KR102299563B1 - Method And Apparatus for Providing Untact Class by using Virtual Teaching Assistant Robot - Google Patents

Abstract

Disclosed are a method and a device for providing a non-face-to-face class using a virtual assistant robot. According to the present embodiment of the method and the device for providing a non-face-to-face class using a virtual assistant robot, in the non-face-to-face class, by using a virtual teaching assistant robot which complements the communication between instructors and learners, individual learning management, group learning management, school life support, and learning emotional care for both learners and instructors can be supported.

Description

Method And Apparatus for Providing Untact Class by using Virtual Teaching Assistant Robot

An embodiment of the present invention relates to a method and apparatus for providing a non-face-to-face class using a virtual assistant robot.

The content described below merely provides background information related to the present embodiment and does not constitute the prior art.

Network-based distance education has been used in schools as a means of coping with classroom education. Distance education does not limit the number of students participating, but allows students to attend classes at home or in an off-site space.

In general distance education, only limited dialogue between students and teachers is possible. For example, in a class provided online, conversation between class participants is impossible. When limited dialogue is possible, it is difficult to record the entire lesson, including dialogue (eg, chart messages) and video and audio programs and other comments inserted by the teacher. It causes complaints from students who miss one during class and those who wish to see all of the discussions that took place during class. It also causes complaints from students who want to revisit discussions related to a particular topic in class.

In addition, in order to respond to COVID-19, universities are avoiding group classes by attending school, and are conducting remote classes (real-time, recording, etc.) or classes using homework. In accordance with real-time lectures through ZOOM and WebEX, and recorded lectures through recording of class materials, both lecturers and learners are conducting non-face-to-face learning in accordance with the changing class format.

Commercial video lecture (conference) solutions have limitations in smooth transmission due to quality degradation and delays due to software, hardware, or network failures despite technological advancements. In addition, in terms of learning, there is a problem of causing confusion to both instructors and learners by not providing attendance processing function and indiscriminate use of unnecessary chatting or file sharing.
A conventional technique for performing education between a lecturer and a learner is disclosed in US Patent Publication No. US2014/0335497.

In this embodiment, in non-face-to-face class, individual learning management, group learning management, school life support, and learning emotion for both learners and lecturers by using a virtual teaching assistant robot that complements communication between the lecturer and the learner An object of the present invention is to provide a method and apparatus for providing a non-face-to-face class using a virtual assistant robot to support care.

According to an aspect of this embodiment, based on the lecture plan received from the lecturer terminal, a chat room opening unit for opening a non-face-to-face private chat room for each study task and a non-face-to-face group chat room for each task; In the non-face-to-face personal chat room and the non-face-to-face group chat room for each learning task, a virtual assistant robot is engaged together with a learner terminal, and the text information input from the non-face-to-face personal chat room and the non-face-to-face group chat room is obtained using the virtual assistant robot. a collecting unit to collect; an analysis unit generating an analysis result of analyzing the text information; a user pattern generator for generating a user pattern for each learner based on the analysis result; an emotion classification unit for classifying emotion information based on the analysis result; and a non-face-to-face class support unit configured to support a non-face-to-face class based on the analysis result, the learner-specific user pattern, and the emotional information using the virtual assistant robot. .

As described above, according to this embodiment, in the non-face-to-face class, individual learning management and group learning management for both the learner and the lecturer using a virtual teaching assistant robot that supplements communication between the lecturer and the learner , it has the effect of supporting school life support and learning emotional care.

1 is a view showing a non-face-to-face class system according to the present embodiment.
2 is a diagram schematically illustrating an apparatus for providing a non-face-to-face class according to the present embodiment.
3 is a diagram illustrating a learner group chatting during class using the virtual assistant robot according to the present embodiment.
4 is a diagram illustrating a study aid for query grouping and lecture reference materials using the virtual assistant robot according to the present embodiment.
5 is a view showing a learning aid for a group task using the virtual assistant robot according to the present embodiment.
6 is a diagram illustrating class-related activities and provision of quizzes using the virtual assistant robot according to the present embodiment.
7 is a view showing a screen for generating a class alarm to the lecturer using the virtual assistant robot according to the present embodiment.

Hereinafter, this embodiment will be described in detail with reference to the accompanying drawings.

1 is a view showing a non-face-to-face class system according to the present embodiment.

The non-face-to-face class system according to the present embodiment includes a non-face-to-face class providing device 110 , a learner terminal 130 , and a lecturer terminal 140 . Components included in the non-face-to-face class system are not necessarily limited thereto.

The non-face-to-face class providing device 110 includes a hardware module identical to that of a typical web server or network server. The non-face-to-face class providing apparatus 110 may be implemented in the form of a web server or a network server.

The non-face-to-face class providing device 110 generally communicates with an unspecified number of clients or other servers via an open computer network such as the Internet. The non-face-to-face class providing device 110 refers to a computer system and computer software (web server program) for deriving and providing a work result corresponding to a request for performing work from a client or other web server.

The non-face-to-face class providing device 110 should be understood as a broad concept including, in addition to the above-described web server program, a series of application programs operating on a web server or various databases built in the device.

The non-face-to-face class providing apparatus 110 receives lecture materials and lecture auxiliary materials from the lecturer terminal 140 , and then transmits the lecture materials and lecture auxiliary materials to the learner terminal 130 according to a learning schedule. The non-face-to-face class providing device 110 requests lecture materials from the lecturer terminal 140 according to the learning schedule. The non-face-to-face class providing device 110 receives lecture materials and lecture auxiliary data from the lecturer terminal 140 .

The non-face-to-face class providing device 110 provides lecture materials and lecture auxiliary materials to the learner terminal 130 for each individual or group of the learner terminal 130 according to a learning schedule. The non-face-to-face class providing device 110 conducts a lecture according to a learning schedule, and collects chat information for each lecture by using the virtual assistant robot 120 .

The non-face-to-face class providing device 110 analyzes the collected chat information for each lecture to generate a chat analysis result, and then extracts only the query content from the chat analysis result. The non-face-to-face class providing apparatus 110 identifies duplicate content among the query content, and generates query aggregate data obtained by combining the duplicate query content into a single query. The non-face-to-face class providing apparatus 110 selects sentences having a preset length or more from among the query collection data, and then generates query summary data summarized within a preset number of sentences or words. The non-face-to-face class providing device 110 transmits the query summary data to the lecturer terminal 140 using the virtual assistant robot 120 .

The non-face-to-face class providing device 110 analyzes chat information for each lecture. The non-face-to-face class providing device 110 extracts learners with a low participation rate or a different context of a conversation in the group chat based on the chat analysis result. The non-face-to-face class providing device 110 transmits a learning participation encouragement message to the learner terminal 130 corresponding to a learner having a low participation rate or a different context of a conversation by using the virtual assistant robot 120 for each lecture.

