WO2019001127A1

WO2019001127A1 - Virtual character-based artificial intelligence interaction method and artificial intelligence interaction device

Info

Publication number: WO2019001127A1
Application number: PCT/CN2018/084879
Authority: WO
Inventors: 伏英娜; 金宇林; 雷宇
Original assignee: 迈吉客科技（北京）有限公司
Priority date: 2017-06-26
Filing date: 2018-04-27
Publication date: 2019-01-03
Also published as: CN107480122A; CN107480122B

Abstract

Provided by the present invention is a virtual character-based artificial intelligence interaction method that is used to solve the technical problem wherein the interactive effect for artificial intelligence in the role of an assistant is poor due to a lack of emotional characteristics in interaction information responses. The embodiments of the present invention comprise: acquiring, from a corpus, a corpus resource for a role and performing vectorization processing so as to form a role corpus for the role; by means of a neural network model, performing natural language semantic recognition on the role corpus to form a determined role, and forming an emotional characteristic for statements by the determined role, then forming a personality characteristic for the determined role. The virtual character-based artificial intelligence interaction method of the embodiments of the present application combines text words or emotional control information of an interaction information input, and, by means of a manner of natural semantic recognition, creates an artificial helper having a personality; the present invention reduces human involvement on a large-scale, reducing the workload in development while producing an artificial intelligence helper having more natural emotions. The present invention further comprises a virtual character-based artificial intelligence interaction device.

Description

Artificial character interaction method based on virtual character and artificial intelligence interaction device

The present invention claims priority from the applicant, the application date is June 26, 2017, the application number is CN2017104987389, and the application name is "an artificial intelligence interaction method and an artificial intelligence interaction device". The entire contents of the above application are hereby incorporated by reference in its entirety.

Technical field

The invention relates to the field of audio and video intelligent control, in particular to an artificial intelligence interaction method based on virtual characters and an artificial intelligence interaction device.

Background of the invention

The assistant product includes the process of interactive information input, interactive information response, and interactive information output. For example, Microsoft's Cortana assistant product and Apple's Siri assistant product use NLP technology (ie, natural language processing technology) to extract keywords from the interactive text in the voice text after obtaining the interactive input voice text through voice recognition technology. The keyword matching/fuzzy matching method retrieves the knowledge list (or knowledge base) to respond to the input interaction information, and responds to the knowledge content as a message, and the message reply is output as the interaction information in the form of voice or text.

In the interactive information response process, since the used knowledge list (or knowledge base) adopts formatting processing, only the necessary knowledge content is included, so that the interactive information output formed by the message reply generated by the same query lacks emotional features, resulting in interaction of the assistant product. The information output is dull and cannot reflect the context-related attributes in the interaction process, and the human-computer interaction experience is poor.

Summary of the invention

In view of this, the embodiments of the present invention provide an artificial intelligence interaction method based on a virtual character and an artificial intelligence interaction device, which are used to solve the technical problem that the interactive information response is poor due to lack of emotional features and artificial intelligence assisted characters.

The virtual character based artificial intelligence interaction method of the invention comprises:

The corpus resources of the character obtained from the corpus are vectorized to form the role corpus of the character;

The role corpus is subjected to natural language semantic recognition by the neural network model to form a determined role, and the sentence emotional characteristics of the determined character are formed to form the character trait of the determined character.

The avatar-based artificial intelligence interaction device of the invention comprises:

The corpus phrase generating module is configured to obtain a corpus resource of the character from the corpus and perform vectorization processing to form a role corpus of the character.

The semantic recognition module is configured to perform natural language semantic recognition on the character corpus through a neural network model to form a determined role, form a sentence emotional feature of the determined character, and form a character feature of the determined character.

a memory, a program code for storing a process of the avatar-based artificial intelligence interaction method as described above;

a processor for executing the program code.

