According to the method, the interaction efficiency of the system and the user experience are improved through sentiment intensity labeling and positive and negative sentiment proportion calculation, the interaction performance of the system is dynamically optimized through multiple interaction feedback weighted analysis, and the method is suitable for a man-machine intelligent interaction system under multiple scenes and has a good application prospect.

Description

Technical Field

[0001] The present invention relates to the technical field of human-computer interaction methods and systems (G06F3), and particularly relates to an optimized interaction method for a human-machine system based on intelligent evaluation of feedback positivity. Background Art

[0002] With the rapid development of artificial intelligence and machine learning technologies, human-computer interaction (HCI) has become one of the core elements in the design of intelligent devices and systems. In traditional human-computer interaction, the interaction between users and systems is usually passive, and the system responds based on the user's input. However, this interaction method often has certain limitations in the face of complex application scenarios, especially in terms of personalized needs and adaptive adjustments. The system's response often fails to meet the diverse and dynamic needs of users.

[0003] To overcome this problem, intelligent systems based on feedback have gradually become a research hotspot. Feedback positivity, that is, the system actively adjusts its interaction strategy and optimizes its functions according to user feedback, has been proven to significantly improve the adaptability of the system and the user's sense of participation. However, how to effectively measure the feedback positivity of users and perform intelligent optimization based on this is still a key challenge in the current field of human-computer interaction.

[0004] Currently, the feedback mechanisms of most human-machine systems still rely on simplified user inputs, such as button clicks, text inputs, etc. This feedback mechanism lacks in-depth analysis of user emotions and behaviors, resulting in the system being unable to dynamically adjust according to the real needs of users, thereby affecting the user experience. The intelligent evaluation method based on feedback positivity attempts to solve this problem by monitoring the user's behavior patterns, emotional feedback, and interaction frequencies, and intelligently adjusting the system's response strategy to achieve a personalized and optimized interaction experience.

[0005] Therefore, an optimized interaction method for a human-machine system based on intelligent evaluation of feedback positivity is proposed, which can better adapt to the personalized needs of users while maintaining the coherence and fluency of the user experience, improve the intelligence level of the system, and provide a more accurate and efficient interaction mechanism for future intelligent devices and services.

[0006] Related technologies are disclosed in the prior art:

[0007] However, the following problems still exist in the above prior art:

[0008] (1) Image recognition based on algorithms such as deep learning requires high hardware computing power; and under low visibility weather conditions, the learning parameters of deep learning are easily affected by factors such as rain and fog diffraction light, resulting in the inaccuracy of the learned model and the reduction of recognition accuracy.

[0009] (2) Based on traditional image recognition technology, under low visibility conditions, if the overall visibility of the image decreases, the recognition ability of the edges of components will be reduced.

[0010] (3) Traditional recognition technology discloses a technical means of using gray gradient for recognition. However, it only recognizes according to the maximum direction of gray gradient change. If some areas are affected by visibility, clouds, fog, light and shadow, the recognition range cannot be expanded and the recognition accuracy is relatively low. Summary of the Invention

[0011] To overcome the shortcomings and deficiencies of the prior art, the present invention provides an optimized interaction method for a human-machine system based on intelligent evaluation of feedback enthusiasm, which specifically involves the following aspects:

[0012] Central idea: Construct a feedback enthusiasm model, introduce the user emotion dimension, and conduct intelligent evaluation of the feedback enthusiasm and efficiency of the current human-machine interaction system for different categories of users. After different categories of users complete the feedback on the current human-machine interaction system, based on the feedback results, generate multiple groups of different user emotion evaluation matrices, and these evaluation matrices form the user emotion evaluation matrix set of this feedback.

[0013] Based on the user emotion evaluation matrix set of this feedback, retrieve the historical user emotion evaluation matrix set of the same human-machine interaction task for matching. Through matching, introduce system descriptive parameters to form a parameter set for optimizing this interaction, and propose improvement suggestions and output suggestions in this human-machine interaction system.

[0014] The optimized interaction method for a human-machine system based on intelligent evaluation of feedback enthusiasm includes the following steps:

[0015] S1. Based on natural speech processing technology, collect information on the speech interaction data between the human-machine interaction system and the user;

[0016] S2. Conduct multi-dimensional quantitative evaluation of user emotions based on the collected information;

[0017] S3. Construct a user classification model based on speech interaction content, semantics and user emotions, classify and identify the accessed users and mark them;

[0018] S4. Based on the feedback enthusiasm model, conduct intelligent evaluation of the feedback enthusiasm and efficiency of the current human-machine interaction system for different categories of users;

[0019] S5. After different categories of users complete the feedback on the current human-machine interaction system, generate multiple groups of different user emotion evaluation matrices based on the feedback results;

[0020] S6. Based on the set of historical user emotion evaluation matrices for the same human-computer interaction task retrieved from the set of user emotion evaluation matrices for the current feedback, perform matching;

[0021] S7. After the human-computer interaction system issues an instruction for this human-computer interaction task and before the next feedback is completed, guide the connected users;

[0022] S8. After the system receives multiple sets of feedback data from users of the same category, based on these feedback data, optimize the human-computer system interaction performance and, as extended data, expand the categories of the user classification model.

[0023] Further, the multi-dimensional quantitative evaluation includes evaluating the current user emotion based on five dimensions: speech rate, pitch, pitch change at the end of a sentence, word choice and recognition, and syntactic structure complexity;

[0024] For each of the five dimensions, collect 20 data points during the current period when the current user is performing voice input, and use the interaction moments of the 20 data points collected during this period as interaction nodes;

[0025] For these five dimensions, set a normal distribution interval for evaluation. If all dimension parameters fall within the interval, evaluate that the user has stable emotions. For values that do not fall within the interval, mark them using one of the five dimensions to which the feature belongs, and record the interaction node at which the value is generated;

[0026] The feature vectors of the five dimensions are:

[0027] Emotion feature vector Y = (y1, y2, y3, y4, y5)

[0028] where y1 represents speech rate; y2 represents pitch; y3 represents pitch change at the end of a sentence; y4 represents word choice and recognition; y5 represents syntactic structure complexity;

[0029] The user emotion feature vector can be expressed as:

[0030] U(i) = (u1(i), u2(i), u3(i), u4(i), u5(i))

[0031] where i is the collection sequence number, satisfying i = 1, 2,..., 20;

[0032] Within a load interval, satisfy:

[0033] u j (i) ∈ [MIN j , MAX j , j = 1, 2, 3, 4, 5

[0034] where MIN jis the lower limit of the positive and negative intervals, MIN j is the upper limit of the positive and negative intervals;

[0035] Furthermore, scale the user emotion feature vectors of each serial number to obtain the scaled user emotion feature vectors:

[0036]

[0037] Among them, represents the scaled user emotion feature component;

[0038] For each of the five dimensions, evaluate the proportion of the duration within the normal distribution probability interval to determine the user emotion characteristics.