The non-face-to-face class providing device 110 analyzes chat information for each lecture (team room within the lecture). The non-face-to-face class providing device 110 extracts only out-of-class content from the chatting analysis results obtained by analyzing the collected chat information for each lecture (team group within the lecture). The non-face-to-face class providing device 110 selects a sentence requiring help or support by analyzing the content outside of class. The non-face-to-face class providing device 110 checks the content of a sentence requiring help or support. The non-face-to-face class providing device 110 selects a category corresponding to a sentence requiring help or support. The non-face-to-face class providing device 110 extracts extra-class support information (eg, part-time job information) corresponding to the corresponding category. The non-face-to-face class providing device 110 transmits support information outside of class to the learner terminal 130 in which a sentence requiring help or support is input using the virtual assistant robot 120 .

The non-face-to-face class providing device 110 analyzes chat information for each lecture (team room within the lecture). The non-face-to-face class providing device 110 generates a chat analysis result obtained by analyzing the collected chat information for each lecture (team room in the lecture). The non-face-to-face class providing device 110 selects an individual learner who matches each individual learning emotion (propensity) to form a new team group based on the chatting analysis result. The non-face-to-face class providing device 110 performs learning emotion care by transmitting a message for recommending a study group or recommending a chat group to individual learners who meet individual learning emotions (propensity) for group chatting.

The non-face-to-face class providing device 110 generates a user pattern for each learner based on the chat contents input from the learner terminal 130 in the group chat room for each task. The non-face-to-face class providing device 110 classifies the UNESCO-defined learning emotions by using the EKMAN model emotional classification based on the user pattern for each learner.

The non-face-to-face class providing device 110 recognizes the user characteristics/patterns of each member in the chat room based on the text input in the group chat room for each task, and learns by user characteristics/patterns based on the graph for building an emotional dictionary for each area categorize emotions.

The non-face-to-face class providing device 110 builds a domain emotional vocabulary list after measuring the importance and frequency of words for a specific domain called foreign language learning. The non-face-to-face class providing device 110 calculates a domain emotional index and builds a domain-customized emotional dictionary using a domain emotional vocabulary. The non-face-to-face class providing device 110 analyzes task participation, frequently used words, emojis, hashtags, and the like, based on conversations input from the learner terminals 130 in the group chat room for each task. The non-face-to-face class providing apparatus 110 classifies emotions by applying at least one information of task participation, frequently used words, emojis, and hashtags to the knowledge base.

The non-face-to-face class providing device 110 recognizes the emotion of each learner based on the change of wavelength in the audio input from the learner terminal 130 in the group chat room for each task and the sentence input by the learner into the chat.

When video and audio are not input from the learner terminal 130 in the group chat room for each task, the non-face-to-face class providing device 110 detects the emotions and emotions of each learner based only on the text. The non-face-to-face class providing device 110 builds a specific emotion dictionary using emotion-emotional information for each word of text input from the learner terminal 130 in the group chat room for each task in order to identify emotion information.

The non-face-to-face class providing device 110 communicates between the learner terminal 130 and the virtual assistant robot 120 in the group chat room for each task, and grasps the emotional change of the learner according to the context of the input text. The non-face-to-face class providing device 110 uses the virtual assistant robot 120 to understand changing emotions in a multi-turn situation. The non-face-to-face class providing device 110 communicates between the learner terminal 130 and the virtual assistant robot 120 using a hierarchical recurrent neural network (RNN) to understand the emotional state that changes according to the context of the input text. do.

When a foreign language text is input in the group chat room for each task, the non-face-to-face class providing device 110 uses the learning data after translating the foreign language text into Korean text using a network-connected translation module. When the non-face-to-face class providing device 110 performs translation by machine learning using the translation module, the translated Korean text may contain noise or may not be a perfect translation sentence according to a situation or context. The non-face-to-face class providing device 110 improves the quality of the translated sentences based on the language model combined with the Korean-specialized tokenizing module and uses them as learning data.

The non-face-to-face class providing device 110 is a service environment for providing the virtual assistant robot 120, and since the content of the conversation is fluidly changed according to the learning content, the level of the learning participant, the topic of the conversation, etc., the emotion is actively expressed in the conversation. In order to do so, emotional elements are classified into EKMAN emotions and seven neutral emotions (neutral, anger, fear, disgust, happiness, sadness, surprise), and the conversation situation is systematized using an ontology tree.

The non-face-to-face class providing device 110 stores the conversation text and voice collected in the non-face-to-face private chat room and the non-face-to-face group chat room by using the virtual assistant robot 120 . The non-face-to-face class providing device 110 classifies the learner's emotional information into seven emotions (neutral, anger, fear, disgust, happiness, sadness, surprise) by analyzing the collected text and voice.

Non-face-to-face class providing device 110 divides the classified emotional information into four domains of academic emotion (Achievement Emotion, Epistemic Emotion, Topic Emotion, Social Emotion ( Social Emotion)).

The non-face-to-face class providing device 110 is based on one or more information of achievement emotion, epistemic emotion, topic emotion, and social emotion confirmed based on emotional information. Recognize inconveniences in learning achievement, challenging learning, topic learning, and interaction. The non-face-to-face class providing device 110 provides a learning guide based on the inconvenience.

The non-face-to-face class providing device 110 is “difficult”, “interesting”, “unfamiliar”, “strange”, “Uh, why is it like this?” Recognizes the learner's emotions that occur during class, including emotional expressions such as "active questions" and learning cognitive activities such as "not submitting assignments".

The non-face-to-face class providing device 110 learns by mapping the classified emotional information to one or more of achievement emotion, epistemic emotion, topic emotion, and social emotion. figure out the mood

Achievement emotion refers to the emotion of success or failure according to the result of learning achievement activity. In other words, the non-face-to-face class providing device 110 classifies the emotion of success or failure due to the achievement activity as the achievement emotion. The non-face-to-face class providing device 110 classifies the pleasant feeling of learning, the hope/pride feeling related to success, and the anxious feeling about failure as the achievement emotion.

Epistemic emotion refers to emotions about the learning process that occur due to cognitive activities, such as surprise, curiosity, and joy when a problem is solved after confusion/frustration at a new task. In other words, the non-face-to-face class providing apparatus 110 classifies emotions generated in the process of learning new, non-routine knowledge into epistemological emotions. The non-face-to-face class providing device 110 classifies the surprising emotion about the new task, the emotion of curiosity, the fearful emotion, and the confused emotion as epistemological emotions.

Topic Emotion refers to the emotion that one's emotions are projected on the positive/negative topics presented in the class. In other words, the non-face-to-face class providing device 110 classifies the emotion related to the topic presented in the class as the topic emotion. The non-face-to-face class providing device 110 classifies all of the emotions of sympathy and affirmative/negative to the emotions of the person depicted in the material as topic emotions.

Social emotion refers to the emotions that arise from interactions with class, lecturers, and learners. In other words, the non-face-to-face class providing device 110 classifies emotions in class, lecturer, interaction with colleagues, and group learning as social emotions. The non-face-to-face class providing device 110 classifies emotions for love, sympathy, longing, respect, contempt, envy, anger, and social anxiety as social emotions.

