CN117541723A - Vocational education tool scene construction method and system based on meta universe - Google Patents
Vocational education tool scene construction method and system based on meta universe Download PDFInfo
- Publication number
- CN117541723A CN117541723A CN202311539913.6A CN202311539913A CN117541723A CN 117541723 A CN117541723 A CN 117541723A CN 202311539913 A CN202311539913 A CN 202311539913A CN 117541723 A CN117541723 A CN 117541723A
- Authority
- CN
- China
- Prior art keywords
- representing
- education
- joint node
- meta
- word segmentation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000001755 vocal effect Effects 0.000 title claims abstract description 40
- 238000010276 construction Methods 0.000 title claims description 29
- 230000009471 action Effects 0.000 claims abstract description 50
- 238000012549 training Methods 0.000 claims abstract description 38
- 238000000034 method Methods 0.000 claims abstract description 30
- 230000011218 segmentation Effects 0.000 claims description 51
- 230000006870 function Effects 0.000 claims description 36
- 239000013598 vector Substances 0.000 claims description 31
- 239000011159 matrix material Substances 0.000 claims description 26
- 230000015654 memory Effects 0.000 claims description 16
- 230000004913 activation Effects 0.000 claims description 14
- 230000001629 suppression Effects 0.000 claims description 7
- 230000000875 corresponding effect Effects 0.000 claims description 6
- 238000005516 engineering process Methods 0.000 claims description 5
- 238000012546 transfer Methods 0.000 claims description 5
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 238000012937 correction Methods 0.000 claims description 4
- 230000005764 inhibitory process Effects 0.000 claims description 3
- 230000035772 mutation Effects 0.000 claims description 3
- 230000017105 transposition Effects 0.000 claims 1
- 238000011161 development Methods 0.000 abstract description 5
- 238000012545 processing Methods 0.000 abstract description 5
- 230000000694 effects Effects 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 230000009286 beneficial effect Effects 0.000 description 5
- 230000018109 developmental process Effects 0.000 description 4
- 238000010521 absorption reaction Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000002068 genetic effect Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 238000003058 natural language processing Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000008485 antagonism Effects 0.000 description 1
- 230000003190 augmentative effect Effects 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/284—Lexical analysis, e.g. tokenisation or collocates
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
- G06F40/295—Named entity recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/20—Education
- G06Q50/205—Education administration or guidance
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Educational Administration (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Tourism & Hospitality (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Strategic Management (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Educational Technology (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method and a system for constructing a professional education tool scene based on meta universe, which belong to the technical field of data processing, wherein the method comprises the following steps: acquiring descriptive information of a vocational education course; analyzing the description information of the vocational education course, and extracting article information and position information; constructing a virtual article according to the article information; the position information is used as constraint, the highest training efficiency is used as a target, the placement positions of the virtual articles are determined, and a vocational education scene is constructed; guiding a learner to perform action input in the vocational education scene, and recognizing the action input of the learner to perform vocational education training. According to the invention, the professional education scene can be automatically constructed, the manual modeling and programming by an engineer are not required, the development cost of the professional education course is reduced, the difficulty of expanding or updating the meta universe is reduced, and the content of the professional education course can be updated in time.
Description
Technical Field
The invention belongs to the technical field of data processing, and particularly relates to a method and a system for constructing a professional education tool scene based on meta universe.
Background
The meta-universe provides a virtual environment through which users can be immersed in the digitized world, interacting with three-dimensional objects and other virtual entities, through virtual reality and augmented reality techniques. In recent years, meta-space technology has rapidly progressed, and has been applied to various fields such as vocational education.
The metauniverse of the current professional education tool scene is mainly modeled and programmed manually by engineers according to the demands of education courses, and a great deal of engineer working time and technical resources are required, so that high development cost is caused. Also, once a meta-universe scene is created, it may become complicated to expand or update it, and if the vocational education content changes or needs to be updated, it needs to be manually adjusted by an engineer, which may cause the content in the scene to lag behind the actual education needs.
Disclosure of Invention
In order to solve the problem that the metauniverse of the current professional education tool scene is manually modeled and programmed mainly by engineers according to the demands of education courses, a great deal of engineer working time and technical resources are required, and high development cost is caused. In addition, once the meta-universe scene is established, the expansion or update of the meta-universe scene may become complicated, and if the professional education content is changed or needs to be updated, an engineer needs to manually adjust the meta-universe scene, which may cause the technical problem that the content in the scene lags behind the actual education requirement.