The avatar-based artificial intelligence interaction method and the artificial intelligence interaction device according to the embodiment of the present invention create an artificial intelligence assistant with a character character through natural semantic recognition, thereby reducing manual participation on a large scale, reducing the workload of development, and at the same time Produce artificial intelligence assistants with more natural emotional effects.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the description of the prior art will be briefly described below. Obviously, the drawings in the following description are Some embodiments of the present invention may also be used to obtain other drawings based on these drawings without departing from the art.

FIG. 1 is a flowchart of a method for artificial intelligence interaction based on virtual characters according to an embodiment of the present invention.

FIG. 2 is a flowchart of determining a character, determining a character's personality, and determining a character's emotion formation based on a virtual person-based artificial intelligence interaction method according to an embodiment of the present invention.

FIG. 3 is a flowchart of an emotional interaction information output of an artificial human intelligent interaction method based on an avatar according to an embodiment of the present invention.

4 is a schematic structural diagram of an artificial intelligence interactive device or program module based on a virtual character.

Mode for carrying out the invention

The technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. It is a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts are protected by the present invention. range.

The step numbers in the drawings are only used as reference numerals for the steps, and do not indicate the order of execution.

FIG. 1 is a flowchart of a method for artificial intelligence interaction based on virtual characters according to an embodiment of the present invention. As shown in Figure 1, it includes:

Step 100: Obtain a corpus resource of the character from the corpus for vectorization processing to form a role corpus of the character.

The corpus resources of the character include literary works such as novels, scripts, poems, or lines that can be retrieved in the corpus. The carrier of literary works can be text or audio, and audio literary works can correspond to literary works that form words through speech recognition technology. The corpus of a character can be a literary work that includes a character, or a literary work that includes several characters. Several characters can have linguistic communication in the same literary work, or they can be quoted in different literary works.

The corpus of the character is vectorized, including the pre-processing of clauses, word segmentation, desensitization, etc., using, for example, word2vec (ie, text-to-vector) techniques to form the vectorization of all words in the corpus, and to maintain the order of all words.

Step 200: Perform natural language semantic recognition on the character corpus through the neural network model to form a determined role, form a sentence emotional characteristic of the determined character, and form a character characteristic for determining the character.

The neural network model for natural language semantic recognition of the role corpus may adopt a CNN (ie, convolutional neural network) model, an RNN (ie, a cyclic neural network) model, or a DNN (ie, deep neural network) model. The content of natural language semantic recognition includes the intrinsic connection of sentence order, the internal relationship of word order, the internal relationship between word frequency and emotional expression level, the internal relationship between sentence and role, the character's personality and the internal relationship between emotion and sentence structure. A specific quantitative internal emotional connection around a literary work that determines the role.

Step 300: Combine the sentence emotional feature of the determined role with the interactive information response by the emotional judgment input on the interaction information to form the emotional interaction information output.

The neural network model is used to extract the intrinsic contact information between the corpus from the role corpus, quantify the character of the character and form a specific linguistic structural feature, and cooperate with the traditional interactive information response to form an interactive information response with contextual emotion association expression. The mood and emotional expression of the interactive information output are adapted to the language mood change when the interactive information is input, and the knowledge content that is the response of the interactive information can be timely adjusted for the emotional change of the input subject in the interaction process, and the affinity of the intelligent assistant product is increased and improved. Language interaction experience.

The avatar-based artificial intelligence interaction method of the embodiment of the present invention creates an AI assistant with a character character through natural semantic recognition, reduces manual participation on a large scale, reduces the workload of development, and can produce a more natural emotional effect. Artificial Intelligence Assistant.

FIG. 2 is a flowchart of determining a character, determining a character's personality, and determining a character's emotion formation based on a virtual person-based artificial intelligence interaction method according to an embodiment of the present invention. Forming the determined roles as shown in FIG. 2 includes:

Step 210: Process the role corpus through the first neural network model, and output the corpus phrase.

The first neural network model in the embodiment of the present invention includes a CNN model and a softmax classifier (ie, a maximum class classifier), and the CNN model includes data formed by the input layer, the convolution layer, the pooling layer, and the fully connected layer CNN model. The softmax classifier forms all the corpus phrases contained in the character corpus. The CNN model can ensure that the intrinsic contact information between words in the role corpus is not lost in the classification process, and the softmax classifier can guarantee the discretization of the output data when forming the corpus.