[0039] Furthermore, the proportion evaluation specifically includes:

[0040] Taking the sampling data moment as a node and the time period composed of all moments as a time unit, for each of the five dimensions, calculate the probability of falling within the preset normal distribution interval as the falling probability P κ , where κ represents the dimension serial number, κ = 1, 2, 3, 4, 5;

[0041] Compare the probability P κ of falling within the preset normal distribution interval with the preset probability threshold G κ ;

[0042] If the judgment result satisfies:

[0043] P κ < G κ

[0044] Then it is considered that the user's emotion is unstable in a certain dimension under investigation, and execute the probability P κ of falling out of the preset normal distribution interval. The value is the numerical identifier, execute the marking, record the current interaction period, mark the user, and the user enters the global database of the human-computer interaction system; search for historical records with the same numerical identifier in the database and retrieve the parameter content.

[0045] Furthermore, the intelligent evaluation of feedback positivity performs a quantitative analysis on the evaluation language, specifically including the following steps:

[0046] S4.1. Perform natural speech processing on the evaluation language. After processing, form a sequence A(σ) of evaluation language words, where σ is the word serial number;

[0047] S4.2. Perform positive and negative emotion annotation on the sequence A(σ) of evaluation language words to obtain the emotion annotation word sequence

[0048] S4.3. Perform sentiment intensity annotation on the sentiment-annotated word sequence to obtain a sentiment intensity word sequence A°(σ);

[0049] S4.4. By statistically calculating the sentiment values of the absolute values of positive and negative bias words and negative sentiment words, comparing which sentiment the absolute value of the cutter is more inclined to, performing length comparison, and at the same time, statistically calculating the orientation of positive and negative sentiment of the evaluation language, synthesizing a vector, weighting the positive and negative sentiment values, obtaining the final sentiment value, forming a positive sentiment value with a positive bias and a negative sentiment value with a negative bias, and using this as a guide to introduce users of this category into the evaluation of human-computer interaction parameters;

[0050] S4.5. By statistically calculating the sentiment intensities of positive sentiment annotation and negative sentiment annotation and performing weighted statistics, obtain positive sentiment intensity and negative sentiment intensity, and form the positive evaluation value and negative evaluation value of the sentiment intensity word sequence A ° (σ).

[0051] Furthermore, the sentiment annotation and sentiment intensity grading scheme are as follows:

[0052] S4.2.1. Sentiment annotation is divided into positive sentiment and negative sentiment, where positive sentiment annotation is "1" and negative sentiment annotation is "-1";

[0053] S4.2.2. Sentiment intensity is divided into three levels, with the intensity increasing in sequence. The grading annotation scheme is as follows:

[0054] The sentiment intensity value of a gentle tone is annotated as b(σ) = ±1;

[0055] The sentiment expression intensity of a moderate tone is annotated as b(σ) = ±2;

[0056] The sentiment expression intensity of an intense tone is annotated as b(σ) = ±3.

[0057] Furthermore, it includes the following steps:

[0058] S4.5.1. Statistically calculate the number of positive sentiment annotation words a(σ+) and the number of negative sentiment annotation words a(σ-) in the feedback evaluation language, with the word numbers being σ+;

[0059] S4.5.2. Invoke the word co-occurrence algorithm to statistically calculate the number of occurrences of each positive sentiment annotation word and negative sentiment annotation word in the sequence store the qualitative sentiment value of evaluating it as a high-frequency word or a low-frequency word, and calculate the total value B(σ+) of positive sentiment annotation words, the total value B(σ-) of negative sentiment annotation words, and the sum, to obtain the positive evaluation value and negative evaluation value of the sentiment intensity word sequence A ° (σ), satisfying:

[0060]

[0061] Among them, b(σ+) represents the emotional intensity value of the positively sentiment-labeled words in A ° (σ), and b(σ-) represents the emotional intensity value of the negatively sentiment-labeled words in A ° (σ);

[0062] T(σ) = B(σ+) - B(σ-)

[0063] Among them, the said t(σ) represents the difference between the intensity of the positive words and the intensity of the negative words in this feedback.

[0064] Furthermore, the user evaluation content describing the system output result was subjected to evaluation quantification analysis, which specifically included the following steps:

[0065] S5.1: Natural language processing: Perform natural language processing on the evaluation content used by the user to describe the system output result to obtain the word sequence A * (δ) of the evaluation language, where δ is the word serial number;

[0066] S5.2: Historical evaluation comparison and emotional intensity classification: According to different user categories, retrieve the word sequence A(σ) of the human-machine system evaluation language used by this user for this feedback; Compare the newly generated word sequence A * (δ) with A(σ), and classify and categorize them according to the emotional intensity;

[0067] S5.3: Calculation of referenceability and algorithm optimization: Based on the user category, calculate the matching rate of the above comparison results to determine the referenceability of the machine system output result in this feedback; If the content output by the system for this category is understood and recognized by the user, it will be weighted in the subsequent standards and comprehensively calculated together with the referenceability submitted in history; When the referenceability exceeds the preset threshold, it indicates that the output result has a high reference value for this category, and the corresponding algorithm can be connected to other human-machine interaction links for further optimization;

[0068] S5.4: Complete the evaluation of this system interaction: According to the comparison results, complete the comprehensive evaluation of this machine system interaction: In the case of multiple output results, compare the referenceabilities of each output result, and select the result with higher referenceability as the solution for this interaction evaluation, input it into the system and return the final optimized result.

[0069] Furthermore, in the process of the interaction feedback of the human-machine system, the obtained interaction content is determined based on the quantification analysis of all user evaluations and through weighted calculation, and the specific steps are as follows:

[0070] S5.4.1: Generate initial reference values and perform weighted comparison: In one machine system interaction, for the same output result, different users will give their respective reference values; the reference values of all users are weighted and summarized, and compared and analyzed with the parameter requirements; the analysis results are usually divided into multiple aspects, each aspect contains positive and negative values, and through weighted calculation of these values, a comprehensive preliminary result is obtained.

[0071] S5.4.2: Positive and negative value statistics and weighting: During the weighting process, the reference values are classified into positive and negative categories according to user categories, and are statistically analyzed and compared respectively; then different weights are applied to these positive and negative reference values, and the proportions of positive and negative values are comprehensively calculated.

[0072] S5.4.3: Form a vector and determine the final reference value: Based on the positive and negative value statistical results, construct a vector with positive and negative values as the main axes; according to the component sizes of the vector, determine the final reference value of the output result.

[0073] S5.4.4: Average processing of multiple interaction output results: If multiple reference values are generated during multiple human-machine system interactions, they can be averaged to obtain a more representative final reference value.

[0074] Furthermore, by collecting the basic information of users and the hardware information of their logged-in hardware devices, user classification and identification are carried out, as well as comparison of emotional characteristics; based on their basic information, evaluation of emotional characteristics is completed, and emotional characteristics of the target population are classified and identified.

[0075] Furthermore, step S6 includes the following steps:

[0076] S6.1: Classification and identification of the set of historical user emotion evaluation matrices: Classify and identify the set of historical user emotion evaluation matrices, and complete the identification; after completion, obtain the unique identifier of the set of historical user emotion evaluation matrices, and the time dimension corresponding to this identifier.

[0077] S6.2: Matching degree analysis between the historical emotion evaluation matrix and the current interaction: Through comparison, retrieve the set of historical user emotion evaluation matrices that meet the requirements; after obtaining it, perform matching degree analysis on the performance of different user categories after entering the human-machine interaction system with the above matrix set; according to the artificially set time identifier, introduce the reference values in the feedback evaluation before the interaction and perform statistics, and finally determine the matching degree between the current user and the set of historical user emotion evaluation matrices.