The non-face-to-face class providing device 110 performs learning analytics and synthesizes data provided to learners such as sentence inspection, sentence writing practice, daily quizzes, and FLEX, learner information, activities, learning history, disposition, etc. Analyze and recommend problems.

The non-face-to-face class providing device 110 supports user voice-based voice chat as an audio chat, and voice conversation within the platform as a voice talk.

The non-face-to-face class providing device 110 applies NLU (Natural Language Understanding) technology for an intelligent conversation service to the virtual assistant robot 120 . The non-face-to-face class providing device 110 reflects real-time feedback for communication between the learner and the lecturer to the virtual assistant robot 120 .

The non-face-to-face class providing device 110 applies NLG (Natural Language Generation) technology for a chatbot conversation with a learner to the virtual assistant robot 120 . The non-face-to-face class providing device 110 reflects real-time feedback for communication between the learning group and the educator in the virtual assistant robot 120 . The non-face-to-face class providing device 110 applies the NLG technology for the learning group and the chatbot conversation to the virtual assistant robot 120 .

The non-face-to-face class providing device 110 is “difficult”, “interesting”, “unfamiliar”, “strange” among the contexts of the text input in the group chat room for each task in order to solve the learners’ anxiety in the non-face-to-face learning environment. , "Uh, why is this like this?" Identify the emotions of learners such as “active questions” and “not submitting assignments”. The non-face-to-face class providing device 110 uses the virtual assistant robot 120 to detect the learner's reaction, analyzes the acquired emotional data, and transmits it to the lecturer terminal 140 . The lecturer terminal 140 may be used as materials to design or supplement lecture contents based on emotion data.

The non-face-to-face class providing device 110 uses the virtual assistant robot 120 to provide a quiz after non-face-to-face learning to the learner terminal 130 to grasp the level of understanding of the learner and at the same time induce the learner to ask additional questions do. The non-face-to-face class providing device 110 uses the virtual assistant robot 120 so that learners can comfortably ask questions.

The non-face-to-face class providing device 110 guides a group of learners as well as individual learning guidance using the virtual assistant robot 120 . The non-face-to-face class providing device 110 helps the learner to “directly or indirectly grasp my level based on the reactions of other learners” in the learner group. The non-face-to-face class providing device 110 identifies the learner's own level and at the same time uses the virtual assistant robot 120 to identify the learner's emotions. The non-face-to-face class providing device 110 encourages more active activities of learners by analyzing a reaction based on emotion.

The non-face-to-face class providing device 110 recognizes emotions based on machine learning in consideration of the learner conversation context and the characteristics of each learner. The non-face-to-face class providing device 110 implements and learns a model based on the emotional response embedding characteristic information for each learner. The non-face-to-face class providing device 110 infers emotion according to an event that occurs as a result of interaction between agents in a multi-agent system by using an emotion reasoning machine. The non-face-to-face class providing device 110 varies according to the interpretation of a situation occurring for each individual learner, and adaptively adjusts each individual's goals, expectations, and learning according to the learner's interpretation of each situation. The non-face-to-face class providing device 110 has different internal models (eg, a total of 24 currently emotional forms) according to the learner's point of view, respectively.

The non-face-to-face class providing device 110 creates a social group in consideration of the characteristics of each learner by using a social agent matching technology to increase the learner's interaction and emotional factors. The non-face-to-face class providing device 110 recommends a social group created through the steps of learner characteristic classification, profile analysis, profile learning, and filtering.

The non-face-to-face class providing device 110 includes student information (gender, age, region, major, courses, grades), interest information (general interest information, self-introduction), machine learning information (participation level) in order to classify the characteristics of the learner. , Q&A, emotional recognition). The non-face-to-face class providing device 110 classifies "student information" on a rule-based basis for profile analysis. The non-face-to-face class providing apparatus 110 classifies "interest information" based on machine learning.

The non-face-to-face class providing device 110 learns the information classified in the profile analysis and the learner's emotional recognition information for profile learning based on machine learning. The non-face-to-face class providing device 110 classifies and recommends the learned information for filtering based on rules and machine learning. The non-face-to-face class providing device 110 continuously learns the feedback of the recommended matching group through profile learning to activate the behavior that satisfies the learner's emotions and emotions.

The learner terminal 130 and the lecturer terminal 140 refer to electronic devices that perform voice or data communication via a network according to a user's key manipulation. The learner terminal 130 and the lecturer terminal 140 are a memory for storing a program or protocol for communicating with the non-face-to-face class providing device 110 via a network, a microprocessor for executing and controlling the corresponding program, etc. to provide

The learner terminal 130 and the lecturer terminal 140 are a smart phone, a tablet, a laptop, a personal computer (PC), a personal digital assistant (PDA), a portable It may be an electronic device such as a portable multimedia player (PMP), a wireless communication terminal, or a media player.

The learner terminal 130 and the lecturer terminal 140 are a communication device such as a communication modem for performing communication with various devices or wired and wireless networks, a memory for storing various programs and data, and a microcomputer for executing and controlling the program. Various devices having a processor and the like. According to at least one embodiment, the memory is a computer such as random access memory (RAM), read only memory (ROM), flash memory, optical disk, magnetic disk, solid state disk (SSD), etc. It may be a readable recording/storage medium. According to at least one embodiment, a microprocessor may be programmed to selectively perform one or more of the operations and functions described herein. According to at least one embodiment, the microprocessor may be fully or partially implemented as hardware such as an Application Specific Integrated Circuit (ASIC) of a specific configuration.

2 is a diagram schematically illustrating an apparatus for providing a non-face-to-face class according to the present embodiment.

The non-face-to-face class providing device 110 according to this embodiment includes a chat room opening unit 210, a collection unit 220, a de-identification unit 222, an intelligent conversation solution providing unit 224, a feedback reflecting unit 226, It includes a natural language response unit 228 , an analysis unit 230 , a user pattern generation unit 240 , an emotion classification unit 250 , a query content summary unit 260 , and a non-face-to-face class support unit 270 . Components included in the non-face-to-face class providing device 110 are not necessarily limited thereto.

Each component included in the non-face-to-face class providing device 110 is connected to a communication path that connects a software module or a hardware module inside the device, so that they can operate organically with each other. These components communicate using one or more communication buses or signal lines.

Each component of the non-face-to-face class providing device 110 shown in FIG. 2 means a unit that processes at least one function or operation, and may be implemented as a software module, a hardware module, or a combination of software and hardware. .

The chat room opening unit 210 opens a non-face-to-face personal chat room for each study task and a non-face-to-face group chat room for each task based on the lecture plan received from the lecturer terminal 140 .

The collection unit 220 engages the virtual assistant robot 120 together with the learner terminal 130 in the non-face-to-face personal chat room and the non-face-to-face group chat room for each learning task. The collection unit 220 collects text information input from the non-face-to-face private chat room and the non-face-to-face group chat room by using the virtual assistant robot 120 .