Step 220: Perform a part-of-speech tag on the words in the corpus phrase.

Part of speech refers to the classification of word classes based on the characteristics of words. The words of modern Chinese can be divided into two categories of 12 parts of speech. One type is real words: nouns, verbs, adjectives, numerals, quantifiers and pronouns. One type is the function word: adverb, preposition, conjunction, auxiliary, interjection and onomatopoeia. The part-of-speech tagging algorithm can use the HanLP algorithm and the Jieba algorithm.

Step 230: Perform statistics on the corpus phrases according to the part-of-speech tag to form a determined role, and form a determined role list.

The part-of-speech tag can further reflect the natural language structure in the corpus phrase, that is, the attributes and connection structures of the subject, predicate, object, attribute, adverbial and complement. Through the statistics of the specific meanings of the subject, predicate and object, the determined role of the subject can be obtained and all the determined roles list can be formed.

As shown in FIG. 2, on the basis of the foregoing embodiment, the statement forming the statement emotional characteristics of the role of the embodiment of the present invention further includes:

The first neural network model in the embodiment of the present invention includes a CNN model and a softmax classifier (ie, a maximum class classifier), and the CNN model includes data formed by the input layer, the convolution layer, the pooling layer, and the fully connected layer CNN model. The softmax classifier (ie, the largest class classifier) forms all the corpus phrases contained in the character corpus. The CNN model can ensure that the intrinsic contact information between words in the role corpus is not lost in the classification process, and the softmax classifier can guarantee the discretization of the output data when forming the corpus.

Step 220: Perform a part-of-speech tag on the words in the corpus phrase.

Step 240: classify the corpus phrases according to the determined role, form a role corpus phrase for determining the role, and obtain the leading dialogue keyword from the role corpus phrase of the determined role.

Determining the role's role corpus phrase includes the content of the conversation associated with determining the role, excluding the corpus phrases that are unrelated to the determined role. The content of each of the determined roles can be completed based on the part-of-speech tag and the determined role list. The leading dialogue keyword in the role corpus phrase is a word, vocabulary or short sentence that is used to determine the character according to characteristics such as frequency.

Step 270: Combine the keyword (or word) list in the existing interaction information response with the leading conversation keyword of the role to form a list of the specific keywords that determine the role.

The embodiment of the present invention associates the response information in the existing interaction information response with the dialogue of the determined role through the proprietary keyword list, so that the response information and the feature information of the determined role can be uniformly processed and integrated.

Step 250: The role corpus phrase of the role is determined by the second neural network model processing, and the corpus phrase is formed into an emotional corpus phrase that determines the character according to the emotional feature, and forms a sentence emotional feature that determines the role.

The second neural network model in the embodiment of the present invention includes a CNN model and an SVM classifier (ie, a support vector machine classifier), and the CNN model includes data formed by the input layer, the convolution layer, the pooling layer, and the fully connected layer CNN model. The two SVM classifiers in series form emotional corpus phrases of different emotion categories. The SVM classifier can ensure accurate emotional classification using the intrinsic emotional connections formed by the second neural network model. For example, there are six basic classifications of emotions, fears and fears. The SVM classifier, as a supervised learning method, can classify extreme sorrow and love emotional corpus phrases through emotional dictionaries, and then form a hi-shock through the first SVM classifier classification. Emotional corpus phrases of fear and anger are classified into four types of emotional corpus phrases: hi, shock, fear and anger after being classified by the second SVM classifier.

As shown in FIG. 2, on the basis of the foregoing embodiment, the character characterization of determining the role of the embodiment of the present invention further includes:

Step 260: Perform emotional frequency statistics on the emotional corpus phrases of the determined characters to form a main character characteristic of the determined character.

Emotional dictionary is used to determine the emotional corpus of the character, and the emotional frequency and the frequency of the emotional words are used to form the emotional frequency histogram data, and the emotional frequency histogram data is input into the SVM classifier to obtain the character classification. For example, the classification of personality characteristics includes:

Type A personality is emotionally stable, socially adaptive and directional, but intellectual performance is general, subjective initiative is general, and communication ability is weak;

Type B personality has the characteristics of extroversion, emotional instability, poor social adaptability, impatience, and interpersonal relationship;

Type C personality has introverted characteristics, emotional stability, and good social adaptability, but it is passive in general;

The D-type personality has the characteristics of extroversion, social adaptability is good or general, interpersonal relationship is good, and organizational ability;

E-type personality has introverted characteristics, emotional instability, poor social adaptability or general, not good communication, but often good at independent thinking, research-oriented.

The most characteristic character classification of weights is used as the character trait of the determined character.

As shown in FIG. 2, on the basis of the foregoing embodiment, the statement emotional characteristics of determining the role in the embodiment of the present invention further includes:

Step 280: Perform statistics on the role corpus phrases of the determined role through the natural language structure, and form an emotional language structure corresponding to the sentence emotional characteristics of the determined role.

The embodiment of the present invention uses the natural language structure as a statistical basis to statistically determine the frequency of different language structures in the emotional corpus phrase of the character, and uses the high-frequency sub-linguistic structure as a language habit to determine the character under different emotional characteristics. Further, a dialog sample template for determining a role is formed according to the language habits under the emotional characteristics.

Step 290: Perform frequency statistics on the specific nouns and the specific noun contexts in the corpus, and form a corpus of related nouns associated with the specific nouns.

The determination of the specific noun in the corpus phrase refers to the keyword list in the existing interaction information response, the frequency statistics in the corresponding information content associated with the keyword list, the determination of the specific noun contextual association in the reference corpus phrase, and the association with the keyword list The grammatical association of statements and words in the corresponding information content. When a special noun is triggered during the interactive information input process of the assistant product, the associated noun and context associated with the specific noun are output as a response.

FIG. 3 is a flowchart of outputting emotional interaction information of a virtual character based artificial intelligence interaction method according to an embodiment of the present invention. As shown in FIG. 3, in the embodiment of the present invention, by combining the emotional judgment of the interaction information input, the determination of the statement's emotional characteristics of the role and the interaction information response include:

Step 310: Receive text vocabulary or emotion control information input by the interaction information.

Step 320: Extract the text vocabulary input by the interaction information, and determine the emotional characteristics of the character by matching and extracting the emotional corpus phrase of the determined character.

Step 350: Determine the emotional language structure of the determined character by determining the emotional characteristics of the character, and combine the leading conversation keyword in the role-specific keyword list with the standard response information in the existing interaction information response according to the emotional language structure.

In the embodiment of the present invention, the emotional corpus phrase is combined with the standard response information in the existing interactive information response, and the emotional expression factor of the role is determined for the standard response information.

As shown in FIG. 3, in the embodiment of the present invention, by combining the emotional judgment of the interaction information input, the determination of the statement's emotional characteristics of the role and the interaction information response include:

Step 340: Extract the text vocabulary input by the interaction information, and determine the specific noun related noun expression corpus of the character by matching with the specific noun in the corpus phrase.

Step 350: Determine the emotional language structure of the determined character by determining the emotional characteristics of the character, and combine the related related noun expression corpus of the specific noun of the determined character with the standard response information in the existing interactive information response according to the emotional language structure.

In the embodiment of the present invention, the specific noun related associative noun expression corpus is combined with the standard response information in the existing interactive information response, and the emotional expression factor of the character is determined by using the associated context in the corpus phrase of the determined role.

Step 330: Extract emotion control information input by the interaction information, match the determined role, and determine the emotional characteristics of the character.

In an embodiment of the invention, an additional step is included:

The emotional language structure of the determined character is obtained by determining the emotional characteristics of the character, and the related noun expression corpus of the specific noun of the determined character is combined with the standard response information in the existing interactive information response according to the emotional language structure.

Step 350 is replaced with an additional step.

In an embodiment of the invention, an additional step and step 350 may be included at the same time, the two steps being in the same parallel configuration in the logic process.

The embodiment of the present invention can continuously input the determined role and the determined emotion as the controllable weight information through the interaction information, so as to determine the degree of performance of the character in determining the emotional performance according to the user experience and the feeling of immediate adjustment. The feedback is formed according to the emotional response between the user and the virtual determined role, and the positive or negative feedback effect of determining the continuous emotional performance of the character is given, thereby further enhancing the human-computer interaction between the character and the user.

4 is a schematic structural diagram of an artificial intelligence interactive device or program module based on a virtual character. The avatar-based artificial intelligence interaction device or the program module deployed in the processor as shown in FIG. 4 includes:

The corpus phrase generating module 10 is configured to obtain a corpus resource of the character from the corpus and perform vectorization processing to form a role corpus of the character.

The semantic recognition module 20 is configured to perform natural language semantic recognition on the character corpus through the neural network model, form a determined role, form a sentence emotional feature that determines the character, and form a character characteristic that determines the character.

The emotion combining module 30 is configured to combine the statement emotional feature of the determined role with the interactive information response by the emotional judgment input on the interactive information to form the emotional interaction information output.

The semantic recognition module 20 includes:

The corpus phrase generating unit 21 is configured to process the role corpus through the first neural network model and output the corpus phrase.

The part of speech tagging unit 22 is configured to perform part-of-speech tagging on words in the corpus phrase.

The determining role generating unit 23 is configured to perform statistical determination on the corpus phrase according to the part-of-speech tag to form a determined role, and form a determined role list.

The role corpus phrase generating unit 24 is configured to classify the corpus phrases according to the determined characters, form a role corpus phrase for determining the role, and obtain the leading dialogue keyword from the role corpus phrase of the determined role.

The emotional corpus phrase generating unit 25 is configured to process the role corpus phrase of the role by the second neural network model processing, and form the corpus phrase to form the emotional corpus phrase of the determined character classified by the emotional feature to form a sentence emotional feature that determines the character.

The personality feature generating unit 26 is configured to perform emotional frequency statistics on the emotional corpus phrases of the determined characters to form a main character feature for determining the character.

The proprietary keyword list generating unit 27 is configured to combine the keyword (or word) list in the existing interactive information response with the leading dialogue keyword of the determined character to form a list of the specific keywords that determine the role.

The emotional language structure generating unit 28 is configured to perform statistics on the role corpus of the determined character through the natural language structure, and form an emotional language structure corresponding to the sentence emotional feature of the determined character.

The special noun generating unit 29 is configured to perform frequency statistics on the contextual terms of the specific nouns and the specific nouns in the corpus phrases, and form a corpus of associated nouns associated with the specific nouns.

The emotion combining module 30 includes:

The emotion information extracting unit 31 is configured to receive text vocabulary or emotion control information input by the interaction information.

The emotion feature recognition unit 32 is configured to extract a text vocabulary input by the interaction information, and determine an emotional feature of the character by matching and extracting the emotional corpus phrase of the character.

The emotion feature control unit 33 is configured to extract emotion control information input by the interaction information, match the determined role, and determine the emotional feature of the character.

The special noun association unit 34 is configured to extract the text vocabulary input by the interaction information, and determine the specific noun related noun expression corpus of the character by matching with the special noun in the corpus phrase.

The response information emotion generating unit 35 is configured to obtain an emotional language structure of the determined character by determining the emotional feature of the character, and determine the leading conversation keyword in the role list of the role and the standard response information in the existing interaction information response. The combination of emotional language structure.

In an embodiment of the present invention, the emotion combining module 30 further includes a second response information emotion generating unit, configured to obtain an emotional language structure of the determined character by determining an emotional feature of the character, and determine related related nouns of the specific noun of the determined character. The corpus and the standard response information in the existing interactive information response are combined according to the emotional language structure.

In an embodiment of the present invention, the second response information emotion generating unit may replace the response information emotion generating unit 35.

In an embodiment of the invention, the second response information emotion generating unit is disposed in parallel with the response information emotion generating unit 35.

The specific implementation and beneficial effects of the avatar-based artificial intelligence interaction device in the embodiment of the present invention can be referred to the avatar-based artificial intelligence interaction method, and details are not described herein again.

The avatar-based artificial intelligence interaction device of the embodiment of the invention includes:

a memory, a program code for storing a process of the virtual character-based artificial intelligence interaction method of the above embodiment;

A processor, program code for executing a process of the avatar-based artificial intelligence interaction method of the above embodiment.

Those of ordinary skill in the art will appreciate that the elements and algorithm steps of the various examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the solution. A person skilled in the art can use different methods for implementing the described functions for each particular application, but such implementation should not be considered to be beyond the scope of the present invention.

A person skilled in the art can clearly understand that for the convenience and brevity of the description, the specific working process of the system, the device and the unit described above can refer to the corresponding process in the foregoing method embodiment, and details are not described herein again.

In the several embodiments provided by the present application, it should be understood that the disclosed systems, devices, and methods may be implemented in other manners. For example, the device embodiments described above are merely illustrative. For example, the division of the unit is only a logical function division. In actual implementation, there may be another division manner, for example, multiple units or components may be combined or Can be integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, and may be in an electrical, mechanical or other form.

The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

The functions may be stored in a computer readable storage medium if implemented in the form of a software functional unit and sold or used as a standalone product. Based on such understanding, the technical solution of the present invention, which is essential or contributes to the prior art, or a part of the technical solution, may be embodied in the form of a software product, which is stored in a storage medium, including The instructions are used to cause a computer device (which may be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and the like, and can store a program check code. Medium.

The above is only the preferred embodiment of the present invention, and is not intended to limit the present invention. Any modifications, equivalent substitutions, etc., which are within the spirit and principles of the present invention, should be included in the scope of the present invention. within.

Industrial applicability

The artificial character intelligent interaction method and device based on the virtual character in the embodiment of the invention creates an artificial intelligence assistant with a character character through natural semantic recognition, reduces manual participation on a large scale, reduces the workload of development, and can generate more Artificial intelligence assistant for natural emotional effects. . Can be widely used in production and living environments.

Claims

An artificial intelligence interaction method based on virtual characters, including:

The corpus resources of the character obtained from the corpus are vectorized to form the role corpus of the character;

The character corpus is subjected to natural language semantic recognition by a neural network model to form a determined character, and the sentence emotional feature of the determined character is formed to form a character characteristic of the determined character.
The avatar-based artificial intelligence interaction method according to claim 1, wherein the forming the determined role comprises:

Processing the role corpus by the first neural network model, and outputting a corpus phrase;

Word-based tagging of words in the corpus;

The corpus phrase is statistically generated according to the part of speech tag to form the determined role.
The avatar-based artificial intelligence interaction method according to claim 2, wherein said forming the sentence emotional characteristics of the determined character comprises:

And classifying the corpus phrases according to the determined role, forming a role corpus phrase of the determined role, and obtaining a leading dialogue keyword from the role corpus phrase of the determined role;

Forming the list of the unique keywords of the determined role in combination with the keyword list in the existing interaction information response and the leading conversation keyword of the determined role.
The avatar-based artificial intelligence interaction method according to claim 3, wherein said forming the sentence emotional characteristics of the determined character comprises:

Processing the role corpus phrase of the determined character by a second neural network model, forming the corpus phrase to form an emotional corpus phrase of the determined character classified according to an emotional feature, forming the statement emotional feature of the determined character ;

The character corpus of the determined character is counted by a natural language structure to form an emotional language structure corresponding to the statement emotional feature of the determined character.
The avatar-based artificial intelligence interaction method according to claim 4, wherein the forming the character traits of the determined character comprises:

Performing emotional frequency statistics on the emotional corpus phrase of the determined character to form the personality trait of the determined character.
The avatar-based artificial intelligence interaction method according to claim 2, wherein said forming the sentence emotional characteristics of the determined character comprises:

Frequency statistics are performed on the contextual terms of the specific nouns and the specific nouns in the corpus, forming a related vocabulary corpus of the specific nouns.
The avatar-based artificial intelligence interaction method according to any one of claims 1 to 6, further comprising:

The emotional relationship of the determined character is combined with the interactive information response by the emotional judgment of the interactive information input to form an emotional interaction information output.
The avatar-based artificial intelligence interaction method according to claim 7, wherein the combining the statement emotion feature of the determined role with the interaction information response by the emotional judgment inputting the interaction information comprises:

Receiving text vocabulary or emotion control information input by interactive information;

Extracting a text vocabulary input by the interaction information, and extracting the emotional feature of the determined character by matching with the emotional corpus phrase of the determined character;

Obtaining the emotional language structure of the determined character by determining the emotional feature of the character, and determining the leading dialogue keyword in the private keyword list of the determined role and a criterion in an existing interaction information response Response information is combined by emotional language structure.
The avatar-based artificial intelligence interaction method according to claim 7, wherein the combining the statement emotion feature of the determined role with the interaction information response by the emotional judgment inputting the interaction information comprises:

Receiving text vocabulary or emotion control information input by interactive information;

Extracting a text vocabulary input by the interaction information, and extracting the emotional feature of the determined character by matching with the emotional corpus phrase of the determined character;

Extracting a text vocabulary input by the interaction information, and extracting related related noun expression corpus of the specific noun of the determined character by matching with the special noun in the corpus phrase;

Obtaining, by the determining the emotional feature of the character, the emotional language structure of the determined character, and determining the related related noun expression corpus of the specific noun of the determined character and the standard response information in the existing interactive information response according to the The combination of emotional language structure.
The avatar-based artificial intelligence interaction method according to claim 7, wherein the combining the statement emotion feature of the determined role with the interaction information response by the emotional judgment inputting the interaction information comprises:

Receiving text vocabulary or emotion control information input by interactive information;

Extracting emotion control information input by the interaction information, matching the determined role, and determining the emotional characteristics of the character;

Obtaining, by the determining the emotional feature of the character, the emotional language structure of the determined character, and using the leading conversation keyword in the specific keyword list of the determined role and the standard response information in the existing interaction information response according to the The combination of emotional language structure.
The avatar-based artificial intelligence interaction method according to any one of claims 8 to 10, wherein the determining the emotional language structure of the determined character by the emotional feature of the determined character, the determining The leading conversation keyword in the role's proprietary keyword list and the standard response information in the existing interaction information response are replaced by the emotional language structure in combination with:

Obtaining an emotional language structure of the determined character by determining the emotional feature of the character, and comparing the related noun expression corpus of the specific noun of the determined character with the standard response information in the existing interactive information response according to the emotional language Structural combination.
An artificial intelligence interaction device based on virtual characters, comprising:

a corpus phrase generating module, configured to obtain a corpus resource of a character from a corpus, perform vectorization processing, and form a role corpus of the character;

The semantic recognition module is configured to perform natural language semantic recognition on the character corpus through a neural network model to form a determined role, form a sentence emotional feature of the determined character, and form a character feature of the determined character.
The avatar-based artificial intelligence interaction device of claim 12, wherein the semantic recognition module comprises:

a corpus phrase generating unit, configured to process the role corpus by the first neural network model, and output a corpus phrase;

a part of speech tagging unit, configured to perform part-of-speech tagging on words in the corpus phrase;

Determining a role generating unit, configured to perform statistics on the corpus phrase according to the part of speech tag to form the determined role.
The avatar-based artificial intelligence interaction device according to claim 13, wherein the semantic recognition module further comprises: a role corpus phrase generating unit, configured to classify the corpus phrases according to the determined role, and form Determining a role corpus phrase of the character, and obtaining a leading conversation keyword from the role corpus phrase of the determined character;

The proprietary keyword list generating unit is configured to combine the keyword list in the existing interaction information response with the leading conversation keyword of the determined role to form the private keyword list of the determined role.
The avatar-based artificial intelligence interaction device of claim 14, wherein the semantic recognition module further comprises:

An emotional corpus phrase generating unit, configured to process the role corpus phrase of the determined character by using a second neural network model, and form the corpus vocabulary phrase to form an emotional corpus phrase of the determined character classified according to an emotional feature, to form the Determining the emotional characteristics of the statement of the character;

The emotional language structure generating unit is configured to perform statistics on the role corpus phrase of the determined character through a natural language structure, and form an emotional language structure corresponding to the sentence emotional feature of the determined character.
The AI-based artificial intelligence interaction device of claim 15, wherein the semantic recognition module further comprises:

The character feature generating unit is configured to perform emotional frequency statistics on the emotional corpus phrase of the determined character to form a personality trait of the determined character.
The AI-based artificial intelligence interaction device of claim 13, wherein the semantic recognition module further comprises:

The special noun generating unit is configured to perform frequency statistics on the specific nouns and the specific noun contextual associations in the corpus phrases, and form related related noun expression corpora of the special nouns.
The avatar-based artificial intelligence interaction device according to any one of claims 12 to 17, further comprising:

The emotion combining module is configured to combine the sentence emotional feature of the determined character with the interactive information response by using the emotional judgment of the interaction information input to form the emotional interaction information output.
The avatar-based artificial intelligence interaction device of claim 18, wherein the emotion combining module further comprises:

An emotion information extracting unit, configured to receive text vocabulary or emotion control information input by the interaction information;

An emotion feature recognition unit, configured to extract a text vocabulary input by the interaction information, and extract the emotional feature of the determined character by matching with the emotional corpus phrase of the determined character;

a response information emotion generating unit, configured to obtain, by using the emotional feature of the determined character, the emotional language structure of the determined character, and the leading dialogue keyword in the list of the unique keywords of the determined character The standard response information in the existing interactive information response is combined by the emotional language structure.
The avatar-based artificial intelligence interaction device of claim 18, wherein the emotion combining module further comprises:

An emotion information extracting unit, configured to receive text vocabulary or emotion control information input by the interaction information;

An emotion feature recognition unit, configured to extract a text vocabulary input by the interaction information, and extract the emotional feature of the determined character by matching with the emotional corpus phrase of the determined character;

a special noun association unit, configured to extract a text vocabulary input by the interaction information, and extract the related noun expression corpus of the specific noun of the determined character by matching with the special noun in the corpus phrase;

a response information emotion generating unit, configured to obtain the emotional language structure of the determined character by determining the emotional feature of the character, and correlate the related noun expression corpus of the specific noun of the determined character with an existing interactive information response The standard response information in the combination is combined according to the emotional language structure.
The avatar-based artificial intelligence interaction device of claim 18, wherein the emotion combining module further comprises:

An emotion information extracting unit, configured to receive text vocabulary or emotion control information input by the interaction information;

An emotion feature control unit, configured to extract emotion control information input by the interaction information, match the determined role, and determine the emotional feature of the character;

a response information emotion generating unit, configured to obtain, by using the emotional feature of the determined character, an emotional language structure of the determined character, and the leading dialogue keyword in the specified keyword list of the determined character and the existing interaction information The standard response information in the response is combined according to the emotional language structure.
The avatar-based artificial intelligence interactive device according to any one of claims 19 to 21, wherein the response information emotion generating unit is replaced by:

a second response information emotion generating unit, configured to obtain, by using the emotional feature of the determined character, the emotional language structure of the determined character, and the related related noun expression corpus of the specific noun of the determined character and the existing interactive information response The standard response information in the combination is combined according to the emotional language structure.
An artificial intelligence interaction device based on virtual characters, comprising:

a program, a program code for storing a process of the virtual character-based artificial intelligence interaction method according to any one of claims 1 to 11;

a processor for executing the program code.