First aspect
The invention provides a construction method of a professional education tool scene based on meta universe, which comprises the following steps:
s1: acquiring descriptive information of a vocational education course;
s2: analyzing the description information of the staff education course, and extracting article information and position information;
s3: constructing a virtual article according to the article information;
s4: the position information is used as constraint, the highest training efficiency is used as a target, the placement positions of the virtual articles are determined, and a vocational education scene is constructed;
s5: guiding the learner to perform action input in the vocational education scene, and recognizing the action input of the learner to perform vocational education training.
Second aspect
The invention provides a meta-universe-based professional education tool scene construction system, which comprises a processor and a memory for storing executable instructions of the processor, wherein the processor is used for executing the instructions of the processor; the processor is configured to invoke the instructions stored by the memory to perform the meta-universe based professional education tool scene construction method of the first aspect.
Compared with the prior art, the invention has at least the following beneficial technical effects:
(1) According to the invention, descriptive information of the vocational education course can be analyzed, article information and position information can be extracted, virtual articles are automatically constructed according to the article information, then the position information is taken as constraint, the highest training efficiency is taken as the target, and the placement position of each virtual article is determined, so that the vocational education scene is automatically constructed, the manual modeling and programming by an engineer are not required to be completely relied on, the development cost of the vocational education course is reduced, the difficulty of expanding or updating the meta universe is reduced, and the vocational education course content can be updated in time.
(2) According to the invention, by constructing the professional education scene, the learner is guided to perform action input in the professional education scene, the action input of the learner is identified to perform professional education training, the learner can perform actual operation and interaction in the virtual scene, simulate the real professional environment, improve the practical ability and skill, and simultaneously improve the participation of the learner, excite the learning interest and strengthen the learning power.
Drawings
The above features, technical features, advantages and implementation of the present invention will be further described in the following description of preferred embodiments with reference to the accompanying drawings in a clear and easily understood manner.
Fig. 1 is a schematic flow chart of a construction method of a professional education tool scene based on metauniverse.
Fig. 2 is a schematic structural diagram of a meta-universe-based professional education tool scene construction system provided by the invention.
Detailed Description
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following description will explain the specific embodiments of the present invention with reference to the accompanying drawings. It is evident that the drawings in the following description are only examples of the invention, from which other drawings and other embodiments can be obtained by a person skilled in the art without inventive effort.
Example 1
In one embodiment, referring to fig. 1 of the specification, a schematic flow chart of a meta-universe-based professional education tool scene construction method provided by the invention is shown.
The invention provides a construction method of a professional education tool scene based on meta universe, which comprises the following steps:
s1: and acquiring descriptive information of the vocational education course.
Wherein, the description information includes: course overview, learning objectives, teaching content, teaching methods, assessment methods, and premised knowledge, among others. The description information provides a foundation for the construction of the subsequent meta-universe scene, and ensures that the virtual environment can accurately reflect the requirements and targets of the actual vocational education courses.
Specifically, the vocational education course may be described in text, or the vocational education course may be described in voice.
S2: analyzing the description information of the vocational education course, and extracting article information and position information.
Specifically, natural scientific language technology (Natural Language Processing, NLP) can be adopted to analyze the descriptive information of the vocational education course, and the article information and the position information can be extracted.
In one possible implementation, S2 specifically includes sub-steps S201 and S202:
s201: and performing word segmentation on the descriptive information of the vocational education course.
In one possible implementation, the substep S201 specifically includes grandchild steps S2011 to S2014:
s2011: and performing word segmentation on the current sentence by a forward maximum matching method.
The positive maximum matching method comprises the steps of selecting word segmentation points from left to right, matching the text to be segmented with words in a word stock, selecting the longest matched word as the word segmentation points, and then carrying out the same matching process on the rest texts until all the words of the text are segmented.
S2012: and performing word segmentation on the current sentence by a reverse maximum matching method.
The reverse maximum matching method is opposite to the forward maximum matching method, word segmentation points are selected from right to left, the longest matched word is selected as the word segmentation points, and the same matching process is carried out on the rest text until all word segmentation of the text is completed.
S2013: judging whether the word segmentation results of the forward maximum matching method and the reverse maximum matching method are the same, if so, outputting the word segmentation result of the current sentence, otherwise, entering the next step.
In the invention, whether the word segmentation results of the forward maximum matching method and the reverse maximum matching method are the same is judged, so that the fault tolerance of the system can be enhanced. If the two methods give similar results, the explanatory segmentation is relatively stable. If the results are different, the method can be more flexibly processed, and a more proper word segmentation result can be selected.
S2014: the current sentence is segmented based on the conditional random field.