The de-identification unit 222 generates de-identification information in which data de-identification is performed on text information. When automatic de-identification is selected, the de-identification unit 222 uses a regular expression from text information to first de-identify important personal information including at least one of email, phone number, ID, and password as the first de-identification target. select The de-identification unit 222 selects the second de-identification target including at least one of a company name, an organization name, an abbreviation, and an abbreviation by using pattern matching, an entity name dictionary, and a recognizer from text information. The de-identification unit 222 de-identifies the first de-identification target and the second de-identification target.

When manual de-identification is selected, the de-identification unit 222 selects only the manually substituted data based on the de-identification guideline established according to the manual de-identification command from the text information as de-identification targets to de-identify. carry out anger

The intelligent dialogue solution providing unit 224 generates a natural language understanding result obtained by understanding a natural language query for de-identified information in real time based on the intelligent dialogue solution. The intelligent dialogue solution providing unit 224 provides an intelligent dialogue solution for performing a learner query domain, an intention classification, and a real-time learner similar query search for de-identified information.

The intelligent conversation solution providing unit 224 generates classification information in which de-identified information is classified according to a learner query domain and an intention classification system. The intelligent conversation solution providing unit 224 reflects the natural language deep understanding result obtained by deep understanding the learner's query in consideration of the dialogue context in the classification information to the natural language understanding result. The intelligent dialogue solution providing unit 224 reflects the result of the learner's deep understanding of the learner's query in consideration of the characteristics of each learner based on the classification information to the natural language understanding result. The intelligent dialogue solution providing unit 224 completes the learner query domain and intention classification system by learning the classification information, the natural language deep implementation result, and the learning deep understanding result.

The intelligent dialogue solution providing unit 224 uses a search engine loaded with a learned classification model and a Korean-specialized tokenizing module when a learner query in de-identified information is given as an input to search for a real-time learner-like query. Extract the semantic feature value of the input query within time (several seconds).

The intelligent conversation solution providing unit 224 measures the similarity value of the learner's query based on machine learning based on the semantic feature value. The intelligent conversation solution providing unit 224 generates a comparison result compared with a specific threshold based on the similarity value.

When the comparison result is lower than a specific reference value, the intelligent dialogue solution providing unit 224 considers that it does not exist in the predefined domain and intention classification system, infers a similar query using the conversation with the learner, and then analyzes the similar query on a case-by-case basis. It is bundled and converted into a DB in real time and reflected in the natural language understanding results.

When the comparison result is greater than or equal to a specific reference value, the intelligent conversation solution providing unit 224 indexes the query related to a predefined category in the real-time similar query DB, considering that it exists in the predefined domain and intention classification system, and results in natural language understanding create The intelligent conversation solution providing unit 224 sets a specific reference value as an optimal value calculated based on repeated performance evaluation.

The feedback reflecting unit 226 outputs real-time feedback information necessary for communication between the learner and the lecturer participating in the non-face-to-face personal chat room and the non-face-to-face group chat room based on the natural language understanding result.

The feedback reflecting unit 226 extracts a response to the learner's query based on the natural language understanding result. The feedback reflecting unit 226 generates a learner characteristic and response analysis result corresponding to a response to the learner's query. The feedback reflecting unit 226 reflects and outputs the learner characteristics and response analysis results in feedback information between the learner and the lecturer.

The feedback reflecting unit 226 extracts a response to a question about the learner from among the natural language understanding results. The feedback reflecting unit 226 extracts important concepts in the lecture that have meaning among the responses to the learner's query. The feedback reflecting unit 226 calculates the difficulty level of the lecture and the degree of education felt by the learner based on important concepts in the lecture having meaning. The feedback reflecting unit 226 extracts from the real-time database, a learning type according to whether the learner is an individual or a group, a preferred language based on the language used by the learner, and characteristics of each learner based on a preferred concept preferred by the learner figure out The feedback reflecting unit 226 outputs feedback information between the learner and the instructor by reflecting the lecture difficulty and education level felt by the learner for each characteristic of each learner.

The natural language response unit 228 generates a natural language response to a learner's query for a conversation between a learner and a chatbot participating in a non-face-to-face personal chat room and a non-face-to-face group chat room based on the natural language understanding result, and then uses the virtual assistant robot 120 to output.

The natural language response unit 228 generates a template-based pre-question for a learner's query and a response-related objective dialogue fingerprint based on the natural language understanding result. The natural language response unit 228 applies the template-based pre-question and the response-related objective dialogue fingerprint to the response generation technology, and then generates a general dialogue fingerprint learned by machine learning for emotional communication with the learner. The natural language response unit 228 reflects the general conversation fingerprint in the natural language response.

The natural language response unit 228 collects additional response information by using the prequery in consideration of the dialogue context and characteristics of each learner within the natural language understanding result in order to generate a template-based pre-question and response-related objective dialogue. The natural language response unit 228 defines a response template for each query case based on the additional response information.

The natural language response unit 228 implements a response sentence generation model based on machine learning in consideration of the conversation context and learner-specific characteristics within the natural language understanding result in order to generate a general conversation fingerprint. The natural language response unit 228 uses a machine learning-based emotion analyzer to identify the contextual emotional state in the dialogue of the natural language understanding result. The natural language response unit 228 embeds machine learning-based learner-specific characteristic information. The natural language response unit 228 implements a machine learning-based emotional response generation model for emotional communication. The natural language response unit 228 generates a natural language response based on the response sentence generation model, the contextual emotional state, the learner-specific characteristic information, and the emotional response generation model.

The analysis unit 230 generates an analysis result by analyzing text information collected from non-face-to-face personal chat rooms and non-face-to-face group chat rooms. When the text information collected from the non-face-to-face personal chat room and the non-face-to-face group chat room is foreign text, the analysis unit 230 generates a translation of the foreign language text into Korean text using a network-connected translation module and then learns the translation. Analyze as data. The analysis unit 230 applies a language model combined with a Korean-specialized tokenizing module to the translation to correct misinterpretation according to situations and contexts.

The user pattern generator 240 generates a user pattern for each learner based on the analysis result. The user pattern generator 240 calculates at least one learning pattern among an attendance rate, a task performance rate, and a learning time based on the analysis result. The user pattern generator 240 calculates a learning tendency of at least one of an area of interest, activeness, and concentration based on the analysis result. The user pattern generator 240 calculates learning outcomes of at least one of achievement, a weak field, and an excellent field based on the analysis result. The user pattern generator 240 generates a user pattern by combining a learning pattern, a learning tendency, and a learning result.

The user pattern generator 240 generates a personal learner pattern based on the analysis result of analyzing personal text information input from the non-face-to-face personal chat room. The user pattern generation unit 240 generates a group learner pattern based on the analysis result of analyzing group text information input from the non-face-to-face group chat room. The non-face-to-face class support unit 270 extracts a non-face-to-face group chat room having a group learning pattern having the same pattern as the individual learner pattern or more than a threshold.