[0078] S6.3. Confirm the correlation time point between the interaction process and the reference value: In the human-computer interaction system, determine the time point corresponding to the reference value of the user's feedback evaluation, that is, which type of machine the user first interacts with during the interaction process, the order of obtaining information, and the corresponding source of the reference value;

[0079] S6.4. Necessity comparison: Compare the reference values of the feedback evaluation to determine the priority in the current interaction requirements;

[0080] S6.5. Make an introduction decision based on the comparison result: According to the comparison result, guide the human-computer interaction system to select appropriate introduction content for the user in this interaction;

[0081] S6.6: System interaction evaluation and content output: When the human-computer interaction system obtains the corresponding interaction evaluation result for this interaction process, based on the effectiveness of further analysis of this result, generate or update corresponding parameters according to the evaluation result, so as to obtain the final interaction content after evaluation.

[0082] Beneficial effects

[0083] 1. Combine user feedback data with system interaction to achieve real-time optimization. Compared with traditional interaction optimization methods, it can respond to user needs more quickly, reduce interaction latency, and improve user satisfaction; by introducing user emotion classification and semantic analysis, it enhances the recognition and understanding of user emotions, thus providing a more user-friendly interaction experience; compared with existing methods, the present invention has a significant improvement in terms of optimization speed, user satisfaction, and system response, providing a more efficient solution for the technical development of human-computer interaction systems.

[0084] 2. The present invention combines user emotion feedback and interaction efficiency. By establishing a feedback positivity model, it dynamically adjusts the interaction performance of the system to meet the needs of different users. Different from the static response mode of the prior art, it can provide personalized optimization solutions through multi-dimensional analysis of user emotions and interaction behaviors, improve the intelligence of the human-machine system and user satisfaction. This method is not only applicable to human-computer interaction systems in multiple fields, but also can greatly improve the user's interaction experience, having broad application prospects and technical advantages.

[0085] 3. Evaluate the current user's emotion based on five dimensions: speech rate, pitch, pitch change at the end of a sentence, word choice and recognition, and syntactic structure complexity, and perform quantitative processing on the evaluation, including annotating common positive and negative emotions and grading their intensities, reducing the amount of data, enabling the human-machine system to quickly complete interaction evaluation, reducing interaction time consumption, and enabling the unmanned aerial vehicle system to quickly optimize its own interaction performance, further improving the user interaction experience. Brief description of the drawings

[0086] Figure 1 Schematic diagram of the method flow of an embodiment of the present invention.

[0087] Figure 2 Schematic diagram of the method flow of another embodiment of the present invention. Detailed implementation manners

[0088] The specific implementation manners of the present invention are described in detail in the form of embodiments. Other implementation manners completed by those skilled in the art without departing from the concept of the present invention shall be understood to fall within the protection scope of the present invention.

[0089] Embodiment 1

[0090] As Figure 1 shown, this embodiment provides an optimized interaction method for a human-machine system based on intelligent evaluation of feedback enthusiasm, including:

[0091] S1. First, through natural speech processing technology, information collection of voice interaction data between the interaction system and the user is performed.

[0092] S2. Based on the above information, a user emotion evaluation matrix is established, and multi-dimensional quantitative evaluation of the user emotion is performed. The specific steps include:

[0093] S2.1. Establish a multi-dimensional quantitative evaluation method for the emotion of the user; the multi-dimensional quantitative evaluation includes evaluating the current user emotion based on five dimensions: speech rate, pitch, pitch change at the end of the sentence, word choice and recognition, and syntactic structure complexity.

[0094] Among them, for each of the five dimensions, 20 data are collected during the current period when the user performs voice input, and the interaction moments of the 20 data collections during the current period are used as interaction nodes.

[0095] The feature vectors of the five dimensions are composed of:

[0096] Emotion feature vector Y = (y1, y2, y3, y4, y5)

[0097] where y1 represents the speech rate; y2 represents the pitch; y3 represents the pitch change at the end of the sentence; y4 represents the word choice and recognition; y5 represents the syntactic structure complexity.

[0098] The user emotion feature vector can be expressed as:

[0099] U(i) = (u1(i), u2(i), u3(i), u4(i), u5(i))

[0100] where i is the collection serial number, satisfying i = 1, 2,..., 20;

[0101] Within a load range, the following is satisfied:

[0102] u j (i) ∈ [MIN j , MAX j , j = 1, 2, 3, 4, 5

[0103] where MIN j is the lower limit of the positive and negative interval, and MAX j is the upper limit of the positive and negative interval;

[0104] Perform scaling processing on the user emotion feature vectors of each serial number to obtain the scaled user emotion feature vectors:

[0105]

[0106] where represents the scaled user emotion feature component;

[0107] For each of the five dimensions, evaluate the proportion of the duration that satisfies the normal distribution probability interval to determine the user emotion feature

[0108] For each of the five dimensions, evaluate the proportion of the duration that satisfies the normal distribution probability interval to determine the user emotion feature, specifically including:

[0109] For each of the five dimensions, calculate the probability of falling into the preset normal distribution interval as the falling probability Pκ, where κ represents the dimension serial number, κ = 1, 2, 3, 4, 5;

[0110] Compare the probability Pκ of falling into the preset normal distribution interval with the preset large probability Gκ;

[0111] If the judgment result satisfies:

[0112] Pκ < Gκ

[0113] Then it is considered that the user's emotion is unstable in a certain dimension under investigation. Execute the value of the probability Pκ of entering and exiting the preset normal distribution interval as a numerical identifier, execute the marking, and record the current interaction period. Mark the user, and the user enters the global database of the human-computer interaction system; search for historical records with the same numerical identifier in the database and retrieve the parameter content;

[0114] S3. Construct a user classification model based on voice interaction content, semantics, and user emotion. Collect the user's basic information and hardware information for the introduced user, and complete classification recognition and identification; the user includes users who log in to the human-computer system for the first time and users who log in to the human-computer system multiple times.

[0115] Including identifying and labeling users who log in to the human-machine system multiple times, including the following steps:

[0116] S3.1. Record the login duration of the user who logs in to the human-machine system multiple times and the hardware device information used by the user to complete the login process; when the human-machine system conducts voice interaction with the user, record the user's voice data;

[0117] S3.2. When the user's voice data is sufficient, call the user classification model to classify, identify, and label the user;

[0118] S3.3. Determine whether the user is a user who logs in to the human-machine system for the first time;

[0119] If so, classify and label the user according to the user's basic information and the hardware information, and collect and store the user's basic information and the hardware information in the database of the human-computer interaction system;

[0120] Otherwise, train the self-learning module in the user classification model to optimize the user classification model;

[0121] S4. Quantitatively analyze the user feedback language evaluation for logging in to the machine system based on the intelligent evaluation of feedback enthusiasm, specifically including:

[0122] S4.1. Perform natural speech processing on the evaluation language. After processing, form a sequence A(σ) of evaluation language words, where σ is the word serial number;

[0123] S4.2. Perform positive and negative emotion annotation on the sequence A(σ) of evaluation language words to obtain an emotion-annotated word sequence

[0124] S4.3. Perform emotion intensity annotation on the emotion-annotated word sequence to obtain an emotion intensity word sequence A°(σ);

[0125] S4.4. By statistically calculating the emotional values of the absolute values of the positive-biased words and negative-biased words, comparing which emotion the absolute value of the comparison cutter is more inclined to, performing length comparison, and at the same time, statistically calculating the positive and negative emotion orientation of the evaluation language, synthesizing a vector, weighting the positive and negative emotion values, obtaining the final emotional value, forming a positive emotion value with a positive bias and a negative emotion value with a negative bias, as a guide, introducing the users of this category into the human-computer interaction parameter evaluation;

[0126] S4.5. By statistically calculating the emotional intensities of the positive emotion annotation and the negative emotion annotation, performing weighted statistics, obtaining the positive emotion intensity and the negative emotion intensity, and forming an emotion intensity word sequence A° (σ) positive evaluation value and negative evaluation value.