Wherein a Conditional Random Field (CRF) is a probabilistic graph model that performs word segmentation by considering the relationships of words in context and the conditional probabilities between them. In the word segmentation process, the CRF model learns the relation among words by learning a large amount of corpus data, and then carries out word segmentation on the text according to the relation.
In the invention, the conditional random field can consider the context information, so that the understanding capability of the context is improved, and the word segmentation method based on the conditional random field can further improve the word segmentation accuracy, especially in the process of some complex language structures and scenes with larger ambiguity.
In one possible implementation, S2014 specifically includes:
based on the conditional random field, calculating the word segmentation marks y of the current sentence x i The following conditional probabilities:
；
where x represents the current statement, y i Represents the ith word-segmentation mark, Y represents a word-segmentation mark set, F 1 () Representing the transfer function omega 1 Weight coefficient representing transfer function, F 2 () Representing a state function, omega 2 The weight coefficient representing the state function exp represents the exponential function based on e.
Wherein, the person skilled in the art can set the weight coefficient omega of the transfer function according to the actual situation 1 And a state function omega 2 The size of the weight coefficient of (2) is not limited in the present invention.
Marking y at each word segmentation according to the current sentence x i And selecting a word segmentation mark with the maximum conditional probability under the conditional probability, and segmenting the current sentence.
In the invention, the Chinese word segmentation based on the conditional random field has the advantages of full consideration of global context, flexibility of characteristics, capability of adapting to different tasks through parameter learning, better processing of ambiguity and integration of multi-level information, so that the model can more accurately understand the language structure in the word segmentation process, and the word segmentation quality and adaptability are improved.
S202: and identifying the article information and the position information from the word segmentation result through an entity identification technology.
In one possible implementation, the substep S202 specifically includes grandchild steps S2021 to S2025:
s2021: extracting word segmentation characteristics of words through a BERT pre-training model to obtain word segmentation characteristic sequences，h i Representing the feature of the i-th word segmentation,n represents the total number of word segmentation features.
The BERT pre-training model adopts a transducer architecture, and language representation is learned from a large-scale text corpus through unsupervised learning.
In the present invention, the use of the BERT pre-training model enables a more comprehensive understanding of the context of the word, rather than just considering the characteristics of a single word. This helps to better understand the relationship of item information and location information during entity recognition, improving the context understanding capability.
S2022: the weights of the individual word segmentation features are determined according to the following formula:
；
wherein w is i The weight of the ith word segmentation feature, V the first parameter matrix, W the second parameter matrix, U the third parameter matrix, tanh () the tanh function, c the current event type,representing a matrix dot product operation () T Representing a matrix transpose operation, d representing the embedding dimension.
In the invention, the importance of different features can be adjusted by determining the weight of each word segmentation feature. The model can more flexibly identify the article information and the position information according to task demands and contexts, and the adaptability and generalization capability of the model are improved.
S2023: determining a term embedding vector according to the following formula:
；
where S represents the word embedding vector and Softmax () represents the Softmax function.
In the present invention, the computation of the word embedding vector utilizes a Softmax function, which helps to transform the conditional probabilities of the word under different event types into corresponding embedding vectors. The processing mode can better capture the distribution information of the words in the context, and the representation effect of the article information and the position information is improved.
S2024: the conditional probabilities under various event types are calculated according to the following formula:
；
wherein p is c Representing conditional probability under event type c, sigmoid () represents a Sigmoid function.
In the invention, the entity identification can be more accurately performed by calculating the conditional probability under various event types and selecting the event type with the largest conditional probability. This helps to correctly identify the article information and the position information in the word segmentation result, and improves the accuracy of entity identification.
S2025: according to the conditional probability of the current word under each event type, selecting the event type with the highest conditional probability, carrying out entity identification on the current word, and identifying article information and position information from the word segmentation result.
According to the invention, through combining the BERT pre-training model and the entity recognition technology, accurate recognition of the article information and the position information in the segmentation result is realized. The BERT model is utilized to understand the context, the feature weights are dynamically adjusted, the embedded vectors are calculated to better represent the word distribution, and accurate entity identification is realized through conditional probability selection. The comprehensive application improves the understanding and adaptability of the model to the complex context, and enhances the effective capturing of the article and the position information in the multitasking scene, thereby improving the accuracy and the comprehensiveness of entity identification.
S3: and constructing a virtual article according to the article information.
Specifically, the virtual item may be generated from item information by a digital twinning technique, and may also be automatically generated by generating an antagonism network (GAN) or a Variational Automatic Encoder (VAE), or the like.
