CN117994101A - Teaching design generation method and device based on large language model - Google Patents
Teaching design generation method and device based on large language model Download PDFInfo
- Publication number
- CN117994101A CN117994101A CN202410399307.7A CN202410399307A CN117994101A CN 117994101 A CN117994101 A CN 117994101A CN 202410399307 A CN202410399307 A CN 202410399307A CN 117994101 A CN117994101 A CN 117994101A
- Authority
- CN
- China
- Prior art keywords
- teaching
- model
- content
- design
- contents
- 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.)
- Granted
Links
- 238000013461 design Methods 0.000 title claims abstract description 150
- 238000000034 method Methods 0.000 title claims abstract description 83
- 238000009877 rendering Methods 0.000 claims abstract description 77
- 238000012216 screening Methods 0.000 claims abstract description 15
- 238000012549 training Methods 0.000 claims description 31
- 238000004590 computer program Methods 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 14
- 238000004458 analytical method Methods 0.000 claims description 7
- 238000010801 machine learning Methods 0.000 abstract description 2
- 230000000694 effects Effects 0.000 description 11
- 238000010586 diagram Methods 0.000 description 7
- 238000004891 communication Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 5
- 230000015572 biosynthetic process Effects 0.000 description 4
- 238000010276 construction Methods 0.000 description 4
- 238000011161 development Methods 0.000 description 4
- 238000011156 evaluation Methods 0.000 description 4
- 238000005457 optimization Methods 0.000 description 4
- 240000002853 Nelumbo nucifera Species 0.000 description 3
- 235000006508 Nelumbo nucifera Nutrition 0.000 description 3
- 235000006510 Nelumbo pentapetala Nutrition 0.000 description 3
- 230000008520 organization Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000009897 systematic effect Effects 0.000 description 3
- 230000006978 adaptation Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000013480 data collection Methods 0.000 description 2
- 230000008451 emotion Effects 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 238000002372 labelling Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 239000007787 solid Substances 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 239000000969 carrier Substances 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000004821 distillation Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Landscapes
- Machine Translation (AREA)
- Electrically Operated Instructional Devices (AREA)
Abstract
The invention provides a teaching design generation method and device based on a large language model, belonging to the technical field of machine learning, wherein the method comprises the following steps: generating a basic document by using the preliminary generation model; based on the basic document, searching a subject data knowledge base to obtain search contents, wherein the subject data knowledge base is a knowledge base of a subject corresponding to the basic document, and the search contents comprise a plurality of teaching contents; based on the retrieval content, performing color rendering by using an expression alignment model to obtain color rendering content, wherein the expression alignment model is a large language model applicable to subjects, and the color rendering content comprises a plurality of color rendering teaching contents; screening high-quality teaching contents from the color-rendering contents based on preset scoring criteria; based on the high-quality teaching content, a teaching design is generated. The teaching design generation method, the device, the electronic equipment and the storage medium provided by the invention reduce the labor cost and the time cost of teaching design writing.
Description
Technical Field
The invention belongs to the technical field of machine learning, and particularly relates to a teaching design generation method and device based on a large language model.
Background
The teaching design describes links such as teaching targets, processes and evaluation through characters, and is a text result of the lesson preparation process of a teacher. The teaching system reflects the pre-planning of teachers on classroom teaching, and is one of key factors influencing teaching effects. Thus, generating a high quality teaching design is particularly important in teaching processes.
The current method for generating the teaching design is mainly characterized in that teaching design is manually written by professionals such as teachers according to teaching targets, teaching contents and the like. But based on teaching design is compiled to professional personnel such as teacher, not only need great human cost and time cost, the teaching design of compiling still receives the influence of the level of compiling personnel.
Therefore, how to efficiently generate reliable teaching designs is a technical problem to be solved.
Disclosure of Invention
The invention provides a teaching design generating method and device based on a large language model, which are used for solving the defects of low efficiency and poor quality of the teaching design generating method in the prior art and realizing efficient and reliable generation of the teaching design.
In a first aspect, the present invention provides a method for generating a teaching design based on a large language model, including:
generating a basic document by using a preliminary generation model, wherein the preliminary generation model is a universal large language model;
Searching a subject data knowledge base based on the basic document to obtain search content, wherein the subject data knowledge base is a subject knowledge base corresponding to the basic document, and is constructed by utilizing a subject data deconstructing scheme based on subject data, and the search content comprises a plurality of teaching contents;
Based on the search content, performing color rendering by using an expression alignment model to obtain color rendering content, wherein the expression alignment model is a large language model applicable to the subject, and the color rendering content comprises a plurality of color rendering teaching contents;
screening high-quality teaching contents from the color-rendering contents based on preset scoring standards;
and generating a teaching design based on the high-quality teaching content.
According to the teaching design generation method based on the large predictive model provided by the invention, after the high-quality teaching content is screened out from the color rendering content based on the preset scoring standard, the method further comprises the following steps:
and if the score of the high-quality teaching content is lower than a preset threshold value, returning to the step of generating the basic document by using the preliminary generation model.