The emotion classification unit 250 classifies the emotion information based on the analysis result. The emotion classification unit 250 classifies the user emotion into positive and negative based on the frequencies of positive words, neutral words, and negative words included in the analysis result. When the user's emotion is positive, the emotion classification unit 250 classifies it into happiness and satisfaction as detailed emotions. When the user's emotion is negative, the emotion classification unit 250 classifies the user's emotion into anger and pain as detailed emotion.

The sentiment classification unit 250 calculates importance based on the result of measuring the frequency of words included in the text information based on the analysis result. The emotion classification unit 250 builds a post-emotional vocabulary list based on the importance. The sentiment classification unit 250 constructs a sentiment dictionary using the sentiment vocabulary by calculating the sentiment index for each sentiment vocabulary list.

The emotion classification unit 250 analyzes at least one of frequently used words, emojis, and hashtags input in text information based on the analysis result. The emotion classification unit 250 classifies the emotion information by matching the emotion vocabulary existing in the emotion dictionary.

The emotion classification unit 250 detects an object from the video data input in the non-face-to-face private chat room and the non-face-to-face group chat room. The emotion classification unit 250 detects a face region of an object in the image data. The emotion classification unit 250 recognizes a change in expression after recognizing an expression in the face region. The emotion classification unit 250 recognizes a change in wavelength of audio data input in a non-face-to-face personal chat room and a non-face-to-face group chat room. The emotion classification unit 250 classifies emotion and emotion information for each learner based on a change in expression, a change in wavelength, and an analysis result.

The emotion classification unit 250 performs communication between the virtual assistant robot 120 and the learner in the non-face-to-face personal chat room and the non-face-to-face group chat room, and calculates the emotional change of the learner according to the context of the text inputted by the hierarchical RNN (Recurrent Neural Network). ) to find out.

The emotion classification unit 250 classifies the emotion clusters generated as a result by a similar cause. The emotion classification unit 250 evaluates the emotion for each event, entity, and agent with respect to the emotion cluster. When evaluating the emotion, the emotion classification unit 250 performs evaluation on whether the agent likes the target related to the satisfaction of the event related to the goal, the degree of approval for the agent, the actions of other agents, and the agent's attitude. Calculate the strength. The emotion classification unit 250 classifies the emotion into a specific emotion according to the interpretation of a given situation based on the intensity of the emotion.

The emotion classification unit 250 recognizes emotion based on machine learning in consideration of the learner conversation context and the characteristics of each learner based on the analysis result. The emotion classification unit 250 learns after implementing a model based on the emotional response in which the characteristic information for each learner is embedded.

The emotion classification unit 250 infers emotion according to an event that occurs due to interaction between agents in a multi-agent system using a sentiment inferrer. The emotional classification unit 250 varies according to the interpretation of a situation that has occurred for each individual learner, and adaptively adjusts each individual's goals, expectations, and learning according to the learner's interpretation of each situation. The emotion classification unit 250 generates different internal emotion models (eg, a total of 24 emotion types currently) according to the learner's point of view, respectively.

The query content summary unit 260 extracts only the query content from the text information based on the analysis result. The query content summary unit 260 identifies duplicate content among the query content and generates query aggregate data in which the duplicate query content is combined into a single query. The query content summary unit 260 selects sentences having a sentence length greater than or equal to a preset sentence length from among the query aggregate data and generates query summary data summarized within a preset number of sentences or words.

The non-face-to-face class support unit 270 uses the virtual assistant robot 120 to support the non-face-to-face class based on the analysis result, user pattern and emotional information for each learner. The non-face-to-face class support unit 270 transmits the query summary data to the lecturer terminal 140 using the virtual assistant robot 120 .

The non-face-to-face class support unit 270 calculates the learning participation level based on the number of sentence input per learner, the sentence length, and the text content based on the analysis result. The non-face-to-face class support unit 270 extracts a learner with the lowest participation rate based on the learning participation level. The non-face-to-face class support unit 270 transmits a learning participation encouragement message to the learner terminal corresponding to the learner with the lowest participation rate by using the virtual assistant robot 120 for each learning task.

The non-face-to-face class support unit 270 extracts only non-class content from the text information based on the analysis result. The non-face-to-face class support unit 270 selects only sentences requiring help or support by analyzing the content outside of class. The non-face-to-face class support unit 270 selects a category corresponding to a sentence requiring help or support. The non-face-to-face class support unit 270 extracts extra-class support information (eg, part-time job information) corresponding to a corresponding category (a category corresponding to a sentence requiring help or support). The non-face-to-face class support unit 270 transmits support information outside of class to the learner terminal 130 in which a sentence requiring help or support is input using the virtual assistant robot 120 .

The non-face-to-face class support unit 270 uses the virtual assistant robot 120 to use the learner terminal 130 corresponding to the individual learner existing in the non-face-to-face private chat room, and the group learning pattern having the same pattern as the individual learner pattern and the threshold value. Participation in non-face-to-face group chat rooms is recommended.

The non-face-to-face class support unit 270 uses the virtual assistant robot 120 to recognize the learner's emotion according to a predefined representative emotion, and generates an appropriate response and emotional expression suitable for the interpretation of the question situation.

The non-face-to-face class support unit 270 applies NLU (Natural Language Understanding) technology to the virtual assistant robot 120 for an intelligent conversation service, so that the virtual assistant robot 120 provides real-time feedback for communication between learners and lecturers. do. The non-face-to-face class support unit 270 applies NLG (Natural Language Generation) technology to the virtual assistant robot 120 for an intelligent conversation service to provide real-time feedback for communication between the learner and the virtual assistant robot 120 .

The non-face-to-face class support unit 270 provides emotional expressions (difficult), “interesting”, “unfamiliar”, “strange”, “Uh, why is this so?”, “active question”, “No assignments submitted”). The non-face-to-face class support unit 270 transmits the emotion data analyzed by the emotional expression to the lecturer terminal 140 so that the lecturer terminal 140 designs or supplements the lecture contents based on the emotion data.

The non-face-to-face class support unit 270 uses the virtual assistant robot 120 to provide a quiz after non-face-to-face learning to the learner terminal 130 to grasp the level of understanding of the learner and induces the learner to ask additional questions. .

The non-face-to-face class support unit 270 uses the virtual assistant robot 120 to guide not only individual learning guidance but also a group of learners, and the learner himself/herself directly or indirectly grasps his/her own level based on the reactions of other learners in the learner group. let it do

The non-face-to-face class support unit 270 grasps the learner's emotions by using the virtual assistant robot 120 at the same time as the learner himself grasps his or her own level. The non-face-to-face class support unit 270 encourages active activities of the learner based on the learner's emotion.

The non-face-to-face class support unit 270 creates a social group in consideration of the characteristics of each learner by using a social agent matching technology to increase the learner's interaction and emotional factors. The non-face-to-face class support unit 270 recommends a social group created through the steps of learner characteristic classification, profile analysis, profile learning, and filtering to individual learners existing in non-face-to-face private chat rooms. do.

The non-face-to-face class support unit 270 includes at least one of student information, general interest information, and self-introduction including at least one of gender, age, region, major, subject, and grade in order to classify the characteristics of the learner. Machine learning information including at least one of interest information, participation level, question and answer, and emotional recognition is acquired. The non-face-to-face class support unit 270 classifies student information based on rules. The non-face-to-face class support unit 270 performs profile analysis for classifying interest information based on machine learning.

The non-face-to-face class support unit 270 performs profile learning based on machine learning based on the information classified in the profile analysis process and the learner's emotional recognition information. The non-face-to-face class support unit 270 filters information on profile learning based on rules and machine learning.

3 is a diagram illustrating a learner group chatting during class using the virtual assistant robot according to the present embodiment.

As shown in FIG. 3 , the non-face-to-face class providing device 110 collects questions during class using the virtual assistant robot 120 .

The non-face-to-face class providing device 110 provides lecture materials and lecture auxiliary materials to the learner terminal 130 for each individual or group of the learner terminal 130 according to a learning schedule. The non-face-to-face class providing device 110 conducts a lecture according to a learning schedule, and collects chat information for each lecture by using the virtual assistant robot 120 .

The non-face-to-face class providing device 110 is a learner scenario, and uses the virtual assistant robot 120 to freely support students' discussions during class through group chat. The non-face-to-face class providing device 110 uses the virtual assistant robot 120 to group the queries based on the conversations the learners had in the group chat and transmits them to the lecturer terminal 140 .

4 is a diagram illustrating a study aid for query grouping and lecture reference materials using the virtual assistant robot according to the present embodiment.

As shown in FIG. 4 , the non-face-to-face class providing device 110 transmits a student group query that occurred in a weekly lecture to the lecturer terminal 140 using the virtual assistant robot 120 as a lecturer scenario.

The non-face-to-face class providing device 110 analyzes the collected chat information for each lecture to generate a chat analysis result, and then extracts only the query content from the chat analysis result. The non-face-to-face class providing device 110 identifies duplicate content among the query content and generates query aggregate data obtained by combining the duplicate query content into one query. The non-face-to-face class providing apparatus 110 selects sentences having a preset length or more from among the query collection data and generates query summary data summarized within a preset number of sentences or words. The non-face-to-face class providing device 110 transmits the query summary data to the lecturer terminal 140 using the virtual assistant robot 120 .

The lecturer terminal 140 may refer to the remaining lecture after receiving the query summary data from the virtual assistant robot 120 . The lecturer terminal 140 may deliver the lecture suspension guidance to the virtual assistant robot 120 . The non-face-to-face class providing device 110 receives class suspension guidance information from the virtual assistant robot 120 and delivers it to the chat room for each task.

The non-face-to-face class providing device 110 opens a non-face-to-face personal chat room for each study task and a non-face-to-face group chat room for each task based on the lecture plan received from the instructor terminal 140 . The non-face-to-face class providing device 110 engages the virtual assistant robot 120 together with the learner terminal 130 in the non-face-to-face personal chat room and non-face-to-face group chat room for each learning task, and uses the virtual assistant robot 120 to the learner terminal Collects text information input from 130 .

The non-face-to-face class providing device 110 generates de-identification information by performing data de-identification on text information. The non-face-to-face class providing device 110 primarily selects important personal information including at least one of an email, a phone number, an ID, and a password as the first de-identification target by using a regular expression from text information. The non-face-to-face class providing device 110 is secondarily selected as a second de-identification target including at least one of a company name, an institution name, an abbreviation and abbreviation by using a pattern matching, an entity name dictionary and a recognizer from text information. . The non-face-to-face class providing apparatus 110 performs de-identification of the first de-identification target and the second de-identification target.

When manual de-identification is selected, the non-face-to-face class providing device 110 selects as a de-identification target only for manually substituted data based on a pre-established de-identification guideline according to a manual de-identification command from text information. perform identification.

The non-face-to-face class providing device 110 generates a natural language understanding result of real-time understanding of a natural language query for de-identified information based on an intelligent conversation solution. The non-face-to-face class providing device 110 provides an intelligent conversational solution for performing a learner query domain, an intention classification, and a real-time learner similar query search for de-identified information.

The non-face-to-face class providing device 110 generates classification information in which de-identified information is classified according to a learner query domain and an intention classification system. The non-face-to-face class providing device 110 reflects the natural language deep understanding result obtained by deep understanding the learner's query in consideration of the dialogue context in the classification information to the natural language understanding result. The non-face-to-face class providing device 110 reflects the result of the learner's in-depth understanding of the learner's query in consideration of the characteristics of each learner based on the classification information to the natural language understanding result. The non-face-to-face class providing device 110 completes the learner query domain and intention classification system by learning the classification information, the natural language in-depth implementation result, and the learning in-depth understanding result.

The non-face-to-face class providing device 110 uses a search engine loaded with a learned classification model and a Korean-specialized tokenizing module when a learner query in non-identified information is given as an input to search for a real-time learner-like query. Extract the semantic feature value of the input query within time (several seconds). The non-face-to-face class providing apparatus 110 measures a similarity value of a learner's query based on machine learning based on the semantic feature value. The non-face-to-face class providing device 110 generates a comparison result compared with a specific threshold based on the similarity value. When the comparison result is lower than a specific reference value, the non-face-to-face class providing device 110 considers that it does not exist in the predefined domain and intention classification system, infers a similar query using a conversation with the learner, and then analyzes the similar query on a case-by-case basis. It is bundled and converted into a DB in real time and reflected in the natural language understanding results. When the comparison result is greater than or equal to a specific reference value, the non-face-to-face class providing device 110 indexes a query related to a predefined category in a real-time similar query DB, considering that it exists in a predefined domain and intention classification system, and results in natural language understanding create The non-face-to-face class providing device 110 sets a specific reference value as an optimal value calculated based on the repeated performance evaluation.

5 is a view showing a learning aid for a group task using the virtual assistant robot according to the present embodiment.

As shown in FIG. 5 , the non-face-to-face class providing device 110 supports group assignments outside of class as a learner scenario. The non-face-to-face class providing device 110 uses the virtual assistant robot 120 to form a team based on the accumulated student data to perform the group task this semester and notify the start of the activity.

The non-face-to-face class providing device 110 uses the virtual assistant robot 120 to select a leader and a member corresponding to the learning task. The non-face-to-face class providing device 110 selects the learner terminal 130 corresponding to the leader and the group and invites them to one group chat room. The non-face-to-face class providing device 110 assists in performing tasks by using the virtual assistant robot 120 to perform tasks for each role in agreement with the leader and the assistants.