S4: and (3) taking the position information as constraint and the highest training efficiency as a target, determining the placement positions of the virtual articles, and constructing a vocational education scene.
In the invention, the virtual articles are placed in an optimal mode, which is beneficial to improving training efficiency. The learner can more easily access and use the required learning resources, reducing unnecessary searches and wasting time. Meanwhile, resource waste can be avoided by reasonably placing virtual articles, each learning element is ensured to be effectively utilized, and the method is beneficial to creating a learning environment with more educational value and effect.
In one possible implementation, S4 specifically includes sub-steps S401 to S404:
s401: setting a training efficiency objective function of a vocational education scene building model by taking the highest training efficiency as a target:
；
wherein, F () represents an objective function, θ represents an article placement vector, T represents a learning duration, λ represents a weight coefficient of the learning duration, and ρ represents an absorptivity.
The absorptivity refers to the knowledge grasping degree of a learner after learning by a course, and can be calculated according to the ratio of the number of learned knowledge points to the number of total knowledge points.
The training efficiency is higher as the learning duration is shorter, and conversely, the training efficiency is lower as the learning duration is longer. The higher the absorption rate, the higher the training efficiency, and conversely, the lower the absorption rate, the lower the training efficiency.
The size of the weight coefficient lambda of the learning duration can be set by a person skilled in the art according to actual conditions, and the invention is not limited.
S402: and setting constraint conditions of the vocational education scene building model according to the position information.
S403: under the constraint of constraint conditions, optimizing the vocational education scene construction model by taking the maximum function value of the training efficiency objective function as the target, and determining the optimal article placement vector.
In the invention, by optimizing the objective function, the system can automatically seek the maximization of efficiency in the process of placing the articles. The scene designed in this way is more likely to transmit more information in the same learning time, so that the learning efficiency is improved.
In one possible implementation, the substep S403 specifically includes grandchild steps S4031 to S4036:
s4031: initializing population Q 1 The population comprises a plurality of individuals X, and each individual X represents a feasible object placement vector theta.
S4032: calculation of initial population Q 1 The fitness value of each individual is a function value of a training efficiency objective function.
S4033: employing elite selection strategy to remove 20% of individuals with lowest fitness value to form new population Q 2 。
S4034: for population Q 2 Performing crossover operation from population H 2 Two individuals are randomly selected as father body to generate a random number, and the random number is combined with the crossover probability p e Comparing the magnitudes, if the random number is smaller than the crossover probability p e Performing cross operation on the parents to generate new individuals so as to form a new population Q 3 The new individuals were generated as follows:
；
wherein Y is 1 、Y 2 Representing a new individual, X 1 Represents a first parent, X 2 Representing the second parent, rand represents a random number between 0 and 1.
S4035: for population Q 3 Performing mutation operation from population Q 3 Randomly selecting an individual as parent, generating a random number, and combining the random number with variation probability p m Comparing the size, if the random number is smaller than the variation probability p m Then the father body is mutated to generate a new individualFormation of a New population Q 4 The new individuals were generated as follows:
；
wherein Y is 3 Representing a new individual, X 3 Representing parent body, X max Represents the individual with the largest fitness value, X min Representing the individual with the smallest fitness value, rand represents a random number between 0 and 1.
S4036: repeating the steps, iterating until the preset iteration times are reached, and outputting a solution with the maximum fitness value as an optimal article placement vector.
According to the invention, the solution space of the object placement vector is comprehensively searched through a genetic algorithm, the individual group is automatically adjusted through the mechanisms of adaptability, diversity and adaptivity, the solution with higher adaptability is reserved, the situation of sinking into a local optimal solution is prevented, and the searching efficiency is improved. Meanwhile, due to the parallelism and automation characteristics of the genetic algorithm, the whole scene construction model is more robust and adaptive, and the professional education scene meeting the aim with highest training efficiency can be constructed efficiently and automatically.
S404: and determining the placement positions of the virtual articles according to the optimal article placement vectors, and constructing a vocational education scene.
According to the invention, the placement position determined based on the optimal article placement vector can enable the layout of the virtual article in the scene to be more optimized, and the target with highest training efficiency is met to the greatest extent. This helps to improve the learning effect of the learner in the virtual environment.
S5: guiding a learner to perform action input in the vocational education scene, and recognizing the action input of the learner to perform vocational education training.
According to the invention, by constructing the professional education scene, the learner is guided to perform action input in the professional education scene, the action input of the learner is identified to perform professional education training, the learner can perform actual operation and interaction in the virtual scene, simulate the real professional environment, improve the practical ability and skill, and simultaneously improve the participation of the learner, excite the learning interest and strengthen the learning power.