[0127] Furthermore, the sentiment annotation and sentiment intensity grading scheme are as follows:

[0128] S4.2.1. Sentiment annotation is divided into positive sentiment and negative sentiment, where positive sentiment is annotated as "1" and negative sentiment is annotated as "-1";

[0129] S4.2.2. Sentiment intensity is divided into three levels, with the intensity increasing in sequence. The grading annotation scheme is as follows:

[0130] The sentiment intensity value of a gentle tone is annotated as b(σ) = ±1;

[0131] The sentiment expression intensity of a moderate tone is annotated as b(σ) = ±2;

[0132] The sentiment expression intensity of an intense tone is annotated as b(σ) = ±3.

[0133] Including the following steps:

[0134] S4.5.1. Count the number of positive sentiment annotation words a(σ+) and the number of negative sentiment annotation words a(σ-) in the language for feedback evaluation, with the word numbers being σ+;

[0135] S4.5.2. Invoke the word co-occurrence algorithm to count the number of occurrences of each positive sentiment annotation word and negative sentiment annotation word in the sequence , evaluate and store their qualitative sentiment values as high-frequency words or low-frequency words, and calculate the total value B(σ+) of positive sentiment annotation words, the total value B(σ-) of negative sentiment annotation words, and the sum, to obtain the positive evaluation value and negative evaluation value of the sentiment intensity word sequence A°(σ), satisfying:

[0136]

[0137] where b(σ+) represents the sentiment intensity value of positive sentiment annotation words in A°(σ), and b(σ-) represents the sentiment intensity value of negative sentiment annotation words in A°(σ);

[0138] T(σ) = B(σ+) - B(σ-)

[0139] where the T(σ) represents the difference between the positive word intensity and the negative word intensity of this feedback.

[0140] S5. After different categories of users complete the feedback on the current human-computer interaction system, based on the feedback results, generate multiple groups of different user emotion evaluation matrices; these evaluation matrices form the user emotion evaluation matrix set of this feedback;

[0141] S6. Retrieve the historical user emotion evaluation matrix set for the same human-computer interaction task for matching; through the matching, introduce system descriptive parameters to form a parameter set for this interaction optimization, propose improvement suggestions in this human-computer interaction system, and output the suggestions.

[0142] Embodiment 2

[0143] Based on Embodiment 1, the specific implementation steps are as follows:

[0144] Through natural language processing technology, collect the voice interaction data from different users. First, take the human-machine system as the server, and the users of the human-machine system as the clients of the human-machine system to actively guide users to express their emotions. At the same time, collect the basic information of the users and the human-machine voice interaction data. These interaction data include the voice files of the users and the specific performances during user interactions. Then, analyze these data to extract the main features of the users' emotions. This collection process can ensure the comprehensiveness and diversity of the data, providing rich information for subsequent emotion analysis. When the user first logs in to the human-machine system, collect the basic information of the user.

[0145] Based on the basic information of the user and their voice data, conduct user classification. The user classification in this method includes two levels. The first level is to classify users based on the basic information of the users. For the developers of the human-machine system, they will pre-control the audience of the human-machine system. For example, only open it to users of a certain age group. Therefore, the human-machine system can determine whether the user legally and compliantly logs in to the system through the basic information of the user. The second level is to determine that the user meets the parameter requirements of the hardware device after determining that the user meets the conditions for logging in to the human-machine system. For example, the human-machine system stipulates that the user needs to use a hardware device with a certain row of microphones to handle a certain business. At this time, the human-machine system needs to check whether the user meets the usage requirements. After the user meets the necessary login conditions, the user can use the human-machine system. The necessary login conditions are optimized with the user's voice data for the human-machine system. This optimization is limited, that is, to optimize the system for the necessary condition-related hardware devices used by the user for this system.

[0146] After completing user classification, based on the user's voice data, a quantitative analysis of the user's emotions is carried out to form a user emotion evaluation matrix. In addition to the user's basic information, the user's voice data is also collected and analyzed. The analysis also includes a quantitative analysis based on five dimensions: speech rate, pitch, pitch change at the end of a sentence, word choice and recognition, and syntactic structure complexity. The eigenvectors of the above five dimensions are introduced to form an emotion eigenvector matrix, completing the feature extraction of the user's voice data. After completing the user language feature extraction, the human-machine system can complete the user emotion determination according to needs and thereby complete the parameter requirements for logging in to the necessary hardware. This determination has a dual purpose for the user. It can identify the emotion characteristics of sensitive groups to clarify whether this group is familiar enough with the human-machine system, enabling the human-machine system to autonomously determine whether this group is suitable for using the human-machine system; for non-sensitive groups, it can complete the auxiliary human-machine system interaction based on their voice characteristics. For example, if the voice is relatively sharp, the volume is reduced while ensuring speech clarity.

[0147] For a human-machine interaction task, the human-machine system will be optimized after obtaining the interaction evaluation. Before that, the human-machine system will generate multiple sets of different user emotion evaluation matrices after different categories of users complete the feedback on the current human-machine interaction system; these evaluation matrices form the set of user emotion evaluation matrices for this feedback. During this process, the content includes the user's feedback, which is generally the result introduced by the human-machine system, that is, the advice given to the user is called the interaction task. After introducing the interaction task, the user will generate feedback, and these feedbacks will be input into the human-machine system and become the optimization content of the human-machine system. Regarding the feedback positivity, that is, the feedback content of the interaction task, the above natural speech processing is completed, and based on the user evaluation data, emotion annotation and intensity grading are completed. The positive and negative emotions are annotated as numerical values, the frequency of each numerical value is counted, and weighted according to the intensity, so as to obtain the feedback positivity of the user evaluation. This positivity, combined with the user emotion evaluation matrix, becomes the content input for the optimization of the machine system.

[0148] The machine system outputs. Before the output, it retrieves the set of historical user emotion evaluation matrices for the same human-machine interaction task; it completes the matching in the historical evaluation set. This matching method is comparison. For example, it compares the emotion data of the historical user and the user emotion data obtained from the current task of the human-machine system to get feedback. If they are the same, it is ignored. If they are different, it retrieves the historical feedback and the current feedback for comparison. If they are the same, it is not processed. If they are different, it retrieves the reference values for the introduction of the human-machine system evaluated in the historical feedback and the reference values for the introduction of the current feedback for comparison. After matching with the current interaction task in the historical introduction result set, it completes the statistics and generates a comment conclusion, and then obtains the introduction suggestion. When the user enters the human-machine system, the above introduction suggestion is displayed. Finally, it obtains the data for this interaction, that is, whether the user browses the suggestion and whether the user is satisfied with the suggestion after entering the system, and finally expands the user category entering the human-machine system; on this basis, it generates the introduced content, that is, the interaction task, and generates interaction data after execution, which is also the performance of machine interaction. Through the introduction, it obtains the content of the human-machine system, involving hardware devices, human emotions, and the introduced content of the human-machine system.

