Intelligent dialogue virtual robot interaction system and method based on AI artificial intelligence

Abstract

The invention relates to the technical field of intelligent interaction, and particularly discloses an intelligent dialogue virtual robot interaction system and method based on AI artificial intelligence, and the system comprises a dialogue data interpretation module which is used for extracting the latest multi-dimensional expression state features and core content texts of a user from dialogue data input by the user; the weight matrix building module is used for building an evaluation assignment conversion matrix of the latest multi-dimensional expression state characteristics of the user; the weight vector determination module is used for generating influence weight vectors of all latest single-dimensional expression state characteristics of the user based on the evaluation assignment conversion matrix of the latest multi-dimensional expression state characteristics of the user; the intelligent dialogue response module is used for obtaining response information of dialogue data input by the user based on the influence weight vector and the core content text of all latest single-dimensional expression state characteristics of the user and an AI dialogue response model; and the response effect of the intelligent dialogue virtual robot is improved.

Description

Technical Field

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

[0002] An AI-powered Conversational Virtual Robot is a natural language processing tool implemented through artificial intelligence technology. It can simulate human conversations and communicate naturally with humans. This robot can understand and generate conversations by learning human language patterns and semantics, so that the robot can better understand and respond to human questions or requests, improving the efficiency and quality of human-computer interaction. An AI-powered Conversational Virtual Robot can help businesses and individuals communicate more effectively with customers, employees, and the public. For example, intelligent customer service robots can provide assistance to customers in call centers, websites, and social media; intelligent assistants can provide users with information and services on smart speakers and smartphones; and intelligent chatbots can provide users with shopping and consulting advice and other services on online platforms.

[0003] However, there are also some challenges for intelligent conversational virtual robots. For example, how to ensure that the robot can correctly understand the meaning of the user's input dialogue does not only rely on high-precision natural language understanding, but also often requires consideration of the language emotions of the dialogue outputter. However, the existing intelligent conversational virtual robot interaction technology is still lacking in natural language sentiment analysis, which leads to the need to improve the response effect of intelligent conversational virtual robots.

[0004] Therefore, the present invention proposes an intelligent dialogue virtual robot interaction system and method based on AI artificial intelligence. Summary of the invention

[0005] The present invention provides an AI-based intelligent dialogue virtual robot interaction system and method, which is used to analyze the multi-dimensional expression state characteristics of dialogue data input by the user based on AI, obtain the influence weight vector of all the user's latest single-dimensional expression state characteristics, and based on the influence weight vector of all the user's latest single-dimensional expression state characteristics and core content text, use artificial intelligence model technology to greatly improve the response effect of the intelligent dialogue virtual robot.

[0006] The present invention provides an AI-based intelligent dialogue virtual robot interaction system, comprising: The dialogue data interpretation module is used to extract the user's latest multi-dimensional expression status features and core content text from the dialogue data input by the user; The weight matrix building module is used to build the evaluation assignment conversion matrix of the user's latest multi-dimensional expression status characteristics; A weight vector determination module, used to generate an influence weight vector of all the latest single-dimensional expression state features of the user based on the evaluation assignment conversion matrix of the latest multi-dimensional expression state features of the user; The conversational intelligent response module is used to obtain the response information of the intelligent conversational virtual robot based on the influence weight vector and core content text of all the user's latest single-dimensional expression status features and the AI ​​conversational response model.

[0007] Preferably, the dialogue data interpretation module includes: The expression state feature extraction submodule is used to extract the user's latest multi-dimensional expression state features from the conversation data input by the user; The core content text extraction submodule is used to extract the user's latest core content text from the conversation data input by the user.

[0008] Preferably, the expression state feature extraction submodule includes: A feature category determination unit, used to determine the data type of the conversation data input by the user, and determine all expression state feature classes corresponding to the conversation data input by the user based on a preset conversation data type-multidimensional expression state feature category list; The state feature extraction unit is used to input the conversation data input by the user into the expression state feature extraction model corresponding to each expression state feature class corresponding to the conversation data, so as to obtain the user's latest multi-dimensional expression state feature.

[0009] Preferably, the weight matrix building module includes: The feature evaluation and assignment submodule is used to evaluate and assign a value to each single-dimensional expression state feature in the user's latest multi-dimensional expression state feature, and obtain the assignment result of each single-dimensional expression state feature of the user; The weight matrix construction submodule is used to build an evaluation assignment conversion matrix of the user's latest multi-dimensional expression state features based on the assignment results of all the user's latest single-dimensional expression state features, wherein the total number of rows and columns of elements in the evaluation assignment conversion matrix is ​​equal to the total number of dimensions of all the user's latest single-dimensional expression state features.

[0010] Preferably, the feature evaluation and assignment submodule includes: An emotional tendency judgment unit, used to judge the emotional tendency of each single-dimensional expression state feature in the user's latest multi-dimensional expression state feature, wherein the emotional tendency includes positive and negative; The tendency degree evaluation unit is used to input each single-dimensional expression state feature in the user's latest multi-dimensional expression state feature into the emotional tendency degree evaluation model corresponding to the emotional tendency, and obtain the emotional tendency degree evaluation value of each single-dimensional expression state feature in the user's latest multi-dimensional expression state feature; The expression state feature assignment unit is used to obtain the latest assignment result of each single-dimensional expression state feature of the user based on the emotional tendency degree evaluation value of each single-dimensional expression state feature in the user's latest multi-dimensional expression state feature.

[0011] Preferably, the expression state feature assignment unit comprises: The first feature assignment subunit is used to, when the emotion tendency of the single-dimensional expression state feature in the latest multi-dimensional expression state feature of the user is positive, use the emotion tendency degree evaluation value of the corresponding single-dimensional expression state feature as the assignment result of the latest corresponding single-dimensional expression state feature of the user; The second feature assignment sub-unit is used to take the negative value of the emotional tendency evaluation value of the corresponding single-dimensional expression state feature as the assignment result of the user's latest corresponding single-dimensional expression state feature when the emotional tendency of the single-dimensional expression state feature in the user's latest multi-dimensional expression state feature is negative.

[0012] Preferably, the weight matrix building submodule includes: The assignment result sorting unit is used to arbitrarily sort the assignment results of all the latest single-dimensional expression state features of the user to obtain the ordinal value of each single-dimensional expression state feature; A first element determination unit is used to determine the row ordinal number corresponding to the ordinal value of the first single-dimensional expression state feature and the column ordinal number corresponding to the ordinal value of the second single-dimensional expression state feature as 1 in the evaluation assignment conversion matrix when the assignment result of the first single-dimensional expression state feature among the latest assignment results of any two single-dimensional expression state features of the user is greater than the assignment result of the second single-dimensional expression state feature; The second element determination unit is used to determine the row ordinal number corresponding to the ordinal value of the first single-dimensional expression state feature and the column ordinal number corresponding to the ordinal value of the second single-dimensional expression state feature as 0 in the evaluation assignment conversion matrix when the assignment result of the first single-dimensional expression state feature among the latest assignment results of any two single-dimensional expression state features of the user is less than the assignment result of the second single-dimensional expression state feature; The third element determination unit is used to determine the row ordinal number corresponding to the ordinal value of the first single-dimensional expression state feature and the column ordinal number corresponding to the ordinal value of the second single-dimensional expression state feature as 0.5 in the evaluation assignment conversion matrix when the assignment result of the first single-dimensional expression state feature among the latest assignment results of any two single-dimensional expression state features of the user is equal to the assignment result of the second single-dimensional expression state feature; The fourth element determination unit is used to determine the elements at all positions where the row ordinal numbers and the column ordinal numbers in the evaluation assignment conversion matrix are the same as each other as 0.5.

[0013] Preferably, the weight vector determination module includes: The positive and negative iteration count submodule is used to treat each column vector in the evaluation assignment conversion matrix as the initial vector of each single-dimensional expression state feature of the user, and count the total number of elements not less than 0.5 in the initial vector of each single-dimensional expression state feature as the positive iteration number of each single-dimensional expression state feature, and count the total number of elements less than 0.5 in the initial vector of each single-dimensional expression state feature as the negative iteration number of each single-dimensional expression state feature; The influence weight vector generation submodule is used to generate the influence weight vector of all the user's latest single-dimensional expression state features based on the positive iteration number and negative iteration number of all single-dimensional expression state features and the initial vector of all single-dimensional expression state features: In the formula, The total number of dimensions that express the user's latest single-dimensional status features. For users of the latest The first of the initial vectors expressing the state features in a single dimension elements, For users of the latest The number of positive iterations of a single dimension to express state features, For users of the latest The number of negative iterations of a single dimension expressing the state feature.

[0014] Preferably, the dialogue intelligent response module includes: The influence weight determination submodule is used to perform dot multiplication calculations on the influence weight vectors of all the latest single-dimensional expression state features of the user and the initial vectors of each single-dimensional expression state feature of the evaluation assignment conversion matrix to obtain the influence weight of each single-dimensional expression state feature; The AI ​​model response submodule is used to input the influence weights and core content texts of all the user's latest single-dimensional expression status features into the AI ​​dialogue response model to obtain the response information of the intelligent dialogue virtual robot.

[0015] The present invention provides an AI-based intelligent dialogue virtual robot interaction method, which is applied to any of the above AI-based intelligent dialogue virtual robot interaction systems, including: S1: Extract the user's latest multi-dimensional expression status features and core content text from the conversation data input by the user; S2: Build an evaluation assignment conversion matrix of the user's latest multi-dimensional expression status characteristics; S3: Generate the influence weight vector of all the latest single-dimensional expression state features of the user based on the evaluation assignment conversion matrix of the user's latest multi-dimensional expression state features; S4: Based on the influence weight vectors of all the latest single-dimensional expression status features of the user and the core content text and the AI ​​dialogue response model, the response information of the intelligent dialogue virtual robot is obtained.

[0016] The beneficial effects of the present invention compared to the prior art are as follows: based on AI artificial intelligence, the multi-dimensional expression state characteristics of the dialogue data input by the user are analyzed to obtain the influence weight vector of all the user's latest single-dimensional expression state characteristics, and based on the influence weight vector of all the user's latest single-dimensional expression state characteristics and the core content text, the artificial intelligence model technology is used to greatly improve the response effect of the intelligent dialogue virtual robot.

[0017] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures specifically pointed out in this application document.

[0018] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 Schematic diagram of the functional modules of the AI-based intelligent conversational virtual robot interaction system in an embodiment of the present invention; Figure 2 Schematic diagram of an intelligent conversation virtual robot interaction system based on AI artificial intelligence in an embodiment of the present invention; Figure 3The present invention is a flowchart of an intelligent conversational virtual robot interaction method based on AI artificial intelligence in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0021] Example 1 The present invention provides an intelligent dialogue virtual robot interaction system based on AI artificial intelligence. Figure 1 and 2 ,include: The dialogue data interpretation module is used to extract the user's latest multi-dimensional expression status features and core content text from the dialogue data input by the user; The weight matrix building module is used to build the evaluation assignment conversion matrix of the user's latest multi-dimensional expression status characteristics; A weight vector determination module, used to generate an influence weight vector of all the latest single-dimensional expression state features of the user based on the evaluation assignment conversion matrix of the latest multi-dimensional expression state features of the user; The conversational intelligent response module is used to obtain the response information of the intelligent conversational virtual robot based on the influence weight vector and core content text of all the user's latest single-dimensional expression status features and the AI ​​conversational response model.

[0022] In this embodiment, the conversation data exists in various forms, such as conversation text data, conversation recording data, conversation audio data, etc.

[0023] In this embodiment, the multi-dimensional expression state feature is a specific feature under all single-dimensional expression state feature categories corresponding to the data type corresponding to the conversation data most recently input by the user.

[0024] In this embodiment, the core content text is the text representing the core expression meaning of the user in the conversation data.

[0025] In this embodiment, the single-dimensional expression state feature is a feature of any one of the dialogue expression states, such as the user's emotion or emotional tendency, the vocal expression state, the body movement expression state, and the language expression habit state, reflected in the dialogue data, for example: emotional expression state feature (whose specific expression features include: anger, happiness, helplessness, confusion, excitement, etc.), emotional expression state feature (whose specific expression features include: high loudness, high pitch, etc.), body movement expression state feature (whose specific expression features include: exaggerated action, no body movement, appropriate body movement, etc.), language expression habit state (whose specific expression features include: rising tone at the end of the sentence, ending the sentence in the form of a question, etc.); In this embodiment, the AI ​​dialogue response model is a pre-trained model that can output reasonable response information based on the influence weights and core content texts of all the latest single-dimensional expression state features of the input user. The training process of the model uses the influence weights and core content texts of all the single-dimensional expression state features in a large amount of dialogue data (as the input in the training process) and the response information for the corresponding dialogue data given by the existing intelligent dialogue virtual robot or manually (as the output of the training process) as training samples for training; The response information may be in various data forms, such as text, audio, video, etc.

[0026] In this embodiment, the response information of the intelligent conversation virtual robot is the response information that the intelligent conversation virtual robot finally gives to the user in response to the most recently input conversation data.

[0027] The beneficial effects of the above technology are: based on AI artificial intelligence, the multi-dimensional expression state characteristics of the dialogue data input by the user are analyzed to obtain the influence weight vector of all the user's latest single-dimensional expression state characteristics, and based on the influence weight vector of all the user's latest single-dimensional expression state characteristics and the core content text, the artificial intelligence model technology is used to greatly improve the response effect of the intelligent dialogue virtual robot.

[0028] Example 2 Based on Example 1, the dialogue data interpretation module includes: The expression state feature extraction submodule is used to extract the user's latest multi-dimensional expression state features from the conversation data input by the user; The core content text extraction submodule is used to extract the user's latest core content text from the conversation data entered by the user by using the knowledge extraction technology in the existing technology (for example, RBC constructed by limited N-gram rules, N-gram rules are n ordered elements that are adjacent to each other in sequential text, and for example, it can be semi-supervised learning, which combines the process of manual labeling and model automatic learning. The system can be trained based on manually provided examples).

[0029] The beneficial effect of the above technology is: completing the process of extracting the user's latest multi-dimensional expression status features and core content text from the conversation data.

[0030] Example 3 Based on Example 2, the expression state feature extraction submodule includes: A feature category determination unit, used to determine the data type of the conversation data input by the user, and determine all expression state feature classes corresponding to the conversation data input by the user based on a preset conversation data type-multidimensional expression state feature category list; The state feature extraction unit is used to input the conversation data input by the user into the expression state feature extraction model corresponding to each expression state feature class corresponding to the conversation data, so as to obtain the user's latest multi-dimensional expression state feature.

[0031] In this embodiment, the data types of the conversation data include: text data, voice data, video data, etc.

[0032] In this embodiment, the preset conversation data type-multi-dimensional expression state feature category list is a pre-prepared list of multi-dimensional expression state feature categories corresponding to each conversation data type.

[0033] In this embodiment, all expression state feature classes corresponding to the conversation data input by the user are the conversation data type corresponding to the conversation data input by the user, and all single-dimensional expression state feature types contained in the corresponding multi-dimensional expression state data category in the preset conversation data type-multi-dimensional expression state feature category list.

[0034] In this embodiment, the expression state feature extraction model corresponding to each expression state feature class is a pre-trained model used to extract specific features of the corresponding type of expression state feature class in the conversation data. The model is trained using a large amount of conversation data (as input in the training process) and manually determined specific features of the corresponding type of expression state feature class contained therein (as output in the training process) as training samples.

[0035] The beneficial effects of the above technology are: determining all single-dimensional expression state feature types corresponding to the dialogue data type corresponding to the dialogue data input by the user, and extracting specific features of all single-dimensional expression state feature types in the dialogue data.

[0036] Example 4 Based on Example 1, the weight matrix building module includes: The feature evaluation and assignment submodule is used to evaluate and assign a value to each single-dimensional expression state feature in the user's latest multi-dimensional expression state feature, and obtain the assignment result of each single-dimensional expression state feature of the user; The weight matrix construction submodule is used to build an evaluation assignment conversion matrix of the user's latest multi-dimensional expression state features based on the assignment results of all the user's latest single-dimensional expression state features, wherein the total number of rows and columns of elements in the evaluation assignment conversion matrix is ​​equal to the total number of dimensions of all the user's latest single-dimensional expression state features.

[0037] The beneficial effect of the above technology is: by evaluating and assigning a value to each single-dimensional expression state feature in the user's latest multi-dimensional expression state feature, an evaluation assignment conversion matrix of the user's latest multi-dimensional expression state feature is constructed.

[0038] Example 5 Based on Example 4, the feature evaluation and assignment submodule includes: An emotional tendency judgment unit, used to judge the emotional tendency of each single-dimensional expression state feature in the user's latest multi-dimensional expression state feature, wherein the emotional tendency includes positive and negative; The tendency degree evaluation unit is used to input each single-dimensional expression state feature in the user's latest multi-dimensional expression state feature into the emotional tendency degree evaluation model corresponding to the emotional tendency, and obtain the emotional tendency degree evaluation value of each single-dimensional expression state feature in the user's latest multi-dimensional expression state feature; The expression state feature assignment unit is used to obtain the latest assignment result of each single-dimensional expression state feature of the user based on the emotional tendency degree evaluation value of each single-dimensional expression state feature in the user's latest multi-dimensional expression state feature.

[0039] In this embodiment, determining the emotional tendency of each single-dimensional expression state feature in the user's latest multi-dimensional expression state feature includes: By retrieving the preset specific feature-emotional tendency list of each single-dimensional expression state feature, the emotional tendency of each single-dimensional expression state feature in the user's latest multi-dimensional expression state feature is determined.

[0040] In this embodiment, the emotional tendency degree assessment model of emotional tendency is obtained by pre-training using a large number of specific features of multiple single-dimensional expression state characteristics (as input in the training process) and the emotional tendency degree of the corresponding emotional tendency represented by the specific features of the corresponding single-dimensional expression state characteristics manually assessed (as output in the training process) as training samples.

[0041] In this embodiment, the emotional tendency evaluation value of each single-dimensional expression state feature represents the degree to which the single-dimensional expression state feature represents the corresponding emotional tendency.

[0042] The beneficial effects of the above technology are: determining the emotional tendency of each single-dimensional expression state feature in the user's latest multi-dimensional expression state feature, and evaluating the emotional tendency degree evaluation value of each single-dimensional expression state feature in the user's latest multi-dimensional expression state feature, based on which a reasonable assignment of each of the user's latest single-dimensional expression state features is achieved.

[0043] Example 6 On the basis of Example 5, the expression state feature assignment unit includes: The first feature assignment subunit is used to, when the emotion tendency of a single-dimensional expression state feature in the latest multi-dimensional expression state feature of the user is positive, use the emotion tendency degree evaluation value of the corresponding single-dimensional expression state feature as the assignment result of the latest corresponding single-dimensional expression state feature of the user; The second feature assignment sub-unit is used to take the negative value of the emotional tendency evaluation value of the corresponding single-dimensional expression state feature as the assignment result of the user's latest corresponding single-dimensional expression state feature when the emotional tendency of the single-dimensional expression state feature in the user's latest multi-dimensional expression state feature is negative.

[0044] The beneficial effect of the above technology is: based on the emotional tendency type and emotional tendency degree evaluation value of the single-dimensional expression state feature in the user's latest multi-dimensional expression state feature, the assignment result of each single-dimensional expression state feature of the user's latest can simultaneously reflect the emotional tendency type and degree.

[0045] Example 7 Based on Example 4, the weight matrix building submodule includes: The assignment result sorting unit is used to arbitrarily sort the assignment results of all the latest single-dimensional expression state features of the user to obtain the ordinal value of each single-dimensional expression state feature; A first element determination unit is used to determine the row ordinal number corresponding to the ordinal value of the first single-dimensional expression state feature and the column ordinal number corresponding to the ordinal value of the second single-dimensional expression state feature as 1 in the evaluation assignment conversion matrix when the assignment result of the first single-dimensional expression state feature among the latest assignment results of any two single-dimensional expression state features of the user is greater than the assignment result of the second single-dimensional expression state feature; The second element determination unit is used to determine the row ordinal number corresponding to the ordinal value of the first single-dimensional expression state feature and the column ordinal number corresponding to the ordinal value of the second single-dimensional expression state feature as 0 in the evaluation assignment conversion matrix when the assignment result of the first single-dimensional expression state feature among the latest assignment results of any two single-dimensional expression state features of the user is less than the assignment result of the second single-dimensional expression state feature; The third element determination unit is used to determine the row ordinal number corresponding to the ordinal value of the first single-dimensional expression state feature and the column ordinal number corresponding to the ordinal value of the second single-dimensional expression state feature as 0.5 in the evaluation assignment conversion matrix when the assignment result of the first single-dimensional expression state feature among the latest assignment results of any two single-dimensional expression state features of the user is equal to the assignment result of the second single-dimensional expression state feature; The fourth element determination unit is used to determine the elements at all positions where the row ordinal numbers and the column ordinal numbers in the evaluation assignment conversion matrix are the same as each other as 0.5.

[0046] In this embodiment, the first single-dimensional expression state feature and the second single-dimensional expression state feature are two single-dimensional expression state features among any two latest single-dimensional expression state features of the user.

[0047] The beneficial effect of the above technology is that by comparing the assignment results of any two of the user's latest single-dimensional expression state features, the constructed evaluation assignment conversion matrix can more prominently represent the type and degree of emotional tendencies in the user's conversation data.

[0048] Example 8 Based on Embodiment 1, the weight vector determination module includes: The positive and negative iteration count submodule is used to treat each column vector in the evaluation assignment conversion matrix as the initial vector of each single-dimensional expression state feature of the user, and count the total number of elements not less than 0.5 in the initial vector of each single-dimensional expression state feature as the positive iteration number of each single-dimensional expression state feature, and count the total number of elements less than 0.5 in the initial vector of each single-dimensional expression state feature as the negative iteration number of each single-dimensional expression state feature; The influence weight vector generation submodule is used to generate the influence weight vector of all the user's latest single-dimensional expression state features based on the positive iteration number and negative iteration number of all single-dimensional expression state features and the initial vector of all single-dimensional expression state features: In the formula, The total number of dimensions that express the user's latest single-dimensional status features. For users of the latest The first of the initial vectors expressing the state features in a single dimension elements, For users of the latest The number of positive iterations of a single dimension to express state features, For users of the latest The number of negative iterations of a single dimension expressing the state feature.

[0049] The beneficial effects of the above technology are: by determining the number of positive iterations and negative iterations based on all single-dimensional expression state features and the initial vectors of all single-dimensional expression state features, the influence weight vector of all the user's latest single-dimensional expression state features is reasonably generated, and the degree of influence of the emotional tendency represented by each element in the initial vector of each of the user's latest single-dimensional expression state features is accurately quantified.

[0050] Example 9 Based on Example 1, the conversational intelligent response module includes: The influence weight determination submodule is used to perform dot multiplication calculations on the influence weight vectors of all the latest single-dimensional expression state features of the user and the initial vectors of each single-dimensional expression state feature of the evaluation assignment conversion matrix to obtain the influence weight of each single-dimensional expression state feature; The AI ​​model response submodule is used to input the influence weights and core content texts of all the user's latest single-dimensional expression status features into the AI ​​dialogue response model to obtain the response information of the intelligent dialogue virtual robot.

[0051] The beneficial effects of the above technology are: the influence weight of each single-dimensional expression state feature is accurately calculated, and based on the influence weights of all the user's latest single-dimensional expression state features and the core content text and the AI ​​dialogue response model, it is achieved by fully considering the types and degrees of user emotional tendencies expressed in user dialogue data, thereby greatly improving the response effect of the intelligent dialogue virtual robot.

[0052] Embodiment 10: The present invention provides an AI-based intelligent dialogue virtual robot interaction method, which is applied to any of the AI-based intelligent dialogue virtual robot interaction systems described in embodiments 1 to 9, with reference to Figure 3 ,include: S1: Extract the user's latest multi-dimensional expression status features and core content text from the conversation data input by the user; S2: Build an evaluation assignment conversion matrix of the user's latest multi-dimensional expression status characteristics; S3: Generate the influence weight vector of all the latest single-dimensional expression state features of the user based on the evaluation assignment conversion matrix of the user's latest multi-dimensional expression state features; S4: Based on the influence weight vectors of all the latest single-dimensional expression status features of the user and the core content text and the AI ​​dialogue response model, the response information of the intelligent dialogue virtual robot is obtained.

[0053] The beneficial effects of the above technology are: based on AI artificial intelligence, the multi-dimensional expression state characteristics of the dialogue data input by the user are analyzed to obtain the influence weight vector of all the user's latest single-dimensional expression state characteristics, and based on the influence weight vector of all the user's latest single-dimensional expression state characteristics and the core content text, the artificial intelligence model technology is used to greatly improve the response effect of the intelligent dialogue virtual robot.

[0054] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. An intelligent dialogue virtual robot interaction system based on AI artificial intelligence, characterized by: include: The dialogue data interpretation module is used to extract the user's latest multi-dimensional expression status features and core content text from the dialogue data input by the user; The weight matrix building module is used to build the evaluation assignment conversion matrix of the user's latest multi-dimensional expression status characteristics; A weight vector determination module, used to generate an influence weight vector of all the latest single-dimensional expression state features of the user based on the evaluation assignment conversion matrix of the latest multi-dimensional expression state features of the user; The conversational intelligent response module is used to obtain the response information of the intelligent conversational virtual robot based on the influence weight vector and core content text of all the user's latest single-dimensional expression status features and the AI ​​conversational response model.

2. The AI-based intelligent dialogue virtual robot interaction system according to claim 1 is characterized in that: Dialogue data interpretation module, including: The expression state feature extraction submodule is used to extract the user's latest multi-dimensional expression state features from the conversation data input by the user; The core content text extraction submodule is used to extract the user's latest core content text from the conversation data input by the user.

3. The AI-based intelligent dialogue virtual robot interaction system according to claim 2 is characterized in that: The expression state feature extraction submodule includes: A feature category determination unit, used to determine the data type of the conversation data input by the user, and determine all expression state feature classes corresponding to the conversation data input by the user based on a preset conversation data type-multidimensional expression state feature category list; The state feature extraction unit is used to input the conversation data input by the user into the expression state feature extraction model corresponding to each expression state feature class corresponding to the conversation data, so as to obtain the user's latest multi-dimensional expression state feature.

4. The AI-based intelligent dialogue virtual robot interaction system according to claim 1 is characterized in that: Weight matrix building module, including: The feature evaluation and assignment submodule is used to evaluate and assign a value to each single-dimensional expression state feature in the user's latest multi-dimensional expression state feature, and obtain the assignment result of each single-dimensional expression state feature of the user; The weight matrix construction submodule is used to build an evaluation assignment conversion matrix of the user's latest multi-dimensional expression state features based on the assignment results of all the user's latest single-dimensional expression state features, wherein the total number of rows and columns of elements in the evaluation assignment conversion matrix is ​​equal to the total number of dimensions of all the user's latest single-dimensional expression state features.

5. The AI-based intelligent dialogue virtual robot interaction system according to claim 4 is characterized in that: Feature evaluation and assignment submodule, including: An emotional tendency judgment unit, used to judge the emotional tendency of each single-dimensional expression state feature in the user's latest multi-dimensional expression state feature, wherein the emotional tendency includes positive and negative; The tendency degree evaluation unit is used to input each single-dimensional expression state feature in the user's latest multi-dimensional expression state feature into the emotional tendency degree evaluation model corresponding to the emotional tendency, and obtain the emotional tendency degree evaluation value of each single-dimensional expression state feature in the user's latest multi-dimensional expression state feature; The expression state feature assignment unit is used to obtain the latest assignment result of each single-dimensional expression state feature of the user based on the emotional tendency degree evaluation value of each single-dimensional expression state feature in the user's latest multi-dimensional expression state feature.

6. The AI-based intelligent dialogue virtual robot interaction system according to claim 5 is characterized in that: The expression state feature assignment unit includes: The first feature assignment subunit is used to, when the emotion tendency of a single-dimensional expression state feature in the latest multi-dimensional expression state feature of the user is positive, use the emotion tendency degree evaluation value of the corresponding single-dimensional expression state feature as the assignment result of the latest corresponding single-dimensional expression state feature of the user; The second feature assignment sub-unit is used to take the negative value of the emotional tendency evaluation value of the corresponding single-dimensional expression state feature as the assignment result of the user's latest corresponding single-dimensional expression state feature when the emotional tendency of the single-dimensional expression state feature in the user's latest multi-dimensional expression state feature is negative.

7. The AI-based intelligent dialogue virtual robot interaction system according to claim 4 is characterized in that: The weight matrix building submodule includes: The assignment result sorting unit is used to arbitrarily sort the assignment results of all the latest single-dimensional expression state features of the user to obtain the ordinal value of each single-dimensional expression state feature; A first element determination unit is used to determine the row ordinal number corresponding to the ordinal value of the first single-dimensional expression state feature and the column ordinal number corresponding to the ordinal value of the second single-dimensional expression state feature as 1 in the evaluation assignment conversion matrix when the assignment result of the first single-dimensional expression state feature among the latest assignment results of any two single-dimensional expression state features of the user is greater than the assignment result of the second single-dimensional expression state feature; The second element determination unit is used to determine the row ordinal number corresponding to the ordinal value of the first single-dimensional expression state feature and the column ordinal number corresponding to the ordinal value of the second single-dimensional expression state feature as 0 in the evaluation assignment conversion matrix when the assignment result of the first single-dimensional expression state feature among the latest assignment results of any two single-dimensional expression state features of the user is less than the assignment result of the second single-dimensional expression state feature; The third element determination unit is used to determine the row ordinal number corresponding to the ordinal value of the first single-dimensional expression state feature and the column ordinal number corresponding to the ordinal value of the second single-dimensional expression state feature as 0.5 in the evaluation assignment conversion matrix when the assignment result of the first single-dimensional expression state feature among the latest assignment results of any two single-dimensional expression state features of the user is equal to the assignment result of the second single-dimensional expression state feature; The fourth element determination unit is used to determine the elements at all positions where the row ordinal numbers and the column ordinal numbers in the evaluation assignment conversion matrix are the same as each other as 0.

5.

8. The AI-based intelligent dialogue virtual robot interaction system according to claim 1 is characterized in that: The weight vector determination module includes: The positive and negative iteration count submodule is used to treat each column vector in the evaluation assignment conversion matrix as the initial vector of each single-dimensional expression state feature of the user, and count the total number of elements not less than 0.5 in the initial vector of each single-dimensional expression state feature as the positive iteration number of each single-dimensional expression state feature, and count the total number of elements less than 0.5 in the initial vector of each single-dimensional expression state feature as the negative iteration number of each single-dimensional expression state feature; The influence weight vector generation submodule is used to generate the influence weight vector of all the user's latest single-dimensional expression state features based on the positive iteration number and negative iteration number of all single-dimensional expression state features and the initial vector of all single-dimensional expression state features: In the formula, The total number of dimensions that express the user's latest single-dimensional status features. For users of the latest The first of the initial vectors expressing the state features in a single dimension elements, For users of the latest The number of positive iterations of a single dimension to express state features, For users of the latest The number of negative iterations of a single dimension expressing the state feature.

9. The AI-based intelligent dialogue virtual robot interaction system according to claim 1 is characterized in that: The dialogue intelligent response module includes: The influence weight determination submodule is used to perform dot multiplication calculations on the influence weight vectors of all the latest single-dimensional expression state features of the user and the initial vectors of each single-dimensional expression state feature of the evaluation assignment conversion matrix to obtain the influence weight of each single-dimensional expression state feature; The AI ​​model response submodule is used to input the influence weights and core content texts of all the user's latest single-dimensional expression status features into the AI ​​dialogue response model to obtain the response information of the intelligent dialogue virtual robot.

10. An intelligent dialogue virtual robot interaction method based on AI artificial intelligence, characterized in that: The AI ​​artificial intelligence-based intelligent dialogue virtual robot interaction system applied to any one of claims 1 to 9 comprises: S1: Extract the user's latest multi-dimensional expression status features and core content text from the conversation data input by the user; S2: Build an evaluation assignment conversion matrix of the user's latest multi-dimensional expression status characteristics; S3: Generate the influence weight vector of all the latest single-dimensional expression state features of the user based on the evaluation assignment conversion matrix of the user's latest multi-dimensional expression state features; S4: Based on the influence weight vectors of all the latest single-dimensional expression status features of the user and the core content text and the AI ​​dialogue response model, the response information of the intelligent dialogue virtual robot is obtained.

Images