In one possible implementation, S5 specifically includes substeps S501 to S506:
s501: and acquiring a joint node data sequence of the learner.
S502: setting action inhibition parameters of each joint node data in the joint node data sequence:
；
wherein a is t The motion suppression parameter at time t, and β represent motion attenuation factors.
In the invention, the setting of the action suppression parameters allows the system to adjust the action response degree of each joint node of the learner, so that the system can be helped to better understand and adapt to individual differences of the learner, and more personalized training experience is provided.
S503: according to the joint node data sequence, calculating action trend parameters:
；
wherein b t Representing action trend parameters at time t, b t-1 The motion trend parameter at time t-1 is represented, gamma represents the motion trend weight coefficient, X t Joint node data, X, representing time t t-1 Joint node data at time t-1 is shown.
The size of the action trend weight coefficient γ can be set by a person skilled in the art according to actual situations, and the invention is not limited.
In the invention, the calculation of the action trend parameters enables the system to recognize the action development trend of the learner, understand the change and evolution of the action, and help the system to more accurately predict the future action of the learner.
S504: correcting the joint node according to the motion suppression parameter and the motion trend parameter:
；
wherein,joint node correction data, a, representing time t t An operation suppression parameter X representing time t t Joint node data, X, representing time t t-1 Joint node data representing time t-1, b t-1 The action trend parameter at time t-1 is shown.
According to the invention, through the comprehensive application of the action inhibition parameters and the action trend parameters, the system can adjust and correct the actions of the learner in real time, thereby being beneficial to correcting potential errors and improving the action accuracy of the learner.
S505: according to the joint node correction data, predicting the joint node:
；
wherein,the joint node prediction data at time t+k is represented, and k represents the prediction step size.
In the invention, the motion of the joint node is predicted to be the estimation of future motion of the learner, which is helpful for the system to recognize possible problems or errors of the learner in advance, even if the learner is righted, and further provides more effective training.
S506: and according to the joint node prediction data, performing action recognition, responding to corresponding actions in the professional education scene, and performing vocational education training.
According to the invention, the system can perform corresponding response change according to the action of the learner, thereby being beneficial to improving the professional skills and application capacity of the learner and further improving the overall training effect.
In one possible implementation, S506 specifically includes:
according to the joint node prediction data, based on long-short-term memory neural network, extracting hidden states of joint nodes at all moments:
；
wherein I is t An activation output vector representing an input gate at time t, sigmoid () representing a Sigmoid activation function, W XI Representing a weight matrix between the joint node data sequence and the input gate, W HI Representing a weight matrix between hidden states and input gates, b I Representing the bias term of the input gate, F t An activation output vector of a forgetting gate at the time t is represented by W XF Weight matrix between data sequence of joint node and forgetting gate, W HF A weight matrix representing the hidden state and forgetting gate, b F Indicating the forgetting of the bias term of the door, O t An activation output vector W representing an output gate at time t XO Representing a weight matrix between the joint node data sequence and the output gate, W HO Representing a weight matrix between hidden states and output gates, C t The activation output vector of the cell memory unit at time t,candidate output vector representing cell memory cell at time t, C t-1 Representing the activation output vector of the cell memory unit at time t-1, and tanh () represents tanh activation function, W XC Representing a weight matrix between a joint node data sequence and a cell storage unit, W HC Representing a weight matrix between hidden states and cell storage units, b C Bias term, h, representing cell memory cell t Represents the hidden state at the time t, h t-1 The hidden state at time t-1 is indicated.
The Long Short-Term Memory neural network (LSTM) is a deep learning model for processing sequence data, and is particularly suitable for capturing Long-Term dependency.
Predicting the probability of each action according to the hidden state of the joint node at each moment:
；
wherein P is i Representing the probability of the ith action, w i Represents the ith action weight parameter, b i The bias parameter representing the ith action, h representing the hidden state.
And selecting the action with the maximum probability value as an action recognition result, and responding to the corresponding action in the professional education scene to perform professional education training.
In the invention, long-term memory neural network is used for helping to capture long-term dependency relationship in the joint node data sequence of the learner, and information is effectively reserved and updated. By doing so, the action mode of the learner can be expressed more accurately, and the accuracy of action recognition is improved.
Example 2
In one embodiment, referring to fig. 2 of the specification, a schematic structural diagram of a meta-universe-based professional education tool scene building system provided by the invention is shown.
