AI intelligent dialogue system based on virtual digital human

Abstract

The invention discloses an AI intelligent dialogue system based on a virtual digital human, which comprises a user dialogue acquisition module, a user dialogue identification module, an identification matching analysis module and a dialogue output analysis module, relates to the technical field of intelligent interaction, and solves the problems of seamless connection and unified processing of voice information and text information. According to the system, voice information is reliably converted and processed in a user dialogue recognition module, so that the compatibility and processing efficiency of the whole system on different types of dialogue input are improved, the dialogue information in the voice mode and the dialogue information in the text mode are processed in a unified mode, and the dialogue information processing efficiency and accuracy are improved. The voice information can be accurately converted into the text for subsequent processing, meanwhile, information highly related to user dialogue can be more accurately screened out from the database through a multi-dimensional and multi-level matching calculation mode, and the accuracy and practicability of a matching result are improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent interaction technology, and in particular to an AI intelligent dialogue system based on virtual digital humans. Background Art

[0002] With the rapid development of artificial intelligence technology, intelligent dialogue systems have been widely used in many fields. As an emerging human-computer interaction interface, virtual digital humans provide users with a more vivid and intuitive interactive experience.

[0003] According to the publication number CN116662511A, an intelligent dialogue scene analysis method and an AI dialogue system using AI model recognition are provided. The service interaction dialogue text generated by the target user in the current intelligent dialogue scene is obtained, and the service interaction dialogue text is mined by using an AI model recognition strategy to generate at least one interest dialogue sentence clustering cluster, and the interest tag corresponding to the interest dialogue sentence cluster with a number of interest dialogue sentences greater than a set number is determined as a target interest tag. Based on the target interest tag, the corresponding dialogue feedback page content is loaded to the intelligent dialogue scene page where the target user is located, thereby combining the number of interest dialogue sentences with different interest tags in the intelligent dialogue scene to mine the interest of the target user. Compared with the related art that only focuses on the mined interest tag itself, the accuracy and effectiveness of interest mining can be improved, thereby improving the matching degree with the target user in the subsequent dialogue feedback process.

[0004] However, existing intelligent dialogue systems often face many challenges when processing user dialogue information. On the one hand, the fusion processing of voice information and text information is not perfect, and it is difficult to efficiently convert voice into text and perform deep semantic analysis. On the other hand, when matching user dialogue content with database information, there is a lack of accurate and efficient matching algorithms, resulting in low accuracy and relevance of matching results. At the same time, in the process of displaying matching results to users, the display method is relatively single and cannot well meet the needs of users in different scenarios. For example, the layout is not friendly enough when displaying text, and there is a lack of effective interaction with the digital human image when displaying voice. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides an AI intelligent dialogue system based on virtual digital humans, which solves the problem of seamless connection and unified processing of voice information and text information, and improves the efficiency and accuracy of dialogue information processing.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: an AI intelligent dialogue system based on virtual digital human, comprising: A user dialogue collection module is used to obtain dialogue information of interactive users and transmit the obtained dialogue information to the user dialogue recognition module; The user conversation recognition module is used to recognize the acquired conversation information to obtain the conversation content, convert the conversation content into text, and perform word segmentation processing according to the conversation content to obtain pre-processed content, and then transmit the pre-processed content to the recognition matching analysis module; The identification and matching module is used to analyze the obtained pre-processed content, match the content in the database to obtain the matching result, calculate the content matching value and the content association value corresponding to the matching result, and filter the results to be analyzed and the pre-selected matching results according to the content matching value and the content association value, and then transmit the two to the comprehensive analysis and processing module; A comprehensive analysis processing module is used to analyze the obtained results to be analyzed and the pre-selected matching results, calculate the matching values ​​of the two with the dialogue information, and filter them according to the matching values ​​to obtain the dialogue display results, and then transmit the dialogue display results to the dialogue output analysis module; The dialogue output analysis module is used to display the dialogue display results to the corresponding interactive users based on the acquired real-time data.

[0007] As a further solution of the present invention, the specific manner in which the user dialogue recognition module obtains the preprocessed content is: Then, the acquired dialogue information is converted into text, and the converted dialogue content is recognized and processed. The subject, predicate and object are used as the key sentence basis to split the dialogue content and obtain multiple groups of sentence words. These multiple groups of sentence words are semantically recognized and simplified to obtain simplified information. The acquired simplified information is integrated and classified with the multiple groups of sentence words to form preprocessed content, and finally the preprocessed content is transmitted to the recognition and matching module.