According to the teaching design generation method based on the large predictive model provided by the invention, the teaching design is generated based on the high-quality teaching content, and the method comprises the following steps:
integrating the high-quality teaching contents to obtain integrated teaching contents;
And based on the integrated teaching content, performing color rendering by using the expression alignment model to obtain the teaching design.
According to the teaching design generation method based on the large predictive model, the expression alignment model is obtained based on training of the following steps:
after data enhancement is carried out on sample data, a training data set is constructed based on the sample data and the data after data enhancement;
and training the pre-training model by using the training data set based on a preset template, and then performing fine adjustment on the model obtained by training by using a preset fine adjustment method to obtain the expression alignment model.
According to the teaching design generation method based on the large predictive model, when the discipline is Chinese, the discipline data comprises teaching designs, text and operation of each text, and the teaching designs comprise learning condition analysis, teaching targets, teaching processes and operation labels.
According to the teaching design generation method based on the large predictive model, the discipline data knowledge base is constructed based on the following steps:
Deconstructing teaching designs, text and operation of the lessons by using a discipline data deconstructing scheme to obtain deconstructing results;
And constructing a discipline data knowledge base based on the deconstructed result.
In a second aspect, the present invention further provides a large language model-based teaching design generating device, including:
A document generation module for: generating a basic document by using a preliminary generation model, wherein the preliminary generation model is a universal large language model;
A content retrieval module for: searching a subject data knowledge base based on the basic document to obtain search content, wherein the subject data knowledge base is a subject knowledge base corresponding to the basic document, and is constructed by utilizing a subject data deconstructing scheme based on subject data, and the search content comprises a plurality of teaching contents;
The content color rendering module is used for: based on the search content, performing color rendering by using an expression alignment model to obtain color rendering content, wherein the expression alignment model is a large language model applicable to the subject, and the color rendering content comprises a plurality of color rendering teaching contents;
A content screening module for: screening high-quality teaching contents from the color-rendering contents based on preset scoring standards;
the teaching plan generating module is used for: and generating a teaching design based on the high-quality teaching content.
In a third aspect, the present invention provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of any one of the above-described large language model based tutorial design generation methods when the program is executed.
In a fourth aspect, the present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the large language model based teaching design generation method as described in any of the above.
In a fifth aspect, the present invention also provides a computer program product comprising a computer program which, when executed by a processor, implements the steps of the large language model based teaching design generation method as described in any of the above.
The invention provides a teaching design generating method and a device based on a large language model, the method firstly uses a preliminary generating model to generate a basic document, the preliminary generating model is a universal large language model, that is, the general model is used to generate the basic document, thereby realizing a teaching scheme for quickly generating a foundation, ensuring the preliminary formation of the teaching design and providing possibility for subsequent optimization; searching a subject data knowledge base based on the basic document to obtain search content, wherein the subject data knowledge base is a subject knowledge base corresponding to the basic document, the subject data knowledge base is constructed based on subject data by using a subject data deconstructing scheme, the search content comprises a plurality of teaching contents, and a subject foundation is provided for subsequent teaching design by using knowledge in the subject data knowledge base; based on the search content, performing color rendering by using an expression alignment model to obtain color rendering content, wherein the expression alignment model is a large language model applicable to the discipline, the color rendering content comprises a plurality of color rendering teaching contents, the color rendering of the expression alignment model enriches the teaching design content, and the specialty and accuracy of the expression are ensured; based on preset scoring standards, high-quality teaching contents are screened from the color-rendering contents, so that the quality and practicability of teaching design are ensured; and generating a teaching design based on the high-quality teaching content. Based on the method, compared with manual writing, the method can obviously reduce the manpower and time cost and efficiently generate the professional and accurate teaching design of the content.
Drawings
In order to more clearly illustrate the technical solutions of the present invention or of the prior art, the following sections will briefly introduce illustrations cited in the examples or the prior art documents. It is evident that the illustrations provided are merely a few embodiments of the present invention and that other relevant illustrations can be derived based on these illustrations without any inventive work to those of ordinary skill in the art.
FIG. 1 is a schematic flow chart of a large language model-based teaching design generation method provided by the invention;
FIG. 2 is a second flow chart of the teaching design generating method based on the large language model provided by the invention;
FIG. 3 is a schematic diagram of an algorithm flow for expressing an alignment model provided by the invention;
FIG. 4 is a schematic diagram of the construction flow of the knowledge base of the Chinese subject matter provided by the invention;
FIG. 5 is a schematic diagram of a large language model based teaching design generating device according to the present invention;
Fig. 6 is a schematic structural diagram of an electronic device provided by the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be noted that in the description of embodiments of the present invention, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element. The orientation or positional relationship indicated by the terms "upper", "lower", etc. are based on the orientation or positional relationship shown in the drawings, are merely for convenience of description and to simplify the description, and are not indicative or implying that the apparatus or elements in question must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as limiting the present invention. Unless specifically stated or limited otherwise, the terms "mounted," "connected," and "coupled" are to be construed broadly, and may be, for example, fixedly connected, detachably connected, or integrally connected; can be mechanically or electrically connected; can be directly connected or indirectly connected through an intermediate medium, and can be communication between two elements. The specific meaning of the above terms in the present invention can be understood by those of ordinary skill in the art according to the specific circumstances.
The terms "first," "second," and the like in this specification are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged, as appropriate, such that embodiments of the present invention may be implemented in sequences other than those illustrated or described herein, and that the objects identified by "first," "second," etc. are generally of a type, and are not limited to the number of objects, such as the first object may be one or more. In addition, "and/or" indicates at least one of the connected objects, and the character "/", generally indicates that the associated object is an "or" relationship.
The teaching design is a systematic decision-making activity on aspects of teaching targets, teaching processes, learning modes, teaching resources, environments, teaching evaluation and the like, is a systematic conception on the teaching activity, and has the following characteristics:
Is a technical process for generating learning experience and learning environment and improving the efficiency and interest of a learner in obtaining specific knowledge and skills;
Is a research activity with discipline, creativity and decision-making property;
Based on the scientific theory of teaching and learning, a wide range of activities are involved.
The teaching design is a basic link that teaching activities can be smoothly unfolded, students can be promoted to effectively learn and talent cultivation can be realized.
The teaching design is an application text oriented to a teaching process, and not only needs to consider teaching styles, but also needs to consider discipline characteristics. Therefore, in order to break through the technical bottlenecks of shallow professional content, dead plates of teaching design style, lack of teaching evaluation guidance and the like faced by the application of the general large language model in the education field, the following embodiments are proposed for the construction of the expression alignment model (Expression Alignment Model, EA).
The method and apparatus for generating a teaching design based on a large language model according to the embodiments of the present invention are described below with reference to fig. 1 to 6.
FIG. 1 is a schematic flow chart of a large language model-based teaching design generating method according to the present invention, as shown in FIG. 1, including but not limited to the following steps:
s110, generating a basic document by using a preliminary generation model, wherein the preliminary generation model is a universal large language model.
For ease of understanding, the following examples are described in terms of the generation of a chinese teaching design.
The language teaching design is generalized speaking, and is the whole teaching arrangement and planning of language course learning; in a narrow sense, the teaching device refers to a teaching theme, a teaching unit or a classroom teaching design, and is the basis for effectively expanding Chinese courses and effectively improving the core literacy of students. The Chinese teaching design is written, and Chinese teaching practice is developed based on the Chinese teaching design, so that the teaching device is a core link for realizing professional development of Chinese teachers and improving teaching ability. The high-quality data marked by a small amount of experts are utilized to develop deep learning training based on an open-source general large model, and a high-quality and high-level Chinese teaching design is generated, so that the method has important practical and theoretical significance. In terms of Chinese teaching practice, the education large model is utilized to automatically generate a Chinese teaching design, so that teaching resource sources of Chinese courses can be greatly expanded, effective teaching resources are enriched, and a wide resource field is provided for teacher teaching and student learning; in terms of professional development of teachers, automatic generation of high-quality and high-level Chinese teaching designs can provide effective expert introduction for all Chinese teachers through high-quality man-machine cooperative dialogue, promote the Chinese teachers to innovate teaching concepts, promote understanding ability and practice ability of teaching, and further develop teaching personality of the teachers, enrich and deepen teaching practice; in the case of educational technology, deep fusion of education and technology can be facilitated, and the continuous transition of technology from educational tools and carriers to educational content generation is facilitated, innovating the technology and educational fusion mode; in terms of theoretical development, the automatic generation of Chinese teaching design can promote the enrichment and deepening of Chinese courses and teaching research theory, promote the digital human research field to develop from calculation to intelligent generation, and promote the development of digital human theory.
It should be noted that, the execution body of the teaching design generating method based on the large language model provided in the embodiment of the present invention may be a server, a computer device, such as a mobile phone, a tablet computer, a notebook computer, a palm computer, an ultra-mobile personal computer (UMPC), a netbook, a Personal Digital Assistant (PDA), or the like.
The large language model is a deep learning model trained (Large Language Models, LLM) based on massive text data, and the type of large language model used herein is not limited.
The universal large language model can be used for quickly generating a basic teaching scheme, so that the preliminary formation of teaching designs is ensured, and the possibility is provided for subsequent optimization and individuation.
And S120, searching a subject data knowledge base based on the basic document to obtain search contents, wherein the subject data knowledge base is constructed based on subject data by using a subject data deconstructing scheme, and the subject data knowledge base is a knowledge base of subjects corresponding to the basic document, and the search contents comprise a plurality of teaching contents.
Taking the Chinese subject as an example, the subject data knowledge base, i.e. the knowledge base corresponding to the Chinese subject, is the knowledge base corresponding to the elements forming the basic document, and the knowledge base stores the related knowledge of the Chinese subject, i.e. the teaching resources related to the Chinese subject, which can include but are not limited to text, operation, high-quality teaching design and other knowledge. The discipline knowledge is systematically arranged and generalized to provide a solid discipline basis for subsequent teaching designs.
In particular, teaching resources cover a variety of categories such as teaching designs, jobs, text of lesson appreciation, and the like. Specifically, numerous items are encompassed under each category: for example, the educational design section encompasses multiple educational designs; the operation part comprises a series of exercises related to courses; the text of text is intended to focus on a plurality of articles.
For Chinese subjects, the subject knowledge base may include knowledge about each text, and the search content includes a plurality of teaching contents of the text corresponding to the base text.
And S130, based on the search content, performing color rendering by using an expression alignment model to obtain color rendering content, wherein the expression alignment model is a large language model applicable to the subject, and the color rendering content comprises a plurality of color rendering teaching contents.
For the Chinese subject, namely, the large language model suitable for the Chinese subject is used for color rendering of the search content, so that teaching design content is enriched, the generated teaching design accords with the education and teaching rule of the Chinese subject, the education concept of the Chinese subject is reflected, and the specialty and accuracy of the text expression of the teaching design are ensured.
Here, the expression alignment model is obtained based on high-quality teaching resource training of the corresponding subject, and the kind of the large language model used for the expression alignment model is not limited.
And S140, screening high-quality teaching contents from the color-rendering contents based on preset scoring criteria.
Specifically, based on preset scoring criteria, scoring and sorting are carried out on the teaching contents after each color rendering, and the teaching contents with optimal value are identified and selected to serve as high-quality teaching contents.
The preset scoring standard is used as a quantitative standard, which is helpful for judging the quality of the teaching design and provides a basis for generating the high-quality teaching design.
In some embodiments, the teaching content reaching a certain scoring value is selected from the color rendering content, and the teaching design reaching a preset scoring value is selected from the teaching design content, taking the teaching content including the teaching design content as an example.
In some embodiments, a certain number of tutorials are selected from the color rendering, for example, the tutorials include tutorial designs, that is, a predetermined number of tutorial designs are selected from the tutorial designs.
The preset score and the preset quantity can be adaptively adjusted according to actual use requirements.
And S150, generating teaching design based on the high-quality teaching content.
The teaching design generating method based on the large language model provided by the embodiment of the invention firstly uses the preliminary generating model to generate the basic document, wherein the preliminary generating model is a universal large language model, that is, the general model is used to generate the basic document, so that a teaching scheme for quickly generating the basis is realized, the preliminary formation of the teaching design is ensured, and the possibility is provided for the subsequent optimization; searching a subject data knowledge base based on the basic document to obtain search content, wherein the subject data knowledge base is a subject knowledge base corresponding to the basic document, the subject data knowledge base is constructed based on subject data by using a subject data deconstructing scheme, the search content comprises a plurality of teaching contents, and a subject foundation is provided for subsequent teaching design by using knowledge in the subject data knowledge base; based on the search content, performing color rendering by using an expression alignment model to obtain color rendering content, wherein the expression alignment model is a large language model applicable to the discipline, the color rendering content comprises a plurality of color rendering teaching contents, the color rendering of the expression alignment model enriches the teaching design content, and the specialty and accuracy of the expression are ensured; based on preset scoring standards, high-quality teaching contents are screened from the color-rendering contents, so that the quality and practicability of teaching design are ensured; and generating a teaching design based on the high-quality teaching content. Based on the method, compared with manual writing, the method can obviously reduce the manpower and time cost and efficiently generate the professional and accurate teaching design of the content.
In an alternative embodiment, after the filtering out the high-quality teaching content from the color-rendering content based on the preset scoring criteria, the method further includes:
And S141, if the score of the high-quality teaching content is lower than a preset threshold value, returning to S110 again.
Specifically, after the high-quality teaching content is screened, if the score of the high-quality teaching content is lower than a preset threshold value, the teaching design is required to be further improved and perfected, and S110 to S140 are executed again; if the score of the high-quality teaching content is greater than or equal to the preset threshold, S150 is executed.
According to the teaching design generation method based on the large language model, high-quality teaching contents with low scores are generated and moistened again, and the quality and practicality of the finally generated teaching design are ensured.
Based on any of the foregoing embodiments, generating a teaching design based on the premium teaching content includes:
integrating the high-quality teaching contents to obtain integrated teaching contents;
And based on the integrated teaching content, performing color rendering by using the expression alignment model to obtain the teaching design.
It will be appreciated that after fine-tuning and integration of the teaching content, further refinement and promotion through the application of the expression alignment model is critical. This process ensures that the resulting educational design is not only structurally complete, but also exhibits profound expertise in content.
Based on any of the above embodiments, the expression alignment model is trained based on the following steps:
after data enhancement is carried out on sample data, a training data set is constructed based on the sample data and the data after data enhancement;
and training the pre-training model by using the training data set based on a preset template, and then performing fine adjustment on the model obtained by training by using a preset fine adjustment method to obtain the expression alignment model.
The teaching is a science and an art, the teaching art is a creative teaching activity with unique style, which is implemented by teachers on classes according to teaching rules and aesthetic scale requirements, and means such as language, expression, action, psychological activities, image organization, regulation and control are flexibly applied, so that the function of teaching emotion is fully exerted, and the best teaching effect is obtained. The teaching design is taken as an application text for describing links such as teaching targets, teaching processes, teaching evaluation and the like by words, and is essentially systematic planning of teaching activities. Thus, the text expression style of the teaching design is called expression alignment. Obviously, this is a requirement for text high-level semantics. In order to generate text conforming to expression alignment, a large amount of high-quality teaching design files and a large amount of calculation force are necessary, and are difficult to realize in practical application.
Aiming at the problems, a fine tuning training method utilizing a prompt project and a large language model is provided, an expression alignment model conforming to education and teaching rules is obtained based on limited computing power resource training, a text conforming to the education and teaching rules and reflecting advanced education ideas is generated on the basis of a general large language model, and the generation of the expression style of a teaching design is ensured.
Fig. 3 is a schematic flow chart of an algorithm for expressing an alignment model, as shown in fig. 3, sample data is subject data, and since the data size of the sample data is smaller, the characteristics of language and text styles of teaching design are analyzed, the subject data and subject expert cooperate, the data size is expanded by using a data enhancement technology based on a large language model, and a training data set (EA training set) is constructed, so that a model obtained by training based on the training data set has a better effect, the generalization capability of the model is enhanced, and the robustness of the model is improved.
The method is characterized in that training data are analyzed from the teaching angles of teaching concepts, core literacy culture targets and the like, a promt template is designed, and the promt template can be constructed manually or automatically, for example, by using a discrete template automatic construction method. The constructed template of promts constitutes a set of promts. Based on the prompt set, selecting a proper large language model as a pre-training model, and performing fine-tuning training by using a preset fine-tuning method to obtain an expression alignment model.
Alternatively, expression alignment models are obtained by fine tuning training using LoRA (Low-Rand Adaptation, low-order Adaptation) methods.
Optionally, in cooperation with the subject matter expert, the large language model is cooperated, the internal rule of the training data set is analyzed, and priori knowledge is introduced to optimize the expression alignment model, so that a final expression alignment model EA_LLM is obtained. Specifically, priori knowledge can be introduced in the input level, namely, data preprocessing and feature extraction are performed based on the priori knowledge; a priori knowledge can also be introduced at the model structure level, i.e. learning the visualizations of the components by means of an attention mechanism; the prior knowledge can also be introduced in the network parameter level, and the knowledge is integrated into the network parameters, such as the model size is compressed by a model distillation method; a priori knowledge may also be introduced at the target constraint level.
According to the teaching design generation method based on the large language model, the expression alignment model conforming to the education and teaching rules is obtained based on limited computing power resource training by utilizing the prompt engineering and the fine tuning training method of the large language model.
Based on any of the above embodiments, when the subject is a Chinese, the subject data includes teaching designs, text and operation for text review, including learning emotion analysis, teaching goals, teaching processes, and operation labels.
The discipline data knowledge base is the basis of a teaching design generating task, fig. 4 is a schematic diagram of a construction flow of the Chinese discipline data knowledge base provided by the invention, and as shown in fig. 4, the discipline data knowledge base negotiates with a Chinese discipline expert to design a discipline data collection and labeling scheme for Chinese teaching design, and the specific used discipline data collection and labeling scheme is not limited herein; then, combining the characteristics of the large language model and the principle of the search enhancement method, designing a discipline data knowledge base/discipline data deconstruction scheme of the task for generating Chinese teaching design; according to the characteristics of the discipline data and the deconstructing scheme, a discipline data analysis deconstructing module is designed to realize automatic analysis of the discipline data; analyzing the characteristics of the discipline data, selecting a knowledge base of an applicable type from SQL, redis, DB2 and the like, and designing a structural organization module according to the discipline rules and the knowledge base organization management characteristics; the comprehensive design of the classification storage module is used for constructing and obtaining a subject data knowledge base oriented to the language teaching generating task, and the support is provided for constructing a language teaching design generating model based on retrieval.
The constructed discipline data knowledge base comprises high-quality teaching design, text and operation for appreciating of each text, wherein the teaching design covers four modules of learning condition analysis, teaching targets, teaching processes and operation according to teaching flow.
Further, the discipline repository is constructed based on the steps of:
Deconstructing teaching designs, text and operation of the lessons by using a discipline data deconstructing scheme to obtain deconstructing results;
And constructing a discipline data knowledge base based on the deconstructed result.
Specifically, text for lesson appreciation is divided into fourteen categories according to the scheme of table 1 below.
Table 1 text content deconstructing scheme
As shown in table 1 above, the 14 class dimensions respectively include a plurality of sub-dimensions, i.e., set up learning tasks or questions, directed to specific content in the sections of learning context analysis, teaching goal statement, analysis explanation of learning tasks/questions, clarity, etc.
It can be understood that when new knowledge points need to be added or existing knowledge points need to be modified, only the new knowledge points need to be added or modified or deleted in the corresponding sub-dimensions, and no large modification is needed to be made to other parts. In addition, each knowledge point exists independently, so that the relation between the knowledge points can be adjusted according to actual requirements so as to adapt to different use scenes.
Taking the example of "the moonlight of lotus pool", the scale of the knowledge points of the lesson in the discipline knowledge base is shown in tables 2 and 3 below.
Table 2 knowledge points of the moonlight of lotus pool
Table 3 deconstructed scheme for text content of the text of the "moonlight of the lotus pool
Optionally, the elements of different types are stored in different folders in a classified mode, the directory level of the folders is consistent with the level relation of the elements, so that a user can understand the content and the relevance of each module conveniently, meanwhile, the user can jump to a required page directly through the directory level, progressive searching layer by layer is not needed, and the searching speed is greatly improved.
Fig. 2 is a second flow chart of the teaching design generating method based on a large language model provided by the present invention, and for convenience of understanding, the following describes the teaching design generating method based on a large language model preferably with reference to fig. 2:
establishing a discipline data knowledge base (Course Information, CI-KB) based on diversified discipline data such as text, operation, high-quality teaching design and the like, so that discipline knowledge is systematically arranged and generalized, and a solid discipline basis is provided for subsequent teaching design;
The general large model LLM is utilized to quickly generate a basic teaching scheme (basic document), so that preliminary formation of teaching design is ensured, and possibility is provided for subsequent optimization and individuation;
On the basis of basic teaching design, the subject data knowledge base is further searched, and the search content is subjected to specialized color rendering by using an expression alignment model EA-LLM, so that the content of the teaching design is enriched, and the specialty and accuracy of expression are ensured;
Sorting and scoring the moistened content, and identifying and selecting the most valuable teaching content;
If the score is lower than a preset threshold k, the teaching design is required to be further improved and perfected, the generation and color rendering of the teaching design are required to be carried out again, and the quality and the practicability of the teaching design are ensured;
Through screening and optimizing of all the steps, the final teaching design content is integrated, and the expressed alignment model EA-LLM is utilized again for color rendering, so that the output teaching design is complete and has high specificity.
The teaching design generating device based on the large language model provided by the embodiment of the invention is described below, and the teaching design generating device based on the large language model described below and the teaching design generating method based on the large language model described above can be correspondingly referred to each other.
Fig. 5 is a schematic structural diagram of a large language model-based teaching design generating device provided by the present invention, and as shown in fig. 5, the present invention also provides a large language model-based teaching design generating device, which may include:
a document generation module 510 for: generating a basic document by using a preliminary generation model, wherein the preliminary generation model is a universal large language model;
Content retrieval module 520 for: searching a subject data knowledge base based on the basic document to obtain search content, wherein the subject data knowledge base is a subject knowledge base corresponding to the basic document, and is constructed by utilizing a subject data deconstructing scheme based on subject data, and the search content comprises a plurality of teaching contents;
A content rendering module 530 for: based on the search content, performing color rendering by using an expression alignment model to obtain color rendering content, wherein the expression alignment model is a large language model applicable to the subject, and the color rendering content comprises a plurality of color rendering teaching contents;
A content screening module 540 for: screening high-quality teaching contents from the color-rendering contents based on preset scoring standards;
a teaching plan generating module 550 for: and generating a teaching design based on the high-quality teaching content.
It should be noted that, in the teaching design generating device based on a large language model provided in the embodiment of the present invention, the teaching design generating method based on a large language model described in any one of the foregoing embodiments may be executed during specific operation, and details of this embodiment are not repeated.
Fig. 6 is a schematic structural diagram of an electronic device according to the present invention, and as shown in fig. 6, the electronic device may include: processor 610, communication interface (Communications Interface) 620, memory 630, and communication bus 640, wherein processor 610, communication interface 620, memory 630 communicate with each other via communication bus 640. Processor 610 may invoke logic instructions in memory 630 to perform a large language model based tutorial design generation method comprising: generating a basic document by using a preliminary generation model, wherein the preliminary generation model is a universal large language model; searching a subject data knowledge base based on the basic document to obtain search content, wherein the subject data knowledge base is a knowledge base of subjects corresponding to the basic document, and the search content comprises a plurality of teaching contents; based on the search content, performing color rendering by using an expression alignment model to obtain color rendering content, wherein the expression alignment model is a large language model applicable to the subject, and the color rendering content comprises a plurality of color rendering teaching contents; screening high-quality teaching contents from the color-rendering contents based on preset scoring standards; and generating a teaching design based on the high-quality teaching content.
Further, the logic instructions in the memory 630 may be implemented in the form of software functional units and stored in a computer-readable storage medium when sold or used as a stand-alone product. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a usb disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
In another aspect, the present invention also provides a computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions which, when executed by a computer, are capable of executing the large language model based teaching design generation method provided by the above embodiments, the method comprising: generating a basic document by using a preliminary generation model, wherein the preliminary generation model is a universal large language model; searching a subject data knowledge base based on the basic document to obtain search content, wherein the subject data knowledge base is a subject knowledge base corresponding to the basic document, and is constructed by utilizing a subject data deconstructing scheme based on subject data, and the search content comprises a plurality of teaching contents; based on the search content, performing color rendering by using an expression alignment model to obtain color rendering content, wherein the expression alignment model is a large language model applicable to the subject, and the color rendering content comprises a plurality of color rendering teaching contents; screening high-quality teaching contents from the color-rendering contents based on preset scoring standards; and generating a teaching design based on the high-quality teaching content.
In still another aspect, the present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, is implemented to perform the large language model based teaching design generation method provided in the above embodiments, the method comprising: generating a basic document by using a preliminary generation model, wherein the preliminary generation model is a universal large language model; searching a subject data knowledge base based on the basic document to obtain search content, wherein the subject data knowledge base is a subject knowledge base corresponding to the basic document, and is constructed by utilizing a subject data deconstructing scheme based on subject data, and the search content comprises a plurality of teaching contents; based on the search content, performing color rendering by using an expression alignment model to obtain color rendering content, wherein the expression alignment model is a large language model applicable to the subject, and the color rendering content comprises a plurality of color rendering teaching contents; screening high-quality teaching contents from the color-rendering contents based on preset scoring standards; and generating a teaching design based on the high-quality teaching content.
The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. A teaching design generation method based on a large language model is characterized by comprising the following steps:
generating a basic document by using a preliminary generation model, wherein the preliminary generation model is a universal large language model;
Searching a subject data knowledge base based on the basic document to obtain search content, wherein the subject data knowledge base is a subject knowledge base corresponding to the basic document, and is constructed by utilizing a subject data deconstructing scheme based on subject data, and the search content comprises a plurality of teaching contents;
Based on the search content, performing color rendering by using an expression alignment model to obtain color rendering content, wherein the expression alignment model is a large language model applicable to the subject, and the color rendering content comprises a plurality of color rendering teaching contents;
screening high-quality teaching contents from the color-rendering contents based on preset scoring standards;
and generating a teaching design based on the high-quality teaching content.
2. The large language model based teaching design generation method according to claim 1, further comprising, after the screening of the good-quality teaching contents from the color contents based on preset scoring criteria:
and if the score of the high-quality teaching content is lower than a preset threshold value, returning to the step of generating the basic document by using the preliminary generation model.
3. The large language model based tutorial design generation method of claim 1, wherein generating a tutorial design based on the premium tutorial content comprises:
integrating the high-quality teaching contents to obtain integrated teaching contents;
And based on the integrated teaching content, performing color rendering by using the expression alignment model to obtain the teaching design.
4. The large language model based teaching design generation method according to claim 1, wherein the expression alignment model is trained based on the steps of:
after data enhancement is carried out on sample data, a training data set is constructed based on the sample data and the data after data enhancement;
and training the pre-training model by using the training data set based on a preset template, and then performing fine adjustment on the model obtained by training by using a preset fine adjustment method to obtain the expression alignment model.
5. The large language model based teaching design generating method according to any one of claims 1-4, wherein when the discipline is a Chinese language, the discipline data includes teaching designs, text and operation for each text, the teaching designs including learning condition analysis, teaching objectives, teaching processes and operation labels.
6. The large language model based teaching design generation method according to claim 5, wherein the discipline repository is constructed based on the steps of:
Deconstructing teaching designs, text and operation of the lessons by using a discipline data deconstructing scheme to obtain deconstructing results;
And constructing a discipline data knowledge base based on the deconstructed result.
7. A large language model based teaching design generation device, comprising:
A document generation module for: generating a basic document by using a preliminary generation model, wherein the preliminary generation model is a universal large language model;
A content retrieval module for: searching a subject data knowledge base based on the basic document to obtain search content, wherein the subject data knowledge base is a subject knowledge base corresponding to the basic document, and is constructed by utilizing a subject data deconstructing scheme based on subject data, and the search content comprises a plurality of teaching contents;
The content color rendering module is used for: based on the search content, performing color rendering by using an expression alignment model to obtain color rendering content, wherein the expression alignment model is a large language model applicable to the subject, and the color rendering content comprises a plurality of color rendering teaching contents;
A content screening module for: screening high-quality teaching contents from the color-rendering contents based on preset scoring standards;
the teaching plan generating module is used for: and generating a teaching design based on the high-quality teaching content.
8. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the large language model based tutorial design generation method of any one of claims 1 to 6 when the computer program is executed.
9. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the steps of the large language model based teaching design generation method of any of claims 1 to 6.
10. A computer program product comprising a computer program which, when executed by a processor, implements the steps of the large language model based teaching design generation method of any of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410399307.7A CN117994101B (en) | 2024-04-03 | 2024-04-03 | Teaching design generation method and device based on large language model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410399307.7A CN117994101B (en) | 2024-04-03 | 2024-04-03 | Teaching design generation method and device based on large language model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117994101A true CN117994101A (en) | 2024-05-07 |
| CN117994101B CN117994101B (en) | 2024-07-12 |
Family
ID=90895818
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410399307.7A Active CN117994101B (en) | 2024-04-03 | 2024-04-03 | Teaching design generation method and device based on large language model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117994101B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110443734A (en) * | 2019-08-13 | 2019-11-12 | 李龙 | A kind of automated teaching design operation system |
| CN115422347A (en) * | 2022-07-25 | 2022-12-02 | 海南科技职业大学 | Knowledge graph-based Chinese course teaching plan generation method |
| KR20230125893A (en) * | 2022-02-22 | 2023-08-29 | ㈜아이윈컴퍼니 | Method of providing educational content that can increase learning efficiency. |
| CN116993544A (en) * | 2023-04-28 | 2023-11-03 | 新大陆数字技术股份有限公司 | LLM-based auxiliary teaching method |
| CN117056538A (en) * | 2023-08-14 | 2023-11-14 | 科大讯飞股份有限公司 | Teaching data generation method, device, equipment and storage medium |
| CN117076693A (en) * | 2023-07-11 | 2023-11-17 | 华中师范大学 | Method for constructing digital human teacher multi-mode large language model pre-training discipline corpus |
| CN117112760A (en) * | 2023-08-26 | 2023-11-24 | 四川欣龙信创科技有限公司 | Intelligent education big model based on knowledge base |
| CN117391902A (en) * | 2023-12-13 | 2024-01-12 | 北京师范大学珠海校区 | Evaluation method and device for Chinese core literacy education based on large language model |
| CN117435689A (en) * | 2023-10-23 | 2024-01-23 | 科大讯飞股份有限公司 | Teaching plan generation method, teaching plan generation device, integrated machine, electronic equipment and storage medium |
| CN117764035A (en) * | 2023-12-18 | 2024-03-26 | 科大讯飞股份有限公司 | Text color rendering and model training method and all-in-one machine |
-
2024
- 2024-04-03 CN CN202410399307.7A patent/CN117994101B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110443734A (en) * | 2019-08-13 | 2019-11-12 | 李龙 | A kind of automated teaching design operation system |
| KR20230125893A (en) * | 2022-02-22 | 2023-08-29 | ㈜아이윈컴퍼니 | Method of providing educational content that can increase learning efficiency. |
| CN115422347A (en) * | 2022-07-25 | 2022-12-02 | 海南科技职业大学 | Knowledge graph-based Chinese course teaching plan generation method |
| CN116993544A (en) * | 2023-04-28 | 2023-11-03 | 新大陆数字技术股份有限公司 | LLM-based auxiliary teaching method |
| CN117076693A (en) * | 2023-07-11 | 2023-11-17 | 华中师范大学 | Method for constructing digital human teacher multi-mode large language model pre-training discipline corpus |
| CN117056538A (en) * | 2023-08-14 | 2023-11-14 | 科大讯飞股份有限公司 | Teaching data generation method, device, equipment and storage medium |
| CN117112760A (en) * | 2023-08-26 | 2023-11-24 | 四川欣龙信创科技有限公司 | Intelligent education big model based on knowledge base |
| CN117435689A (en) * | 2023-10-23 | 2024-01-23 | 科大讯飞股份有限公司 | Teaching plan generation method, teaching plan generation device, integrated machine, electronic equipment and storage medium |
| CN117391902A (en) * | 2023-12-13 | 2024-01-12 | 北京师范大学珠海校区 | Evaluation method and device for Chinese core literacy education based on large language model |
| CN117764035A (en) * | 2023-12-18 | 2024-03-26 | 科大讯飞股份有限公司 | Text color rendering and model training method and all-in-one machine |
Non-Patent Citations (1)
| Title |
|---|
| 魏顺平;路秋丽;何克抗;李宇翔;: "教学设计自动化语义模型及其实现方法", 开放教育研究, no. 06, 5 December 2009 (2009-12-05), pages 52 - 60 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117994101B (en) | 2024-07-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Chang et al. | A corpus-based approach to online materials development for writing research articles | |
| Madiba | Promoting concept literacy through multilingual glossaries: A translanguaging approach | |
| Thomas et al. | A diachronic analysis of explicit definitions and implicit conceptualizations of language learning strategies | |
| Erwen et al. | Application of computer-aided translation technology in translation teaching | |
| Malherbe et al. | Bridge the terminology gap between recruiters and candidates: A multilingual skills base built from social media and linked data | |
| Hadifar et al. | EduQG: A multi-format multiple-choice dataset for the educational domain | |
| CN116821377A (en) | Primary school Chinese automatic evaluation system based on knowledge graph and large model | |
| CN116756340A (en) | Test question automatic generation method, system and equipment | |
| CN117391902B (en) | Evaluation method and device for Chinese core literacy education based on large language model | |
| Cai | RETRACTED: Practice of Hybrid Teaching Mode of English Writing Based on Artificial Intelligence | |
| Slater et al. | Using correlational topic modeling for automated topic identification in intelligent tutoring systems | |
| CN117994101B (en) | Teaching design generation method and device based on large language model | |
| Chen | Strategies for improving the effectiveness of English translation teaching in higher vocational colleges based on data mining | |
| Vințe et al. | Sustainable development in education–automating curriculum assessment | |
| Wang | A Study of Applying Automated Assessment in Teaching College English Writing Based on Juku Correction Network. | |
| Hwang et al. | The Implementation of ChatGPT-assisted Writing Instruction in ESL/EFL Classrooms | |
| Tolmachev et al. | Automatic extraction of high-quality example sentences for word learning using a determinantal point process | |
| Jing | A teaching system of English online course based on artificial intelligence | |
| Azouaou et al. | Semantic annotation for the teacher: models for a computerized memory tool | |
| Wu et al. | Quality evaluation system of online courses based on review data | |
| Olsen et al. | Integrating data cleansing with popular culture: A novel SQL character data tutorial | |
| Niu | New Applications of Online Education Content Analysis Based on Natural Language Processing Technology | |
| Lan et al. | Research on the construction of a teaching skills graph of information technology normal students | |
| León | Guidelines for the design and generation of corpus-based terminology-oriented instructional material | |
| Sarmiento et al. | Representation of an electrical engineering curriculum using an ontology of controlled quality |
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 | ||
| GR01 | Patent grant |