The non-face-to-face class providing device 110 encourages team members with low participation in the group chat room by using the virtual assistant robot 120 . The non-face-to-face class providing device 110 collects and analyzes group chat information. The non-face-to-face class providing device 110 extracts the member with the lowest participation rate in the group chat based on the chat analysis result. The non-face-to-face class providing device 110 transmits a learning participation encouragement message to the learner terminal 130 corresponding to the learner with the lowest participation rate by using the virtual assistant robot 120 .

For example, the non-face-to-face class providing device 110 uses the virtual assistant robot 120 to select learners who do not input any comments for a preset time (eg, about 12 hours) in the group chat room. The non-face-to-face class providing device 110 transmits a participation encouragement message to the selected low participation team members.

The non-face-to-face class providing device 110 transmits information about the test in the group chat room using the virtual assistant robot 120 . The non-face-to-face class providing device 110 transmits a notice about the test in the group chat room using the virtual assistant robot 120 .

The non-face-to-face class providing device 110 generates a natural language response to a learner's query for a conversation between a learner and a chatbot participating in a non-face-to-face personal chat room and a non-face-to-face group chat room based on the natural language understanding result, and then outputs it using a virtual assistant robot let it do

The non-face-to-face class providing device 110 generates a template-based pre-question for a learner's query and a response-related conversational fingerprint based on the natural language understanding result. The non-face-to-face class providing device 110 applies the template-based pre-question and the response-related purpose dialogue fingerprint to the response generation technology. The non-face-to-face class providing device 110 generates a general conversation fingerprint learned by machine learning for emotional communication with the learner. The non-face-to-face class providing device 110 reflects the general conversation fingerprint in the natural language response.

The non-face-to-face class providing device 110 collects additional response information by using a pre-question in consideration of the dialogue context and characteristics of each learner within the natural language understanding result in order to generate a template-based pre-question and response-related objective dialogue. The non-face-to-face class providing device 110 defines a response template for each query case based on the additional response information.

The non-face-to-face class providing device 110 implements a response sentence generation model based on machine learning in consideration of the conversation context and learner-specific characteristics within the natural language understanding result in order to generate a general conversation fingerprint. The non-face-to-face class providing device 110 uses a machine learning-based emotion analyzer to identify the contextual emotional state in the dialogue of the natural language understanding result. The non-face-to-face class providing device 110 embeds machine learning-based learner-specific characteristic information. The non-face-to-face class providing device 110 implements a machine learning-based emotional response generation model for emotional communication. The non-face-to-face class providing device 110 generates a natural language response based on the response sentence generation model, the contextual emotional state, the characteristic information for each learner, and the emotional response generation model.

6 is a diagram illustrating class-related activities and provision of quizzes using the virtual assistant robot according to the present embodiment.

As shown in FIG. 6 , the non-face-to-face class providing device 110 performs support outside of class as a learner scenario. The non-face-to-face class providing device 110 comprehensively supports matters necessary for school life, such as academic schedule guidance and study information, using the virtual assistant robot 120 .

The non-face-to-face class providing device 110 delivers lecture materials and lecture auxiliary data according to a learning schedule for each group chat room. The non-face-to-face class providing device 110 delivers a lecture material guide message to the group chat room before a preset schedule (eg, 3 days) before the corresponding lecture day arrives according to the learning schedule.

The non-face-to-face class providing apparatus 110 delivers the contents of the lecture contents to the group chat room before a preset schedule (eg, 2 days) before the arrival of the corresponding lecture day according to the learning schedule. The non-face-to-face class providing device 110 delivers lecture materials and lecture auxiliary materials before a preset schedule (eg, 1 day) before the corresponding lecture day arrives according to the learning schedule.

The non-face-to-face class providing device 110 analyzes group chat information.

The non-face-to-face class providing device 110 extracts only out-of-class content from among the chatting analysis results of analyzing the collected group chat information. The non-face-to-face class providing device 110 selects a sentence requiring help or support by analyzing the content outside of class. The non-face-to-face class providing device 110 checks the content of a sentence requiring help or support. The non-face-to-face class providing device 110 selects a category corresponding to a sentence requiring help or support. The non-face-to-face class providing device 110 extracts extra-class support information (eg, part-time job information) corresponding to the corresponding category. The non-face-to-face class providing device 110 transmits extra-class support information to the learner terminal 130 in which a sentence requiring help or support is input using the virtual assistant robot 120 .

The non-face-to-face class providing device 110 analyzes group chat information. The non-face-to-face class providing device 110 generates a chat analysis result by analyzing the collected group chat information. The non-face-to-face class providing device 110 selects individual learners who have not participated in the group chat based on the chat analysis result. The non-face-to-face class providing device 110 takes care of the learning emotion of transmitting a message for recommending a study group or recommending a chat group as an individual learner who does not participate in the group chat.

The non-face-to-face class providing device 110 outputs a quiz (task) for each lecture week to the group chat room by using the virtual assistant robot 120 . The non-face-to-face class providing device 110 uses the virtual assistant robot 120 to notify the quiz deadline for each parking group in the group chat room. The non-face-to-face class providing device 110 receives an answer (task result) corresponding to the quiz from the learner terminal 130 participating in the group chat room by using the virtual assistant robot 120 . The non-face-to-face class providing device 110 transmits the scoring result to the personal chat room after confirming the correct answer for the answer (task result) corresponding to the quiz received from the learner terminal 130 . The non-face-to-face class providing device 110 attaches a link for viewing an explanation of an incorrect problem to the scoring result.

The non-face-to-face class providing device 110 encourages the learner's participation in activities such as class-related tasks and quizzes. The non-face-to-face class providing device 110 automatically scores multiple-choice tasks and immediately provides the results to the learner. The non-face-to-face class providing device 110 collects the subjective tasks input by the learners and then delivers them to the lecturer.

The non-face-to-face class providing device 110 outputs real-time feedback information necessary for communication between the learner and the lecturer participating in the non-face-to-face personal chat room and the non-face-to-face group chat room based on the natural language understanding result.

The non-face-to-face class providing device 110 extracts a response to the learner's query based on the natural language understanding result. The non-face-to-face class providing device 110 generates a learner characteristic and response analysis result corresponding to a response to the learner's query. The non-face-to-face class providing device 110 reflects and outputs the learner characteristics and response analysis results in feedback information between the learner and the instructor.

The non-face-to-face class providing device 110 extracts a response to a question about the learner from the natural language understanding result. The non-face-to-face class providing device 110 extracts important concepts in the lecture that have meaning among the responses to the learner's query. The non-face-to-face class providing device 110 calculates the difficulty level of the lecture and the degree of education felt by the learner based on important concepts in the lecture having meaning. The non-face-to-face class providing device 110 is a learning type according to whether the learner is an individual or a group extracted from the real-time database, a preferred language based on the language used by the learner, and a learner based on a preferred concept preferred by the learner. Identify the star characteristics. The non-face-to-face class providing device 110 outputs feedback information between the learner and the instructor by reflecting the lecture difficulty and education level felt by the learner for each characteristic of each learner.

7 is a view showing a screen for generating a class alarm to the lecturer using the virtual assistant robot according to the present embodiment.

As shown in FIG. 7 , the non-face-to-face class providing device 110 generates a class alarm to the lecturer as a lecturer scenario. The non-face-to-face class providing device 110 supports the lecturer to recognize that there is a class and to prepare in advance by using the virtual assistant robot 120 .

The non-face-to-face class providing device 110 transmits a notification to the lecturer terminal 140 before the lecture starts according to the learning schedule. The non-face-to-face class providing device 110 transmits the contents of the current lecture to the lecturer terminal 140 based on the lecture plan. The non-face-to-face class providing device 110 detects the progress of the train and transmits it to the lecturer terminal 140 .

The non-face-to-face class providing device 110 guides the progress situation to the lecturer terminal 140 using the virtual assistant robot 120 . The non-face-to-face class providing device 110 summarizes and delivers the contents of the current lecture to the lecturer terminal 140 using the virtual assistant robot 120 . The non-face-to-face class providing device 110 uses the virtual assistant robot 120 to capture and deliver the contents of the last lecture to the lecturer terminal 140 .

The non-face-to-face class providing device 110 helps a lecturer to comfortably conduct a class by using the virtual assistant robot 120 .

The above description is merely illustrative of the technical idea of this embodiment, and various modifications and variations will be possible by those skilled in the art to which this embodiment belongs without departing from the essential characteristics of the present embodiment. Accordingly, the present embodiments are intended to explain rather than limit the technical spirit of the present embodiment, and the scope of the technical spirit of the present embodiment is not limited by these embodiments. The protection scope of this embodiment should be interpreted by the following claims, and all technical ideas within the equivalent range should be interpreted as being included in the scope of the present embodiment.

110: non-face-to-face class providing device
120: virtual assistant robot
130: learner terminal
140: lecturer terminal
210: chat room opening section
220: collection unit
230: analysis unit
240: user pattern generation unit
250: false classification unit
260: query content summary part
270: Non-face-to-face class support department

Claims (7)

A chat room opening unit for opening a non-face-to-face private chat room for each study task and a non-face-to-face group chat room for each task based on the syllabus received from the lecturer's terminal;
In the non-face-to-face personal chat room and the non-face-to-face group chat room for each learning task, a virtual assistant robot is engaged together with a learner terminal, and text information input from the non-face-to-face personal chat room and the non-face-to-face group chat room is obtained using the virtual assistant robot. a collecting unit to collect;
an analysis unit generating an analysis result of analyzing the text information;
a user pattern generator for generating a user pattern for each learner based on the analysis result;
an emotion classification unit for classifying emotion information based on the analysis result; and
A non-face-to-face class support unit that supports non-face-to-face classes based on the analysis result, the learner-specific user pattern and the emotional information using the virtual assistant robot
Including, wherein the sentiment classification unit calculates importance based on a result of measuring the frequency of words included in the text information based on the analysis result, builds a sentiment vocabulary list based on the importance, and the sentiment vocabulary A sentiment dictionary is constructed by calculating the emotional index for each list, and after analyzing at least one of frequently used words, emojis, and hashtags in the text information based on the analysis result, the sentiment vocabulary existing in the sentiment dictionary Classifies the emotional information by matching with , detects an object from the video data input in the non-face-to-face personal chat room and the non-face-to-face group chat room, detects the face region of the object, and recognizes the expression in the face region After recognizing the change in the facial expression, recognizing the wavelength change of the audio data input in the non-face-to-face personal chat room and the non-face-to-face group chat room, each learner based on the change in the facial expression, the wavelength change, and the analysis result Classify emotion and the emotional information, and hierarchical RNN (Recurrent) Neural Network) is used to identify
The non-face-to-face class support unit applies NLU (Natural Language Understanding) technology to the virtual assistant robot for intelligent conversation service so that the virtual assistant robot provides real-time feedback for communication between learners and lecturers, or NLG (Natural Language Generation) ) technology is applied to the virtual assistant robot to provide real-time feedback for communication between the learner and the virtual assistant robot. According to claim 1,
The user pattern generation unit,
Based on the analysis result, at least one learning pattern of attendance rate, task performance rate, and learning time is calculated, and based on the analysis result, at least one learning tendency of interest, activeness, and concentration is calculated, and the analysis result is A non-face-to-face class providing apparatus, characterized in that after calculating at least one learning result among achievement, weak field, and excellent field based on the combination of the learning pattern, the learning tendency, and the learning result, the user pattern is generated. 3. The method of claim 2,
The emotion classification unit,
User emotion is classified into positive and negative based on the frequency of positive words, neutral words, and negative words included in the analysis result, and when the user emotion is positive, it is classified into happiness and satisfaction as detailed emotions, and the user emotion is In the case of negative, non-face-to-face class providing device, characterized in that it is classified into anger and suffering as detailed sensibility. 4. The method of claim 3,
Extracting only the query contents from the text information based on the analysis result, checking the duplicate contents among the query contents to generate query aggregate data in which the duplicate query contents are combined into one query, and a preset sentence among the query collection data A query content summary unit that selects sentences having a length or more and generates query summary data summarized within a preset number of sentences or words
In addition, the non-face-to-face class providing device, characterized in that the non-face-to-face class support unit transmits the query summary data to the lecturer terminal using the virtual assistant robot. 5. The method of claim 4,
The non-face-to-face class support department,
Based on the analysis result, the learning participation rate is calculated based on the number of sentence input per learner, the sentence length, and the text content, and the learner with the lowest participation rate is extracted based on the learning participation rate, and then the virtual assistant robot is used for each learning task. A non-face-to-face class providing device, characterized in that it transmits a message encouraging participation in learning to a learner terminal corresponding to a learner with the lowest participation rate by using it. 6. The method of claim 5,
The non-face-to-face class support department,
After extracting only non-class content from the text information based on the analysis result
Selecting only sentences requiring help or support by analyzing the out-of-class content, selecting a category corresponding to the sentences requiring help or support,
Non-face-to-face, characterized in that the out-of-class support information corresponding to the corresponding category is extracted, and the out-of-class support information is transmitted to the learner terminal in which the sentence requiring the help or support is input using the virtual assistant robot class provider. 7. The method of claim 6,
The user pattern generation unit generates a personal learner pattern based on the analysis result of analyzing the personal text information input from the non-face-to-face personal chat room, and based on the analysis result of analyzing the group text information input from the non-face-to-face group chat room create a group learner pattern,
The non-face-to-face class support unit extracts a non-face-to-face group chat room having a group learning pattern having the same pattern as the individual learner pattern or more than a threshold, and responds to individual learners existing in the non-face-to-face personal chat room using the virtual assistant robot Non-face-to-face class providing device, characterized in that it recommends participation in a non-face-to-face group chat room having a group learning pattern having the same pattern as the individual learner pattern or more than a threshold with a learner terminal.

Images