[0008] As a further solution of the present invention, the specific manner in which the recognition and matching module analyzes the preprocessed content is: Acquire preprocessed content and a database, and match sentence words in the preprocessed content with the database; Match the entire sentence word with the database, and label the matching result as the first matching result a, where a=1, 2, ..., n, where n represents the number of the first matching results; The simplified information in the preprocessed content is matched with the database, and the matching method is: the simplified information is obtained, and the simplified information as a whole is matched with the database, and the matching result is labeled and recorded as the second matching result b, and b=1, 2, ..., m, where m represents the quantity label corresponding to the second matching result.

[0009] As a further solution of the present invention, the specific method in which the identification and matching module calculates the content matching value corresponding to the matching result is: Get the first matching result a, and take one group of data in it as the analysis object, then get the text data in the analysis object that is the same as the sentence word, and label it as o, where o=1, 2, ..., p, where p represents the type of text data, and at the same time get the quantity corresponding to the text data o as Lo, then calculate the total quantity of text data o as L1o, and calculate the proportion of the total quantity of text data o L1o as Zo, sum the calculated proportions of all text data o, and calculate the mean of the sum of the proportions, and use the calculated mean as the content matching value of the analysis object.

[0010] As a further solution of the present invention, the specific method in which the identification and matching module calculates the content association value corresponding to the matching result is: Obtain all the second matching results b, and take one set of data as the target object, identify the semantics of the target object, and simplify the semantics of the target object to obtain simplified semantics, then calculate the similarity value between the simplified semantics and the simplified information, and so on, calculate the similarity values ​​corresponding to all the second matching results b.

[0011] As a further solution of the present invention, the specific manner in which the identification and matching module obtains the results to be analyzed and the pre-selected matching results according to the content matching value and the content association value is: Compare the content matching value with the threshold value. If the content matching value is greater than the threshold value, the corresponding first matching result is recorded as the result to be analyzed. Otherwise, if the content matching value is less than the threshold value, the corresponding first matching result is recorded as the eliminated matching result. The similarity value of the second matching result b is compared with the preset value, and the second matching result corresponding to the similarity value greater than the preset value is recorded as the pre-selected matching result, and the second matching result corresponding to the similarity value less than the preset value is recorded as the eliminated matching result, and the eliminated results are not processed.

[0012] As a further solution of the present invention, the specific manner in which the comprehensive analysis processing module analyzes the results to be analyzed and the pre-selected matching results is: Obtain the result to be analyzed and the pre-selected matching result, and combine the two to obtain a combined matching result and label it as r, where r=1, 2, ..., e, where e represents the number of combined matching results, and the combined matching result here includes all the results of the result to be analyzed and the pre-selected matching result. Then, obtain any combined matching result as the target object, and segment the target object to obtain multiple sentences, and calculate the similarity between the multiple sentences and the dialogue information respectively, then calculate the sum of all similarities, and calculate the corresponding similarity mean; Similarly, all combination matching results are processed, and the corresponding similarity mean is calculated, and the similarity mean is compared with the preset similarity to obtain the dialogue display result.

[0013] As a further solution of the present invention, the comprehensive analysis and processing module compares the similarity mean with the preset similarity to obtain the dialogue display result in the following specific manner: Filter the combination matching results whose average similarity value is greater than the preset similarity value and mark them as dialogue display results.

[0014] The present invention provides an AI intelligent dialogue system based on virtual digital human. Compared with the prior art, it has the following beneficial effects: The present invention reliably converts and processes voice information in the user dialogue recognition module, which helps to improve the compatibility and processing efficiency of the entire system for different types of dialogue inputs, and uniformly processes dialogue information in both voice and text modes. Voice information can be accurately converted into text and subsequently processed, thereby achieving seamless connection of multimodal information. At the same time, through a multi-dimensional and multi-level matching calculation method, information that is highly relevant to user dialogues can be more accurately screened out from the database, thereby improving the accuracy and practicality of the matching results. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a block diagram of the system principle of the present invention. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] For example, see Figure 1 The present application provides an AI intelligent dialogue system based on virtual digital human, including a user dialogue acquisition module, a user dialogue recognition module, a recognition matching analysis module and a dialogue output analysis module, and at the same time, in combination with the attached Figure 1 It can be known that the above functional units are electrically connected in a unidirectional manner.

[0018] First, build a virtual digital human, and determine the style of the virtual digital human according to the application scenario and target audience. If used in a game scenario, the style may be fantasy, realistic or cartoon-like; if used in customer service, education and other fields, the image is usually more realistic, friendly and professional. For example, a customer service digital human designed for a financial institution may have a well-dressed professional suit, a simple hairstyle, and a gentle and professional facial expression.

[0019] The user dialogue collection module is used to obtain the dialogue information of the interactive users and transmit the obtained dialogue information to the user dialogue recognition module. The dialogue information here includes text dialogue information and voice dialogue information. The specific text dialogue information is obtained through the corresponding text input box, and the voice dialogue information is obtained through the corresponding voice input button.

[0020] The user conversation recognition module is used to identify the acquired conversation information to obtain the conversation content, convert the conversation content into text, and perform word segmentation processing based on the conversation content to obtain pre-processed content, and then transmit the pre-processed content to the recognition matching analysis module.

[0021] First, the conversation information of the interactive user is obtained, and the conversation information is identified and judged. If the conversation information is voice conversation information, it is classified as voice information; if the conversation information is text conversation information, it is classified as text information; Then, the acquired dialogue information is converted into text, and here the conversion is mainly performed for voice information. If the input is text information, no conversion is required, and the converted dialogue content is recognized and processed, with the subject, predicate and object as the sentence points, and the dialogue content is divided into sentences according to the sentence points to obtain multiple groups of sentence words. For example, when the user asks the intelligent dialogue system through voice, "I want to know the functions of the new mobile phone", the system will automatically obtain the subject, predicate and object in the sentence, specifically the subject "I", the predicates include "want" and "know about it", the objects "new mobile phone", "of" and "function", and the multiple groups of sentence words obtained are shown above. Then, the semantics of the obtained multiple groups of sentence words are recognized, and the multiple groups of sentence words are simplified to obtain simplified information. For example, in the above example, the obtained multiple groups of sentence words are simplified to obtain "new mobile phone functions", and the obtained simplified information and multiple groups of sentence words are sorted to obtain preprocessed content, and then the preprocessed content is transmitted to the recognition and matching module.

[0022] An identification and matching module is used to analyze the acquired pre-processed content, match it with the content of the database to obtain the matching result, calculate the content matching value and content association value corresponding to the matching result, and filter the results to be analyzed and the pre-selected matching results according to the content matching value and content association value, and then transmit the two to the comprehensive analysis and processing module.

[0023] The preprocessed content and the database are obtained, and the clause words in the preprocessed content are matched with the database, and the matching method is: the clause words are matched with the database as a whole, and the matching results are labeled and recorded as the first matching result a, and a=1, 2, ..., n, where n represents the number label corresponding to the first matching result. For example, if the clause words in the preprocessed content contain "mobile phone", the database is searched for information that completely matches it. If information such as parameters and sales data of the mobile phone is found, they are marked as a1, a2, etc. respectively, and the simplified information in the preprocessed content is matched with the database, and the matching method is: the simplified information is obtained, and the simplified information is matched with the database as a whole, and the matching results are labeled and recorded as the second matching result b, and b=1, 2, ..., m, where m represents the number label corresponding to the second matching result; Obtain a first matching result a and a second matching result b, and simultaneously obtain corresponding clause words and simplified information, and then calculate a content matching value between the first matching result a and the clause words and a content association value between the second matching result b and the simplified information; The specific calculation method for the content matching value is as follows: obtain the first matching result a, and take one group of data as the analysis object, then obtain the text data that is the same as the sentence word in the analysis object, and the text data here represents the corresponding text, and is labeled as o, and o=1, 2, ..., p, where p represents the type of text data, and at the same time obtain the number corresponding to the text data o and record it as Lo, then calculate the total number of text data o and record it as L1o, and calculate the number ratio of the total number of text data o L1o and record it as Zo, and the number ratio here is obtained by dividing the total number of text data by the total number of analysis objects, summing up the calculated number ratios of all text data o, and calculating the mean of the sum of the number ratios, and using the calculated mean as the content matching value of the analysis object; For example, a group of analysis objects in the first matching result a is "XXX mobile phone has a high-definition screen, a powerful processor and a large-capacity battery, the appearance design of XXX mobile phone is very exquisite, and the operating system of XXX mobile phone is smooth", where the sentence word is "XXX mobile phone", and the text data type p=3 (respectively "high-definition screen", "powerful processor", and "large-capacity battery"). Assume that "high-definition screen" appears once, that is, L1 (high-definition screen) = 1; "powerful processor" appears once, that is, L1 (powerful processor) = 1; "large-capacity battery" appears once, that is, L1 ( Large-capacity battery) = 1, the total number of analysis objects is assumed to be 20 words (this is a simple example, and the actual calculation can be based on more precise text units), then the proportion of "high-definition screen" Z1 = 1 / 20 = 0.05; the proportion of "powerful processor" Z2 = 1 / 20 = 0.05; the proportion of "large-capacity battery" Z3 = 1 / 20 = 0.05, the sum of the proportions is 0.05+0.05+0.05=0.15, and the average is 0.15 / 3=0.05, and this 0.05 is the content matching value of the analysis object.