The invention provides a meta-universe-based professional education tool scene construction system, which comprises a processor 201 and a memory 202 for storing executable instructions of the processor 201. The processor 201 is configured to invoke the instructions stored in the memory 202 to perform the meta-universe based professional education tool scene construction method of embodiment 1.
The professional education tool scene construction system based on the meta-universe provided by the invention can realize the steps and effects of the professional education tool scene construction method based on the meta-universe in the embodiment 1, and the invention is not repeated for avoiding repetition.
The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The foregoing examples illustrate only a few embodiments of the invention, which are described in detail and are not to be construed as limiting the scope of the invention. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the invention, which are all within the scope of the invention. Accordingly, the scope of protection of the present invention is to be determined by the appended claims.
Claims (10)
1. A meta-universe-based professional education tool scene construction method, comprising the steps of:
s1: acquiring descriptive information of a vocational education course;
s2: analyzing the description information of the staff education course, and extracting article information and position information;
s3: constructing a virtual article according to the article information;
s4: the position information is used as constraint, the highest training efficiency is used as a target, the placement positions of the virtual articles are determined, and a vocational education scene is constructed;
s5: guiding the learner to perform action input in the vocational education scene, and recognizing the action input of the learner to perform vocational education training.
2. The meta-universe-based professional education tool scene construction method according to claim 1, wherein the S2 specifically includes:
s201: word segmentation is carried out on the description information of the staff education courses;
s202: and identifying the article information and the position information from the word segmentation result through an entity identification technology.
3. The meta-universe-based professional education tool scene construction method according to claim 2, wherein the S201 specifically includes:
s2011: word segmentation is carried out on the current sentence through a forward maximum matching method;
s2012: word segmentation is carried out on the current sentence through a reverse maximum matching method;
s2013: judging whether the word segmentation results of the forward maximum matching method and the reverse maximum matching method are the same, if so, outputting the word segmentation result of the current sentence, otherwise, entering the next step;
s2014: the current sentence is segmented based on the conditional random field.
4. The meta-universe based professional education tool scene construction method according to claim 2, wherein the S2014 specifically comprises:
based on the conditional random field, calculating the word segmentation marks y of the current sentence x i The following conditional probabilities:
；
where x represents the current statement, y i Represents the ith word-segmentation mark, Y represents a word-segmentation mark set, F 1 () Representing the transfer function omega 1 Weight coefficient representing transfer function, F 2 () Representing a state function, omega 2 The weight coefficient representing the state function, exp represents the exponential function based on e;
marking y at each word segmentation according to the current sentence x i And selecting a word segmentation mark with the maximum conditional probability under the conditional probability, and segmenting the current sentence.
5. The meta-universe-based professional education tool scene construction method according to claim 2, wherein the S202 specifically includes:
s2021: extracting word segmentation characteristics of words through a BERT pre-training model to obtain word segmentation characteristic sequences，h i Representing the ith word segmentation feature, +.>N represents the total number of word segmentation features;
s2022: the weights of the individual word segmentation features are determined according to the following formula:
；
wherein w is i The weight of the ith word segmentation feature, V the first parameter matrix, W the second parameter matrix, U the third parameter matrix, tanh () the tanh function, c the current event type,representing a matrix dot product operation () T Representing a matrix transposition operation, d representing an embedding dimension;
s2023: determining a term embedding vector according to the following formula:
；
where S represents a word embedding vector, softmax () represents a Softmax function;
s2024: the conditional probabilities under various event types are calculated according to the following formula:
；
wherein p is c Representing conditional probability under event type c, sigmoid () represents a Sigmoid function;
s2025: according to the conditional probability of the current word under each event type, selecting the event type with the highest conditional probability, carrying out entity identification on the current word, and identifying article information and position information from the word segmentation result.
6. The meta-universe-based professional education tool scene construction method according to claim 1, wherein the S4 specifically includes:
s401: setting a training efficiency objective function of a vocational education scene building model by taking the highest training efficiency as a target:
；
wherein F () represents an objective function, θ represents an article placement vector, T represents a learning duration, λ represents a weight coefficient of the learning duration, and ρ represents an absorptivity;
s402: setting constraint conditions of a vocational education scene construction model according to the position information;
s403: under the constraint of the constraint condition, optimizing the construction model of the staff education scene by taking the maximum function value of the training efficiency objective function as a target, and determining the optimal article placement vector;
s404: and determining the placement positions of the virtual articles according to the optimal article placement vectors, and constructing a vocational education scene.