[0149] Embodiment 3

[0150] Based on Embodiment 1, it includes the following steps:

[0151] Through natural language processing technology, it collects the voice interaction data from different users. First, taking the human-machine system as the server and the users of the human-machine system as the clients of the human-machine system, it actively guides the users to express their emotions, and at the same time collects the basic information of the users and the human-machine voice interaction data. These interaction data include the voice files of the users and the specific performances during the user interaction. Then it analyzes these data to extract the main features of the user emotions. This collection process can ensure the comprehensiveness and diversity of the data, providing rich information for the subsequent emotion analysis. When the user first logs in to the human-machine system, it collects the basic information of the user.

[0152] Based on the user's basic information and their voice data, user classification is carried out. The user classification in this method includes two levels. At the first level, user classification is based on the user's basic information. For the developers of the human-machine system, they will pre-control the audience of the human-machine system. For example, it is only open to users in a certain age group. Therefore, the human-machine system can determine whether the user logs in to the system legally and compliantly through the user's basic information. At the second level, after determining that the user meets the conditions for logging in to the human-machine system, it is determined whether the user needs to respond to the parameter requirements of the hardware device. For example, the human-machine system stipulates that the user needs to use a hardware device with a certain row of microphones to handle a certain business. At this time, the human-machine system needs to check whether the user meets the usage requirements. After the user meets the necessary login conditions, the human-machine system can be used. The necessary login conditions are optimized with the user's voice data. This optimization is limited, that is, the system is optimized for the necessary hardware devices related to the user's usage conditions of this system.

[0153] After completing the user classification, based on the user's voice data, a quantitative analysis of the user's emotions is carried out to form a user emotion evaluation matrix. In addition to the user's basic information, the user's voice data is also collected and analyzed. It also includes a quantitative analysis based on five dimensions: speech rate, pitch, pitch change at the end of the sentence, word choice and recognition, and syntactic structure complexity. The feature vectors of the above five dimensions are introduced to form an emotion feature vector matrix to complete the feature extraction of the user's voice data. After completing the user language feature extraction, the human-machine system can complete the user emotion determination as needed and thereby complete the parameter requirements of the necessary login hardware. This determination has a dual purpose for the user. It can identify the emotion characteristics of sensitive populations to clarify whether this population is familiar enough with the human-machine system, so that the human-machine system can independently determine whether this population is suitable for using the human-machine system; for non-sensitive populations, it can complete the auxiliary human-machine system interaction according to their voice characteristics. For example, if the voice is relatively sharp, the volume is reduced while ensuring speech clarity.

[0154] For a human-computer interaction task, the human-computer system will be optimized after obtaining the interaction evaluation. Before that, after different categories of users complete the feedback on the current human-computer interaction system, the human-computer system will generate multiple groups of different user emotion evaluation matrices; these evaluation matrices form the set of user emotion evaluation matrices for this feedback. During this process, the content includes the feedback from the users. This feedback is generally the result introduced by the human-computer system, that is, the suggestions given to the users are called interaction tasks. After introducing the interaction tasks, the users will generate feedback, and these feedbacks will be input into the human-computer system and become the optimization content of the human-computer system. Regarding the feedback positivity, that is, the feedback content of the interaction tasks, the above-mentioned natural language processing is completed, and based on the user evaluation data, emotion annotation and intensity grading are completed. The positive and negative emotions are annotated as numerical values, the frequency of each numerical value is counted, and weighted according to the intensity, so as to obtain the feedback positivity of the user evaluation. This positivity, combined with the user emotion evaluation matrix, becomes the content input for the optimization of the machine system.

[0155] Taking the machine system as the output, before the output, the historical user emotion evaluation matrix set of the same human-computer interaction task is retrieved; matching is completed in the historical evaluation set. This matching method is comparison. For example, the emotion data of the historical users is compared with the user emotion data obtained from the current task of the human-computer system to get feedback. If they are the same, it is ignored. If they are different, the historical feedback and the current feedback are retrieved for comparison. If they are the same, no processing is done. If they are different, the reference values for introduction of the human-computer system evaluated in the historical feedback and the reference values for introduction of the current feedback are retrieved for comparison. After matching with the current interaction task in the historical introduction result set, statistics are completed to generate a comment conclusion, and thus the introduction suggestions can be obtained. When the user enters the human-computer system, the above-mentioned introduction suggestions are displayed. Finally, the data for this interaction is obtained, that is, after the user enters the system, whether the suggestions are browsed and whether the suggestions are satisfied. Finally, the user categories entering the human-computer system are expanded; on this basis, the introduced content, that is, the interaction tasks, is generated. After execution, interaction data is generated, which is also the performance of machine interaction. Through introduction, the content of the human-computer system is obtained, involving hardware devices, human emotions, and the human-computer system introduces content. If the service of this content ends, the feedback result enters the feedback process and is stored in the database, so that one introduction is completed; if the introduced content is retrieved and used in the docking of a service, the interaction performance of the human-computer system can be optimized through these feedback results; after one feedback, the introduction process ends, and hardware devices are introduced during this process.

[0156] Embodiment 4

[0157] As Figure 2 shown, this embodiment gives an optimized interaction method for an English teaching human-computer system based on intelligent evaluation of feedback positivity, including the following steps:

[0158] S1. Collect information on the voice interaction data between the English teaching human-computer interaction system and students based on natural language processing technology;

[0159] S2. Conduct multi-dimensional quantitative evaluation of students' emotions based on the collected information;

[0160] S3. Build a student classification model based on voice interaction content, semantics, and students' emotions, classify and identify the accessed students and label them;

[0161] S4. Based on the feedback positivity model, conduct intelligent evaluation of the feedback positivity and efficiency of the current English teaching human-computer interaction system for different categories of students;

[0162] S5. After different categories of students complete the feedback on the current English teaching human-computer interaction system, generate multiple groups of different student emotion evaluation matrices based on the feedback results;

[0163] S6. Based on the set of student emotion evaluation matrices from this feedback, retrieve the set of historical student emotion evaluation matrices for the same English teaching human-computer interaction task and perform matching;

[0164] S7. After the English teaching human-computer interaction system issues an instruction for this English teaching human-computer interaction task and before the next feedback is completed, guide the accessed students;

[0165] S8. After the system receives feedback data from multiple students of the same category, optimize the interaction performance of the English teaching human-computer system based on these feedback data and use it as extended data to expand the categories of the student classification model.