[0024] Similarly, the content matching values ​​of all first matching results a are calculated and sorted from large to small according to the content matching values. Then, the content matching value is compared with a threshold value, and the specific value of the threshold value is set by the operator. If the content matching value is greater than the threshold value, the corresponding first matching result is recorded as a result to be analyzed. Conversely, if the content matching value is less than the threshold value, the corresponding first matching result is recorded as a discarded matching result. The specific calculation method of the content association value is as follows: obtain all the second matching results b, and take one set of data as the target object, identify the semantics of the target object, and simplify the semantics of the target object to obtain the simplified semantics. Then, the similarity value between the simplified semantics and the simplified information is calculated. The calculation of the similarity value here is mainly reflected by the cosine similarity. First, the simplified semantics and the simplified information are converted into vectors through the pre-trained word vector model, and the vector modulus corresponding to the two is calculated at the same time, and according to the formula The cosine similarity between the two is calculated, and similarity values ​​corresponding to all second matching results b are calculated in the same way; The similarity value of the second matching result b is compared with a preset value, and the specific value of the preset value is set by the operator. The second matching result corresponding to the similarity value greater than the preset value is recorded as the pre-selected matching result, and the second matching result corresponding to the similarity value less than the preset value is recorded as the eliminated matching result, and the eliminated result is not processed.

[0025] A comprehensive analysis and processing module is used to analyze the obtained results to be analyzed and the pre-selected matching results, calculate the matching values ​​of the two with the dialogue information, and filter them according to the matching values ​​to obtain the dialogue display results, and then transmit the dialogue display results to the dialogue output analysis module.

[0026] Obtain the result to be analyzed and the pre-selected matching result, and combine the two to obtain a combined matching result and label it as r, and r=1, 2, ..., e, where e represents the number of combined matching results, and the combined matching result here includes all the results of the result to be analyzed and the pre-selected matching results, then obtain any combined matching result as the target object, and perform sentence segmentation processing on the target object to obtain multiple sentences (then, select any combined matching result as the target object, and perform sentence segmentation operation on the target object according to punctuation marks (such as "," ";" ".", etc.) to obtain multiple sentences), and the segmentation here is segmentation according to punctuation marks, such as "," ";" or "." For example, for the combined matching result The result is "This phone has a high configuration, its processor has strong performance, and its camera pixel is also very good." According to the punctuation, it can be divided into three sentences: "This phone has a high configuration", "Its processor has strong performance", and "And its camera pixel is also very good". The similarities between the multiple sentences and the conversation information are calculated respectively, and then the sum of all similarities is calculated, and the corresponding similarity mean is calculated. Assuming that the similarities between the above three sentences and the original conversation information are 0.3, 0.4, and 0.2 respectively, then their similarity sum is 0.3+0.4+0.2=0.9, and the similarity mean is 0.9÷3=0.3. By analogy, all the combination matching results are processed and the corresponding similarity mean is calculated; The calculated similarity mean is compared with a preset similarity value, and the value of the preset similarity value is set by the operator. The combination matching results corresponding to the similarity mean greater than the preset similarity value are screened and marked as dialogue display results, and the dialogue display results are transmitted to the dialogue output analysis module.

[0027] The dialogue output analysis module is used to display the dialogue display results to the corresponding interactive users based on the acquired real-time data.

[0028] Specifically, the real-time data is mainly the conversation mode corresponding to the interactive users, and the conversation mode includes language conversation, text conversation and the combination of language and text. In this embodiment, it is displayed in the form of text display. The specific display mode is as follows: Text display Dialogue window design: Design a dedicated dialog window in the web page or mobile application interface. The location of the dialog window should be convenient for users to see, usually in the middle or bottom of the screen. You can adjust the window's size, color, transparency and other properties to coordinate with the overall interface style. For example, in a simple style application, the dialog window can be white and translucent with a light gray border.

[0029] Distinguish the text input by the user and the text replied by the digital person. You can distinguish them by different font colors, font sizes or indentation methods. For example, the text input by the user is displayed in black font, and the text replied by the digital person is displayed in blue font. The text replied by the digital person can be appropriately indented, making it easier for users to distinguish the two parties in the conversation.

[0030] Text formatting and layout: For longer text content, appropriate segmentation and line breaks are performed to improve readability. For example, when a digital person responds to a complex product introduction, different product features are introduced in paragraphs to avoid a large paragraph of text causing reading pressure on users.

[0031] Supports hyperlink display. If the content of the digital human's reply involves relevant web resources (such as product links, knowledge reference links, etc.), the hyperlink can be displayed in the text in a clickable form. For example, when the digital human recommends a product, a hyperlink is added to the product name, and the user can jump directly to the product details page after clicking it.

[0032] Embodiment 2, as Embodiment 2 of the present invention, is implemented on the basis of Embodiment 1, and differs from Embodiment 1 in that: This embodiment displays the dialogue results by means of language display, and the specific display method is as follows: Speech Visualization: Add voice visualization elements such as volume icons or audio waveforms to the conversation window. When the digital person speaks, the volume icon can dynamically display the volume, and the audio waveform can show the rhythm and intensity changes of the voice, allowing users to intuitively feel the state of the voice.

[0033] Synchronize with the digital human's lip movements. Ensure that the opening and closing of the digital human's mouth shape matches the speech content when speaking. This can be achieved by setting the synchronization relationship between speech and lip shape during the animation production stage. For example, when the digital human makes the sound of a vowel, the mouth shape should show a corresponding opening shape.

[0034] Voice prompts and controls: Provide a voice play / pause button to facilitate users to control voice playback. For example, when the user is in a noisy environment or does not want to listen to the voice temporarily, they can pause the voice playback and view the text content instead.

[0035] Display the voice progress bar to let users know the progress of voice playback. The progress bar can help users quickly locate a certain position in the voice. For example, if the user wants to listen to a key point just said by the digital human again, the progress bar can be used to quickly jump to the corresponding position.

[0036] Embodiment 3, as the embodiment 3 of the present invention, focuses on combining the implementation processes of embodiment 1 and embodiment 2 for implementation.

[0037] At the same time, the virtual digital human image display in this application includes the following different ways: If it is a 3D digital human, the digital human image can be displayed in all directions through operations such as rotation, scaling, and translation. For example, when the digital human first appears, it can be rotated slowly to show the user its overall image. During the conversation, the posture and angle of the digital human can be appropriately adjusted according to the content and emotions of the conversation to enhance the visual appeal.

[0038] Use lighting effects to enhance the realism of 3D digital humans. For example, add appropriate ambient light, point light or spotlight to the digital human to simulate real lighting scenes and make the details of the digital human's skin and clothing more vivid.

[0039] For 2D digital humans, dynamic effects can be displayed by switching different expressions and action frames. A series of animation sequences can be created, such as nodding, blinking, smiling, etc., and these animations can be played regularly during the conversation or according to the content of the conversation. For example, when the digital human agrees with the user's point of view, a nodding animation is played.

[0040] Using layered animation technology, different parts of the digital human (such as the head, body, limbs, etc.) are animated separately and then combined together to make the movements of the 2D digital human more natural and smooth.

[0041] Some of the data in the above formulas are dimensionless and numerically calculated. Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0042] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. An AI intelligent dialogue system based on virtual digital human, characterized in that: include: A user dialogue collection module is used to obtain dialogue information of interactive users and transmit the obtained dialogue information to the user dialogue recognition module; The user conversation recognition module is used to recognize the acquired conversation information to obtain the conversation content, convert the conversation content into text, and perform word segmentation processing according to the conversation content to obtain pre-processed content, and then transmit the pre-processed content to the recognition matching analysis module; The identification and matching module is used to analyze the obtained pre-processed content, match the content in the database to obtain the matching result, calculate the content matching value and the content association value corresponding to the matching result, and filter the results to be analyzed and the pre-selected matching results according to the content matching value and the content association value, and then transmit the two to the comprehensive analysis and processing module; A comprehensive analysis processing module is used to analyze the obtained results to be analyzed and the pre-selected matching results, calculate the matching values ​​of the two with the dialogue information, and filter them according to the matching values ​​to obtain the dialogue display results, and then transmit the dialogue display results to the dialogue output analysis module; The dialogue output analysis module is used to display the dialogue display results to the corresponding interactive users based on the acquired real-time data.

2. The AI ​​intelligent dialogue system based on virtual digital human according to claim 1, characterized in that: The specific method for the user dialogue recognition module to obtain the preprocessed content is: Then, the acquired dialogue information is converted into text, and the converted dialogue content is recognized and processed. The subject, predicate and object are used as the key sentence basis to split the dialogue content and obtain multiple groups of sentence words. These multiple groups of sentence words are semantically recognized and simplified to obtain simplified information. The acquired simplified information is integrated and classified with the multiple groups of sentence words to form preprocessed content, and finally the preprocessed content is transmitted to the recognition and matching module.

3. The AI ​​intelligent dialogue system based on virtual digital human according to claim 1, characterized in that: The specific method in which the recognition and matching module analyzes the preprocessed content is as follows: Acquire preprocessed content and a database, and match sentence words in the preprocessed content with the database; Match the entire sentence word with the database, and label the matching result as the first matching result a, where a=1, 2, ..., n, where n represents the number of the first matching results; The simplified information in the preprocessed content is matched with the database, and the matching method is: the simplified information is obtained, and the simplified information as a whole is matched with the database, and the matching result is labeled and recorded as the second matching result b, and b=1, 2, ..., m, where m represents the quantity label corresponding to the second matching result.

4. The AI ​​intelligent dialogue system based on virtual digital human according to claim 1, characterized in that: The specific method in which the identification and matching module calculates the content matching value corresponding to the matching result is: Get the first matching result a, and take one group of data in it as the analysis object, then get the text data in the analysis object that is the same as the sentence word, and label it as o, where o=1, 2, ..., p, where p represents the type of text data, and at the same time get the quantity corresponding to the text data o as Lo, then calculate the total quantity of text data o as L1o, and calculate the proportion of the total quantity of text data o L1o as Zo, sum the calculated proportions of all text data o, and calculate the mean of the sum of the proportions, and use the calculated mean as the content matching value of the analysis object.

5. The AI ​​intelligent dialogue system based on virtual digital human according to claim 1, characterized in that: The specific method in which the identification and matching module calculates the content association value corresponding to the matching result is: Obtain all the second matching results b, and take one set of data as the target object, identify the semantics of the target object, and simplify the semantics of the target object to obtain simplified semantics, then calculate the similarity value between the simplified semantics and the simplified information, and so on, calculate the similarity values ​​corresponding to all the second matching results b.

6. The AI ​​intelligent dialogue system based on virtual digital human according to claim 1, characterized in that: The specific method of the identification and matching module to obtain the results to be analyzed and the pre-selected matching results according to the content matching value and the content association value is: Compare the content matching value with the threshold value. If the content matching value is greater than the threshold value, the corresponding first matching result is recorded as the result to be analyzed. Otherwise, if the content matching value is less than the threshold value, the corresponding first matching result is recorded as the eliminated matching result. The similarity value of the second matching result b is compared with the preset value, and the second matching result corresponding to the similarity value greater than the preset value is recorded as the pre-selected matching result, and the second matching result corresponding to the similarity value less than the preset value is recorded as the eliminated matching result, and the eliminated results are not processed.

7. The AI ​​intelligent dialogue system based on virtual digital human according to claim 1, characterized in that: The specific way in which the comprehensive analysis processing module analyzes the results to be analyzed and the pre-selected matching results is as follows: Obtain the result to be analyzed and the pre-selected matching result, and combine the two to obtain a combined matching result and label it as r, where r=1, 2, ..., e, where e represents the number of combined matching results, and the combined matching result here includes all the results of the result to be analyzed and the pre-selected matching result. Then, obtain any combined matching result as the target object, and segment the target object to obtain multiple sentences, and calculate the similarity between the multiple sentences and the dialogue information respectively, then calculate the sum of all similarities, and calculate the corresponding similarity mean; Similarly, all combination matching results are processed, and the corresponding similarity mean is calculated, and the similarity mean is compared with the preset similarity to obtain the dialogue display result.

8. The AI ​​intelligent dialogue system based on virtual digital human according to claim 7, characterized in that: The specific method in which the comprehensive analysis and processing module compares the similarity mean with the preset similarity to obtain the dialogue display result is: Filter the combination matching results whose average similarity value is greater than the preset similarity value and mark them as dialogue display results.

Images