7. The meta-universe based professional education tool scene construction method as claimed in claim 6, wherein the S403 specifically includes:
s4031: initializing population Q 1 The population comprises a plurality of individuals X, wherein each individual X represents a feasible object placement vector theta;
s4032: calculating the initial population Q 1 The fitness value of each individual is a function value of a training efficiency objective function;
s4033: employing elite selection strategy to remove 20% of individuals with lowest fitness value to form new population Q 2 ；
S4034: for population Q 2 Performing crossover operation from population H 2 Two individuals are randomly selected as father body to generate a random number, and the random number is combined with the crossover probability p e Comparing the magnitudes, if the random number is smaller than the crossover probability p e Performing cross operation on the parents to generate new individuals so as to form a new population Q 3 The new individuals were generated as follows:
；
wherein Y is 1 、Y 2 Representing a new individual, X 1 Represents a first parent, X 2 Representing a second parent, rand representing a random number between 0 and 1;
s4035: for population Q 3 Performing mutation operation from population Q 3 Randomly selecting an individual as parent, generating a random number, and combining the random number with variation probability p m Comparing the size, if the random number is smaller than the variation probability p m Performing mutation operation on parent body to generate new individual to form new population Q 4 The new individuals were generated as follows:
；
wherein Y is 3 Representing a new individual, X 3 Representing parent body, X max Represents the individual with the largest fitness value, X min Representing the individual with the smallest fitness value, rand represents a random number between 0 and 1;
s4036: repeating the steps, iterating until the preset iteration times are reached, and outputting a solution with the maximum fitness value as an optimal article placement vector.
8. The meta-universe-based professional education tool scene construction method according to claim 1, wherein the S5 specifically comprises:
s501: acquiring a joint node data sequence of a learner;
s502: setting action inhibition parameters of each joint node data in the joint node data sequence:
；
wherein a is t An operation suppression parameter at time t, t indicating time, and β indicating an operation attenuation factor;
s503: according to the joint node data sequence, calculating action trend parameters:
；
wherein b t Representing action trend parameters at time t, b t-1 The motion trend parameter at time t-1 is represented, gamma represents the motion trend weight coefficient, X t Joint node data, X, representing time t t-1 Joint node data representing time t-1;
s504: correcting the joint node according to the action suppression parameter and the action trend parameter:
；
wherein,joint node correction data, a, representing time t t An operation suppression parameter X representing time t t Joint node data, X, representing time t t-1 Joint node data representing time t-1, b t-1 Action trend parameters at time t-1 are represented;
s505: according to the joint node correction data, predicting the joint node:
；
wherein,joint node prediction data at time t+k, k representing a prediction step length;
s506: and according to the joint node prediction data, performing action recognition, responding to corresponding actions in the professional education scene, and performing vocational education training.
9. The meta-universe based professional education tool scene construction method as claimed in claim 8, wherein S506 specifically includes:
according to the joint node prediction data, based on long-short-term memory neural network, extracting hidden states of joint nodes at all moments:
；
wherein I is t An activation output vector representing an input gate at time t, sigmoid () representing a Sigmoid activation function, W XI Representing a weight matrix between the joint node data sequence and the input gate, W HI Representing a weight matrix between hidden states and input gates, b I Representing the bias term of the input gate, F t An activation output vector of a forgetting gate at the time t is represented by W XF Weight matrix between data sequence of joint node and forgetting gate, W HF A weight matrix representing the hidden state and forgetting gate, b F Indicating the forgetting of the bias term of the door, O t An activation output vector W representing an output gate at time t XO Representing a weight matrix between the joint node data sequence and the output gate, W HO Representing a weight matrix between hidden states and output gates, C t The activation output vector of the cell memory unit at time t,candidate output vector representing cell memory cell at time t, C t-1 Representing the activation output vector of the cell memory unit at time t-1, and tanh () represents tanh activation function, W XC Representing a weight matrix between a joint node data sequence and a cell storage unit, W HC Representing a weight matrix between hidden states and cell storage units, b C Bias term, h, representing cell memory cell t Represents the hidden state at the time t, h t-1 The hidden state at the time t-1 is represented;
predicting the probability of each action according to the hidden state of the joint node at each moment:
；
wherein P is i Representing the probability of the ith action, w i Represents the ith action weight parameter, b i A bias parameter representing an ith action, h representing a hidden state;
and selecting the action with the maximum probability value as an action recognition result, and responding to the corresponding action in the professional education scene to perform professional education training.