[0166] Furthermore, the multi-dimensional quantitative evaluation includes evaluating the current student emotions based on five dimensions: speech rate, pitch, pitch change at the end of a sentence, word choice and recognition, and grammatical structure complexity;

[0167] For each of the five dimensions, collect 20 data points during the current period when the current student is performing voice input, and use the interaction moments of the 20 data points collected during this period as interaction nodes;

[0168] For these five dimensions, set a normal distribution interval for evaluation. If all dimension parameters fall within the interval, evaluate the student as having stable emotions. For values that do not fall within the interval, mark them using one of the five dimensions where the feature lies and record the interaction node at which the value is generated;

[0169] The feature vector group of the five dimensions is:

[0170] Emotion feature vector Y = (y1, y2, y3, y4, y5)

[0171] Among them, y1 represents the speech rate; y2 represents the pitch; y3 represents the pitch change at the end of a sentence; y4 represents the word choice and recognition; y5 represents the complexity of the grammatical structure;

[0172] The student emotion feature vector can be expressed as:

[0173] U(i) = (u1(i), u2(i), u3(i), u4(i), u5(i))

[0174] Among them, i is the acquisition serial number, satisfying i = 1, 2,..., 20;

[0175] Within a load range, it satisfies:

[0176] u j (i) ∈ [MIN j , MAX j , j = 1, 2, 3, 4, 5

[0177] Among them, MIN j is the lower limit of the positive and negative interval, and MAX j is the upper limit of the positive and negative interval;

[0178] Furthermore, the scaling process of the student emotion feature vectors of each serial number is performed to obtain the scaled student emotion feature vectors:

[0179]

[0180] Among them, represents the scaled student emotion feature component;

[0181] For each of the five dimensions, the duration satisfying the normal distribution probability interval is evaluated proportionally to determine the student emotion feature.

[0182] Furthermore, the proportional evaluation specifically includes:

[0183] Taking the sampling data moment as a node and the time period composed of all moments as a time unit, for each of the five dimensions, calculate the probability of falling into the preset normal distribution interval as the falling probability P κ , where κ represents the dimension serial number, κ = 1, 2, 3, 4, 5;

[0184] Compare the probability P κ of falling into the preset normal distribution interval with the preset probability threshold G κ ;

[0185] The judgment result is as follows if it satisfies:

[0186] P κ < G κ

[0187] It is considered that the student is emotionally unstable in a certain dimension under investigation, and the probability P of deviating from the preset normal distribution interval is κ The value of is a numerical identifier. Perform a mark and record the current interaction period. Mark the student, and the student enters the global database of the human-computer interaction system; search for historical records with the same numerical identifier in the database and retrieve the parameter content.

[0188] Furthermore, the intelligent evaluation of feedback positivity conducts a quantitative analysis on the evaluation language, specifically including the following steps:

[0189] S4.1. Perform natural speech processing on the evaluation language. After processing, a sequence A(σ) of evaluation language words is formed, where σ is the word serial number;

[0190] S4.2. Perform positive and negative emotion annotation on the sequence A(σ) of evaluation language words to obtain an emotion-annotated word sequence

[0191] S4.3. Perform emotion intensity annotation on the emotion-annotated word sequence to obtain an emotion intensity word sequence A ° (σ);

[0192] S4.4. By statistically calculating the emotional values of the absolute values of positive-biased words and negative-biased emotion words, comparing the absolute values of the cutter, determining which emotion it is more inclined to, performing length comparison, and at the same time, statistically calculating the positive and negative emotion orientation of the evaluation language, synthesizing a vector, weighting the positive and negative emotion values, obtaining the final emotional value, forming a positive emotion value biased towards the positive and a negative emotion value biased towards the negative, as a guide, introducing students of this category into the human-computer interaction parameter evaluation;

[0193] S4.5. By statistically calculating the emotion intensities of positive emotion annotation and negative emotion annotation and performing weighted statistics, obtaining the positive emotion intensity and negative emotion intensity, forming the positive evaluation value and negative evaluation value of the emotion intensity word sequence A ° (σ).

[0194] Furthermore, the emotion annotation and emotion intensity grading scheme are as follows:

[0195] S4.2.1. Emotion annotation is divided into positive emotion and negative emotion, where positive emotion annotation is "1" and negative emotion annotation is "-1";

[0196] S4.2.2. Emotion intensity is divided into three levels, with the intensity increasing in sequence. The grading annotation scheme is as follows:

[0197] The emotion intensity value of a gentle tone is annotated as b(σ) = ±1;

[0198] The emotional expression intensity with a moderate tone is labeled as b(σ) = ±2;

[0199] The emotional expression intensity with an intense tone is labeled as b(σ) = ±3.

[0200] Furthermore, it includes the following steps:

[0201] S4.5.1: Count the number of positively emotional labeled words a(σ+) and negatively emotional labeled words a(σ-) in the language for feedback evaluation, where the word numbers are σ+ and σ- respectively;

[0202] S4.5.2: Invoke the word co-occurrence algorithm to count the occurrence times of each positively emotional labeled word and negatively emotional labeled word in the sequence and store the qualitative emotional values evaluating them as high-frequency or low-frequency words, and calculate the total value B(σ+) of positively emotional labeled words, the total value B(σ-) of negatively emotional labeled words, and the sum, to obtain the positive evaluation value and negative evaluation value of the emotional intensity word sequence A ° (σ), satisfying:

[0203]

[0204] where b(σ+) represents the emotional intensity value of positively emotional labeled words in A ° (σ), and b(σ-) represents the emotional intensity value of negatively emotional labeled words in A ° (σ);

[0205] T(σ) = B(σ+) - B(σ-)

[0206] where the T(σ) represents the difference between the positive word intensity and the negative word intensity of this feedback.

[0207] Furthermore, a quantitative analysis of the evaluation of the student evaluation content describing the system output result is carried out, specifically including the following steps:

[0208] S5.1: Natural language processing: Perform natural language processing on the evaluation content used by students to describe the system output result to obtain the word sequence A * (δ) of the evaluation language, where δ is the word number;

[0209] S5.2: Historical evaluation comparison and emotional intensity classification: According to different student categories, retrieve the word sequence A(σ) of the evaluation language used by the student for the English teaching human-computer system evaluation of this feedback; Compare the newly generated word sequence A * (δ) with A(σ) and classify and categorize them according to the emotional intensity;

[0210] S5.3: Calculation of Reference Degree and Algorithm Optimization: Based on the student category, calculate the matching rate of the above comparison results to determine the reference degree of the output result of the machine system in this feedback. If the content output by the system for this category is understood and recognized by the students, it will be weighted in the subsequent standards and comprehensively calculated together with the reference degrees submitted in history. When the reference degree exceeds the preset threshold, it indicates that the output result has high reference value for this category, and the corresponding algorithm can be connected to other human-computer interaction links for further optimization.

[0211] S5.4: Completion of the Evaluation of this System Interaction: According to the comparison results, complete the comprehensive evaluation of this machine system interaction. In the case of multiple output results, compare the reference degrees of each output result, and select the result with a higher reference degree as the solution for this interaction evaluation, input it into the system and return the final optimized result.

[0212] Furthermore, in the process of interactive feedback of the human-machine system, the obtained interactive content is determined based on the quantitative analysis of all students' evaluations and through weighted calculation. The specific steps are as follows:

[0213] S5.4.1: Generation of the Initial Reference Value and Weighted Comparison: In a machine system interaction, for the same output result, different students will give their respective reference values. Aggregate the reference values of all students through weighting and compare and analyze them with the parameter requirements. The analysis results are usually divided into multiple aspects, and each aspect contains positive and negative values. By performing weighted calculation on these values, a comprehensive preliminary result is obtained.

[0214] S5.4.2: Statistics and Weighting of Positive and Negative Values: During the weighting process, divide the reference values into positive and negative categories according to the student category, and conduct statistics and comparison respectively. Subsequently, apply different weights to these positive and negative reference values and comprehensively calculate the proportions of the positive and negative values.

[0215] S5.4.3: Formation of a Vector and Determination of the Final Reference Value: Based on the statistical results of positive and negative values, construct a vector with positive and negative values as the main axes. Determine the final reference value of the output result according to the component sizes of the vector.

[0216] S5.4.4: Average Processing of the Output Results of Multiple Interactions: If multiple reference values are generated during multiple interactions of the human-machine system, they can be averaged to obtain a more representative final reference value.

[0217] Furthermore, through the collection of students' basic information and the hardware information of their logged-in hardware devices, conduct student classification and identification, as well as comparison of emotional characteristics. Based on their basic information, complete the evaluation of emotional characteristics, and conduct classification and identification of emotional characteristics for the target population.

[0218] Further, step S6 includes the following steps:

[0219] S6.1. Classification and identification of the set of historical student emotion evaluation matrices: Classify and identify the set of historical student emotion evaluation matrices, and complete the identification. After completion, obtain the unique identifier of the set of historical student emotion evaluation matrices and the corresponding time dimension.

[0220] S6.2. Analysis of the matching degree between the historical emotion evaluation matrix and the current interaction: By comparison, retrieve the set of historical student emotion evaluation matrices that meet the requirements. After obtaining it, analyze the matching degree between the performance of different student categories after entering the human-computer interaction system and the above matrix set. According to the artificially set time identifier, introduce the reference values in the feedback evaluation before the interaction and conduct statistics, and finally determine the matching degree between the current student and the set of historical student emotion evaluation matrices.

[0221] S6.3. Confirm the time point associated with the reference value in the interaction process: In the English teaching human-computer interaction system, determine the time point corresponding to the reference value of the student's feedback evaluation, that is, which type of machine the student first interacts with, the order of obtaining information, and the corresponding source of the reference value during the interaction process.

[0222] S6.4. Necessity comparison: Compare the reference values of the feedback evaluation to determine the priority in the current interaction requirements.

[0223] S6.5. Make an introduction decision based on the comparison result: According to the comparison result, guide the English teaching human-computer interaction system to select appropriate introduction content for the student in this interaction.

[0224] S6.6: System interaction evaluation and content output: When the English teaching human-computer interaction system obtains the corresponding interaction evaluation result for this interaction process, based on the effectiveness of the further analysis of this result, generate or update the corresponding parameters according to the evaluation result, so as to obtain the final interaction content after evaluation.

[0225] The specific implementation manners disclosed for the above embodiments are only used to help understand the core idea of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the core idea of the present invention, several modifications and improvements can be made, and these modifications and improvements also fall within the scope protected by the claims of the present invention.

Claims

1. An optimized interaction method for a human-machine system based on intelligent evaluation of feedback enthusiasm, characterized in that It includes the following steps: S1. Based on natural speech processing technology, collect information on the speech interaction data between the human-computer interaction system and the user; S2. Based on the collected information, conduct a multi-dimensional quantitative evaluation of the user's emotion; S3. Construct a user classification model based on speech interaction content, semantics, and user emotion, classify and identify the accessed users, and mark them; S4. Based on the feedback positivity model, conduct an intelligent evaluation of the feedback positivity and efficiency of the current human-computer interaction system for different categories of users; S5. After different categories of users complete the feedback on the current human-computer interaction system, based on the feedback results, generate multiple groups of different user emotion evaluation matrices; S6. Based on the set of user emotion evaluation matrices of the current feedback, retrieve the set of historical user emotion evaluation matrices of the same human-computer interaction task for matching; S7. After the human-computer interaction system issues an instruction for this human-computer interaction task and before the next feedback is completed, guide the accessed users; S8. After the system receives multiple groups of feedback data from users of the same category, based on these feedback data, optimize the interaction performance of the human-machine system and, as extended data, expand the categories of the user classification model.

2. The optimized interaction method for a human-machine system based on intelligent evaluation of feedback enthusiasm according to claim 1, characterized in that, The multi-dimensional quantitative evaluation includes evaluating the current user emotion based on five dimensions: speech rate, pitch, pitch change at the end of a sentence, word choice and recognition, and syntactic structure complexity; For each of the five dimensions, collect 20 times of data during the current period when the current user makes a speech input, and use the interaction moments of the 20 times of data collection during this period as interaction nodes; For these five dimensions, set a normal distribution interval for evaluation. If all dimension parameters fall within the interval, evaluate that the user has stable emotions. For values that do not fall within this interval, mark them using one of the five dimensions where the feature is located, and record the interaction node where the value is generated; The feature vector group of the five dimensions is: Emotion feature vector Y=(y1,y2,y3,y4,y5) Among them, y1 represents the speech rate; y2 represents the pitch; y3 represents the pitch change at the end of a sentence; y4 represents the word choice and recognition; y5 represents the syntactic structure complexity; The user emotion feature vector can be expressed as: U(i)=(u1(i),u2(i),u3(i),u4(i),u5(i)) Among them, i is the collection serial number, satisfying i = 1, 2,..., 20; Within a load interval, it satisfies: u j (i) ∈ [MIN j , MAX j , j = 1, 2, 3, 4, 5 Among them, MIN j is the lower limit of the positive and negative interval, and MAX j is the upper limit of the positive and negative interval.

3. The optimized interaction method for a human-machine system based on intelligent evaluation of feedback enthusiasm according to claim 2, wherein Perform scaling processing on the user emotion feature vectors of each serial number to obtain the scaled user emotion feature vectors; Among them, represents the scaled user emotion feature component; For each of the five dimensions, conduct a proportional evaluation of the duration that satisfies the normal distribution probability interval to determine the user emotion feature.

4. The optimized interactive method for a human-machine system based on intelligent evaluation of feedback enthusiasm according to claim 3, wherein The proportional evaluation specifically includes: Taking the sampling data moment as a node and the time period composed of all moments as the time unit, for each of the five dimensions, calculate the probability of falling into the preset normal distribution interval as the falling probability P κ , where κ represents the dimension serial number, κ = 1, 2, 3, 4, 5; The probability P that falls within a preset normal distribution interval κ is compared with a preset probability threshold G κ ; The judgment result satisfies: P κ <G κ It is considered that the user's emotion is unstable in a certain dimension under investigation, and the probability P of executing out of the preset normal distribution interval κ The value is a numerical identifier, execute the marking, and record the current interaction period, mark the user, and the user enters the global database of the human-computer interaction system; search for historical records with the same numerical identifier in the database and retrieve the parameter content.

5. The optimized interaction method for a human-machine system based on intelligent evaluation of feedback enthusiasm according to claim 4, characterized in that The intelligent evaluation of the feedback positivity conducts a quantitative analysis on the evaluation language, specifically including the following steps: S4.

1. Conduct natural speech processing on the evaluation language. After the processing is completed, form a sequence A(σ) of the words of the evaluation language, where σ is the word serial number; S4.

2. Perform sentiment positive / negative annotation on the sequence A(σ) of evaluation language words to obtain a sentiment-annotated word sequence S4.

3. Perform sentiment intensity annotation on the sentiment-annotated word sequence to obtain a sentiment intensity word sequence A ° (σ); S4.

4. By statistically calculating the sentiment values of the absolute values of positive and negative bias words and negative sentiment words, comparing the absolute values of the mowers, determining which sentiment is more inclined, conducting length comparison, and at the same time, statistically calculating the positive and negative sentiment orientation of the evaluation language, synthesizing a vector, weighting the positive and negative sentiment values, obtaining the final sentiment value, forming a positive sentiment value with a positive bias and a negative sentiment value with a negative bias, and using it as a guide to introduce users of this category into the evaluation of human-computer interaction parameters; S4.

5. By statistically analyzing the sentiment intensities of positive sentiment annotations and negative sentiment annotations, performing weighted statistics to obtain the positive sentiment intensity and the negative sentiment intensity, and forming a sentiment intensity word sequence A ° (σ)'s positive evaluation value and negative evaluation value.

6. The human-machine system optimization interaction method based on the intelligent evaluation of feedback positivity according to claim 5, characterized in that, The sentiment annotation and sentiment intensity grading scheme are as follows: S4.2.

1. Sentiment annotation is divided into positive sentiment and negative sentiment, where positive sentiment is annotated as "1" and negative sentiment is annotated as "-1"; S4.2.

2. Sentiment intensity is divided into three levels, with the intensity increasing in sequence. The grading annotation scheme is as follows: The sentiment intensity value with a gentle tone is annotated as b(σ) = ±1; The sentiment expression intensity with a moderate tone is annotated as b(σ) = ±2; The sentiment expression intensity with a strong tone is annotated as b(σ) = ±3.

7. The method for optimizing the interaction of a human-machine system based on intelligent evaluation of feedback enthusiasm according to claim 6, characterized in that, It includes the following steps: S4.5.

1. Statistically calculate the number of positive sentiment annotation words a(σ+) in the feedback evaluation language, with the word serial number σ+ and the number of negative sentiment annotation words a(σ-); S4.5.

2. Call the word co-occurrence algorithm to count the number of occurrences of each positively sentiment-annotated word and negatively sentiment-annotated word in the sequence store the qualitative sentiment value evaluated as a high-frequency word or a low-frequency word, calculate the total value B(σ+) of the positively sentiment-annotated words, the total value B(σ-) of the negatively sentiment-annotated words, and obtain the sentiment intensity word sequence A ° (σ)'s positive evaluation value and negative evaluation value, satisfying: Among them, b(σ+) represents A ° (σ) the emotional intensity value of the positively sentiment-annotated words, and b(σ-) represents A ° (σ) the emotional intensity value of the negatively sentiment-annotated words; T(σ) = B(σ+) - B(σ-) Among them, the T(σ) represents the difference between the positive word intensity and the negative word intensity of this feedback.

8. The method for optimizing the interaction of a human-machine system based on intelligent evaluation of feedback enthusiasm according to claim 7, characterized in that, Quantitative analysis and evaluation were carried out on the user evaluation content describing the system output results, specifically including the following steps: S5.1: Natural language processing: Perform natural language processing on the evaluation content used by the user to describe the system output result to obtain the word sequence A of the evaluation language * (δ), where δ is the word serial number; S5.2: Historical evaluation comparison and emotional intensity classification: According to different user categories, retrieve the word sequence A(σ) of the human-machine system evaluation language used by the user for this feedback; Compare the newly generated word sequence A * (δ) with A(σ), and classify and categorize them according to the emotional intensity; S5.3: Calculation of referenceability and algorithm optimization: Based on the user category, calculate the matching rate of the above comparison results to determine the referenceability of the machine system output result in this feedback; if the content output by the system for this category is understood and recognized by the user, it will be weighted in the subsequent standards and comprehensively calculated together with the referenceability submitted in history; when the referenceability exceeds the preset threshold, it indicates that the output result has a high reference value for this category, and the corresponding algorithm can be connected to other human-computer interaction links for further optimization; S5.4: Complete the evaluation of the current system interaction: According to the comparison results, complete the comprehensive evaluation of the current machine system interaction: in the case of multiple output results, compare the referenceability of each output result, and select the result with higher referenceability as the evaluation scheme for this interaction, input it into the system and return the final optimized result.

9. The method for optimizing the interaction of a human-machine system based on intelligent evaluation of feedback positivity according to claim 8, characterized in that, In the process of interactive feedback of the human-machine system, the obtained interactive content is determined based on the quantitative analysis of all user evaluations and through weighted calculation. The specific steps are as follows: S5.4.1: Generate an initial reference value and conduct weighted comparison: In a machine system interaction, for the same output result, different users will give their respective reference values; weight and sum up all users' reference values and conduct comparison and analysis with the parameter requirements; the analysis results usually fall into multiple aspects, each aspect containing positive and negative values. By performing weighted calculation on these values, a comprehensive preliminary result is obtained; S5.4.2: Positive and Negative Value Statistics and Weighting: During the weighting process, the reference values are classified into positive and negative categories according to the user category, and statistics and comparison are carried out separately; then different weights are applied to these positive and negative reference values, and the proportions of positive and negative values are calculated comprehensively; S5.4.3: Forming a Vector and Determining the Final Reference Value: Based on the positive and negative value statistics results, a vector with positive and negative values as the main axes is constructed; according to the component sizes of the vector, the final reference value of the output result is determined; S5.4.4: Average Processing of Multiple Interaction Output Results: If multiple reference values are generated during multiple human-machine system interactions, they can be averaged to obtain a more representative final reference value.

10. The method for optimizing the interaction of a human-machine system based on intelligent evaluation of feedback enthusiasm according to claim 9, characterized in that, By collecting the basic information of the user and the hardware information of the hardware device logged in by the user, user classification and identification, as well as comparison of emotional characteristics are carried out; Based on its basic information, the evaluation of emotional characteristics is completed, and the emotional characteristics of the target population are classified and identified; The step S6 includes the following steps: S6.1: Classification and Identification of the Set of Historical User Emotional Evaluation Matrices: The set of historical user emotional evaluation matrices is classified and identified, and identification is completed; after completion, the unique identifier of the set of historical user emotional evaluation matrices is obtained, as well as the time dimension corresponding to this identifier; S6.2: Analysis of the Matching Degree between the Historical Emotional Evaluation Matrix and the Current Interaction: Through comparison, the set of historical user emotional evaluation matrices that meet the requirements is retrieved; after obtaining it, the performance of different user categories after entering the human-machine interaction system is analyzed for matching degree with the above matrix set; according to the artificially set time identifier, the reference values in the feedback evaluation before the interaction are introduced and statistically analyzed, and finally the matching degree between the current user and the set of historical user emotional evaluation matrices is determined; S6.3: Confirming the Correlation Time Point between the Interaction Process and the Reference Value: In the human-machine interaction system, the time point corresponding to the reference value of the user's feedback evaluation is judged, that is, which type of machine the user first interacts with during the interaction process, the order of obtaining information, and the corresponding source of the reference value; S6.4: Necessity Comparison: The reference values of the feedback evaluation are compared for necessity to judge the priority in the current interaction requirements; S6.5: Making an Introduction Decision Based on the Comparison Result: According to the comparison result, the human-machine interaction system is guided to select appropriate introduction content for the user in this interaction; S6.6: System Interaction Evaluation and Content Output: When the human-machine interaction system obtains the corresponding interaction evaluation result for this interaction process, based on the effectiveness of the further analysis of this result, corresponding parameters are generated or updated according to the evaluation result, so as to obtain the final interaction content after evaluation.

Images