10. A meta-universe-based professional education tool scene construction system, comprising a processor and a memory for storing instructions executable by the processor; the processor is configured to invoke the instructions stored in the memory to perform the meta-universe based professional education tool scene construction method of any one of claims 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311539913.6A CN117541723A (en) | 2023-11-19 | 2023-11-19 | Vocational education tool scene construction method and system based on meta universe |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311539913.6A CN117541723A (en) | 2023-11-19 | 2023-11-19 | Vocational education tool scene construction method and system based on meta universe |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117541723A true CN117541723A (en) | 2024-02-09 |
Family
ID=89783720
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311539913.6A Pending CN117541723A (en) | 2023-11-19 | 2023-11-19 | Vocational education tool scene construction method and system based on meta universe |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117541723A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110598770A (en) * | 2019-08-30 | 2019-12-20 | 华中师范大学 | Multi-space fusion learning environment construction method and device |
| CN110909531A (en) * | 2019-10-18 | 2020-03-24 | 平安科技（深圳）有限公司 | Method, device, equipment and storage medium for discriminating information security |
| CN115933930A (en) * | 2022-11-16 | 2023-04-07 | 福建天晴数码有限公司 | Method, terminal and device for analyzing attention of learning object in education meta universe |
| CN116152020A (en) * | 2023-01-08 | 2023-05-23 | 福州卓立信息科技有限公司 | IOT education platform based on meta universe and construction method thereof |
| US20230343326A1 (en) * | 2022-04-26 | 2023-10-26 | Walmart Apollo, Llc | Systems and methods to identify products from verbal utterances |
-
2023
- 2023-11-19 CN CN202311539913.6A patent/CN117541723A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110598770A (en) * | 2019-08-30 | 2019-12-20 | 华中师范大学 | Multi-space fusion learning environment construction method and device |
| CN110909531A (en) * | 2019-10-18 | 2020-03-24 | 平安科技（深圳）有限公司 | Method, device, equipment and storage medium for discriminating information security |
| US20230343326A1 (en) * | 2022-04-26 | 2023-10-26 | Walmart Apollo, Llc | Systems and methods to identify products from verbal utterances |
| CN115933930A (en) * | 2022-11-16 | 2023-04-07 | 福建天晴数码有限公司 | Method, terminal and device for analyzing attention of learning object in education meta universe |
| CN116152020A (en) * | 2023-01-08 | 2023-05-23 | 福州卓立信息科技有限公司 | IOT education platform based on meta universe and construction method thereof |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112163426B (en) | Relationship extraction method based on combination of attention mechanism and graph long-time memory neural network | |
| CN113544703B (en) | Efficient off-policy credit allocation | |
| CN108647233B (en) | Answer sorting method for question-answering system | |
| Raissi et al. | Deep multi-fidelity Gaussian processes | |
| CN112116092B (en) | Interpretable knowledge level tracking method, system and storage medium | |
| CN113852432B (en) | Spectrum Prediction Sensing Method Based on RCS-GRU Model | |
| CN113361685B (en) | Knowledge tracking method and system based on learner knowledge state evolution expression | |
| CN113591988B (en) | Knowledge cognitive structure analysis method, system, computer equipment, medium and terminal | |
| CN110210656B (en) | Shared bicycle flow prediction method and system based on station behavior analysis | |
| CN110377707B (en) | Cognitive diagnosis method based on depth item reaction theory | |
| CN115545160B (en) | Knowledge tracking method and system for multi-learning behavior collaboration | |
| Ku et al. | A study of the Lamarckian evolution of recurrent neural networks | |
| CN114254127A (en) | Student ability portrayal method and learning resource recommendation method and device | |
| CN115018193A (en) | Time series wind energy data prediction method based on LSTM-GA model | |
| CN115906997A (en) | Knowledge tracking method based on comparative learning and memory mechanism | |
| CN112508177A (en) | Network structure searching method and device, electronic equipment and storage medium | |
| CN114548591A (en) | Time sequence data prediction method and system based on hybrid deep learning model and Stacking | |
| CN114971066A (en) | Knowledge tracking method and system integrating forgetting factor and learning ability | |
| CN111985560B (en) | Knowledge tracking model optimization method, system and computer storage medium | |
| CN117541723A (en) | Vocational education tool scene construction method and system based on meta universe | |
| CN115984025A (en) | Influence propagation estimation method and system based on deep learning graph network model | |
| CN109815323B (en) | Human-computer interaction training question-answer generation algorithm | |
| Lockett et al. | Temporal convolution machines for sequence learning | |
| CN113051353A (en) | Attention mechanism-based knowledge graph path reachability prediction method | |
| CN112750466A (en) | Voice emotion recognition method for video interview |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |