Calligraphy teaching auxiliary system and method based on language chain and dialogue generation model

Abstract

The present invention discloses a calligraphy teaching auxiliary system and method based on a language chain and a dialogue generation model, which relates to the field of artificial intelligence education technology, and includes: a database module, a knowledge matching engine and an interaction module; wherein the database module is used to store a calligraphy knowledge base; the interaction module is used to obtain user interaction information; the knowledge matching engine includes a language chain model and a dialogue generation model; the language chain model is used to analyze interaction information and provide context-related information of interaction information; the dialogue generation model is used to generate corresponding response information based on the calligraphy knowledge base, interaction information and context-related information of interaction information; the interaction module is also used to display the response information to the user. The present invention alleviates the problem of insufficient combination of theory and practice that learners often face in the calligraphy learning process in the prior art, as well as the technical problem of lack of real-time feedback and guidance in the learning process.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence education technology, and in particular to a calligraphy teaching auxiliary system and method based on a language chain and dialogue generation model. Background Art

[0002] In the process of calligraphy art education and learning, traditional teaching methods usually rely on direct teaching and individual guidance from teachers, which is limited in resources and time. In order to improve learning efficiency and quality, modern calligraphy teaching attempts to introduce digital tools, but these tools often lack the ability of personalized teaching and deep interaction. With the development of computer science and information technology, especially the breakthrough of natural language processing (NLP) technology, new opportunities have emerged to use artificial intelligence to assist calligraphy teaching.

[0003] Existing calligraphy teaching applications or systems, although they can provide a certain degree of interactivity, have difficulty understanding complex calligraphy terms, providing accurate calligraphy knowledge answers, or adapting teaching content to learners' personalized needs. In addition, learners often face the problem of insufficient integration of theory and practice in the calligraphy learning process, as well as a lack of real-time feedback and guidance during the learning process. Summary of the invention

[0004] The purpose of the present invention is to provide a calligraphy teaching auxiliary system and method based on a language chain and a dialogue generation model in order to solve at least one of the above technical problems.

[0005] In the first aspect, an embodiment of the present invention provides a calligraphy teaching auxiliary system based on a language chain and a dialogue generation model, comprising: a database module, a knowledge matching engine and an interaction module; wherein the database module is used to store a calligraphy knowledge base; the interaction module is used to obtain user interaction information; the knowledge matching engine includes a language chain model and a dialogue generation model; the language chain model is used to analyze the interaction information and provide context-related information of the interaction information; the dialogue generation model is used to generate corresponding response information based on the calligraphy knowledge base, the interaction information and the context-related information of the interaction information; the interaction module is also used to display the response information to the user.

[0006] Furthermore, the calligraphy knowledge base includes: basic calligraphy knowledge, calligraphy work information and calligraphy teaching information.

[0007] Furthermore, the language chain model is also used to clean and standardize the text data in the calligraphy knowledge base and convert it into vector form.

[0008] Furthermore, the dialogue generation model includes a ChatGLM model.

[0009] Furthermore, it also includes an information collection module and an update module; wherein the information collection module is used to collect the calligraphy knowledge information uploaded by the user; and the update module is used to update the calligraphy knowledge base and the knowledge matching engine based on the calligraphy knowledge information.

[0010] In the second aspect, an embodiment of the present invention also provides a calligraphy teaching auxiliary method based on a language chain and a dialogue generation model, the method comprising: constructing a calligraphy knowledge base based on preset database resources; obtaining user interaction information; analyzing the interaction information based on a preset language chain model to obtain context-related information of the interaction information; using the calligraphy knowledge base, the interaction information and the context-related information of the interaction information as inputs of a preset dialogue generation model to generate corresponding response information; and displaying the response information to the user.

[0011] Furthermore, the context encoder of the preset language chain model includes Transformer-XL; the input processing mechanism of the preset language chain model includes a multimodal input processing mechanism, and the multimodal input processing mechanism includes: M=F multi (T, I p , A); where M represents the integrated multimodal information, F multi is the fusion function of multimodal information, T represents text information, I p represents image information, and A represents audio information.

[0012] Furthermore, the expression of the preset dialogue generation model includes: R=M ChatGLM (K calligraphy , I, C, E, P); where R represents the response information, E represents the emotional state, P represents the user portrait and interaction data, and K calligraphy represents the calligraphy knowledge base, I represents the interactive information, and C represents the context-related information.

[0013] In a third aspect, an embodiment of the present invention further provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in the second aspect when executing the computer program.

[0014] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the method described in the second aspect is implemented.

[0015] The present invention provides a calligraphy teaching auxiliary system and method based on a language chain and dialogue generation model. The system and method combine the language chain model and the dialogue generation model to respond to learners' inquiries in real time and provide personalized guidance and feedback to improve the efficiency of calligraphy teaching and the learning accuracy of learners, thereby alleviating the problems existing in the prior art of insufficient integration of theory and practice faced by learners in the calligraphy learning process, as well as the technical problems of lack of real-time feedback and guidance in the learning process. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the implementation methods of the present application or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1 A schematic diagram of a calligraphy teaching auxiliary system based on a language chain and a dialogue generation model provided by an embodiment of the present invention;

[0018] Figure 2 A schematic diagram of a knowledge matching engine provided by an embodiment of the present invention;

[0019] Figure 3 A schematic diagram of another calligraphy teaching auxiliary system based on a language chain and a dialogue generation model provided by an embodiment of the present invention;

[0020] Figure 4 A flowchart of a calligraphy teaching auxiliary method based on a language chain and a dialogue generation model provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] Embodiment 1: Figure 1 Schematic diagram of a calligraphy teaching auxiliary system based on a language chain and a dialogue generation model provided according to an embodiment of the present invention. Figure 1 As shown, the system includes: a database module 10, a knowledge matching engine 20 and an interaction module 30.

[0023] Specifically, the database module 10 is used to store a calligraphy knowledge base;

[0024] An interaction module 30, used to obtain user interaction information;

[0025] Figure 2 FIG. 1 is a schematic diagram of a knowledge matching engine provided according to an embodiment of the present invention. Figure 2 As shown, the knowledge matching engine 20 includes a language chain model 21 and a dialogue generation model 22 .

[0026] Specifically, the language chain model 21 is used to analyze the interaction information and provide context-related information of the interaction information;

[0027] Preferably, in the embodiment of the present invention, the language chain model 21 includes an improved LangChain model, which adopts a more powerful context encoder Transformer-XL. XL , enhances the sensitivity to long-term conversation history, thereby improving the ability to understand long conversations. The context-awareness of the improved LangChain model can be expressed as follows:

[0028] C=T XL (H)

[0029] Among them, H represents the historical dialogue sequence. In addition, the improved LangChain model realizes the understanding of multiple data formats such as text T, image I and audio A through the multimodal input processing mechanism, which is expressed as:

[0030] M=F multi (T, I, A)

[0031] Among them, M represents the integrated multimodal information, and F multi It is the fusion function of multimodal information.

[0032] This system, which integrates the improved LangChain model and the ChatGLM model, not only provides structured logical reasoning, but also optimizes the naturalness of conversations and emotional matching, ensuring that the generated conversations are more natural and more in line with user expectations, while also having a high degree of adaptability and customized service capabilities.

[0033] The dialogue generation model 22 is used to generate corresponding response information based on the calligraphy knowledge base, the interaction information and the context-related information of the interaction information.

[0034] Specifically, the interaction module 30 is also used to display response information to the user.

[0035] Preferably, the language chain model 21 is also used to clean and standardize the text data in the calligraphy knowledge base and convert it into a vector form.

[0036] Preferably, the dialogue generation model includes a ChatGLM model.

[0037] Among them, ChatGLM is a large bilingual model in Chinese and English, based on the GLM-130B, a 100 billion base model. The following are some key points of the ChatGLM model:

[0038] (1) Technical details: ChatGLM is developed based on the Transformer structure. Unlike models such as GPT and BERT, ChatGLM learns a large amount of dialogue corpus during training, enabling it to perform well in dialogue generation and understanding tasks.

[0039] (2) Application areas: ChatGLM is widely used in natural language processing, dialogue systems, language learning, and knowledge graph construction.

[0040] (3) Performance optimization: In terms of server deployment, ChatGLM supports fine-tuning techniques, including using GPU for computing and multi-threaded concurrent processing of requests, to optimize model performance.

[0041] (4) Features: ChatGLM inherits the advantages of the GLM series and optimizes the model architecture, lowering the threshold for deployment and application, so that large models can also be used for reasoning applications on consumer-grade graphics cards. It uses ChatGPT's reinforcement learning human alignment strategy to make the generation effect closer to human value.

[0042] Specifically, calligraphy knowledge base K calligraphy , the interaction information I and its context-related information C are provided as input to the preset dialogue generation model M ChatGLM , to generate the response information R. The preset dialogue generation model adjusts the style and content of the reply by integrating emotional sensitivity E and user personalized features P:

[0043] R=M ChatGLM (K calligraphy , I, C, E, P)

[0044] Among them, E represents emotional state, and P represents user portrait and interaction data. At the same time, the model introduces an enhanced error handling mechanism to handle D error , optimize response generation to ensure reasonable responses are provided even when receiving partially ambiguous or incorrect input:

[0045] ;

[0046] Even when the model receives partially ambiguous or erroneous input, This ensures that the generated responses remain reasonable.

[0047] Preferably, the calligraphy knowledge base includes: basic calligraphy knowledge, calligraphy works information and calligraphy teaching information. For example, the basic calligraphy knowledge includes the introduction of various calligraphy styles, the introduction of various calligraphy techniques, the characteristics and evolution history of different calligraphy styles, etc.; the calligraphy works information includes the works of famous calligraphers of various periods and their artistic evaluations, etc.; the calligraphy teaching information includes various brushstroke explanations, related calligraphy teaching materials, video tutorials and learning guides, etc. Each entry in the calligraphy knowledge base is equipped with a detailed description, classification label and relevance score so that the database module 10 can accurately retrieve and recommend content.

[0048] Specifically, let D = {d1, d2, …, d n} as the document collection of the knowledge base, each document d i Contains calligraphy-related knowledge points, and the representation of knowledge points K i Can be converted into a vector K through word embedding i =WordEmbedding(d i ).

[0049] In the embodiment of the present invention, the knowledge matching engine 20 can analyze the information provided by the user, and intelligently match the corresponding calligraphy teaching materials and knowledge points from the database according to the user's learning history and preferences, generate personalized learning paths and course plans, and provide a one-to-one teaching experience. The core is to capture the long-distance dependencies between input sequences through the self-attention mechanism. In the calligraphy teaching auxiliary system, the language chain LangChain can be used to understand and generate calligraphy knowledge related to the learner's query.

[0050] In the embodiment of the present invention, it is assumed that a given input sequence X=(x1, x2, …, x n ), where this sequence is the sample data after text segmentation, the self-attention mechanism first converts the text input sequence into three vector sets: queries (Queries, Q), keys (Keys, K) and values ​​(Values, V), then calculates the similarity between each query and all keys, converts it into weight parameters, and finally generates weighted values, outputting the sequence Y=(y1,y2,…,y n ), where the Y sequence corresponds to the previously inputted calligraphy data related slice text, which is the knowledge point, question, answer or explanation of a specific topic related to the user query. In the transformer model, both the X sequence and the Y sequence are embedded in a high-dimensional space to obtain rich semantic information.

[0051] Specifically, the mathematical modeling of the ChatGLM model is as follows: Assume that Q={q1,q2,…,q n} is a series of questions asked by the user, where each q is a natural language string. The goal of the ChatGLM model is to generate an answer a for each question q. Then the question q can be mapped to a high-dimensional vector space, and then the sequence-to-sequence framework is used to process the embedded vector, and finally a vector sequence is generated. The vector sequence can be converted into natural language to answer the user's question.

[0052] In the embodiment of the present invention, the interaction module 30 enables the user to ask specific calligraphy questions in a natural language.

[0053] Figure 3 FIG. 1 is a schematic diagram of another calligraphy teaching auxiliary system based on a language chain and a dialogue generation model according to an embodiment of the present invention. Figure 3 As shown, the system provided by the embodiment of the present invention further includes an information collection module 40 and an update module 50.

[0054] Specifically, the information collection module 40 is used to collect calligraphy knowledge information uploaded by users;

[0055] The specific information collection module 40 allows users to upload calligraphy works and query calligraphy-related questions. This module uses image recognition and natural language processing technology to extract key information of the uploaded content, including but not limited to brushstroke style, structural layout, ink application skills, etc., and can also extract key terms and concepts from text descriptions.

[0056] The updating module 50 is used to update the calligraphy knowledge base and the knowledge matching engine based on the calligraphy knowledge information.

[0057] Specifically, the update module 50 allows calligraphy educators and experts to upload teaching experience, historians to upload the latest research results, artists to share their creative experience and other resources to the designated shared folder. These resources will be automatically identified and classified, and imported into the corresponding classification knowledge base in the database according to the naming rule of "calligraphy field + knowledge point type + serial number" to ensure the timeliness and diversity of teaching content.

[0058] From the above description, it can be seen that the embodiment of the present invention provides a calligraphy teaching auxiliary system based on language chain and dialogue generation model, which can provide learners with an interactive and personalized learning environment, meet their learning needs at any time, and improve learning efficiency and interest. At the same time, teachers and experts can better understand and adapt to learners' needs through this system, and realize real-time updating and optimization of teaching content and methods.

[0059] Embodiment 2: Figure 4 1 is a flowchart of a calligraphy teaching auxiliary method based on a language chain and a dialogue generation model according to an embodiment of the present invention. Figure 4 As shown, the method specifically comprises the following steps:

[0060] Step S402: construct a calligraphy knowledge base based on preset database resources.

[0061] Specifically, a comprehensive calligraphy knowledge base is established using the knowledge of experts and public database resources. The calligraphy knowledge base includes: basic calligraphy knowledge, calligraphy works information, and calligraphy teaching information. For example, basic calligraphy knowledge includes the introduction of various calligraphy styles, the introduction of various calligraphy techniques, the characteristics and evolution history of different calligraphy styles, etc.; calligraphy works information includes the works of famous calligraphers of various periods and their artistic evaluations, etc.; calligraphy teaching information includes various brushwork explanations, related calligraphy teaching materials, video tutorials, and learning guides. Each entry in the calligraphy knowledge base is equipped with a detailed description, classification label, and relevance score so that the database module 10 can accurately retrieve and recommend content.

[0062] Specifically, let D = {d1, d2, …, d n} as the document collection of the knowledge base, each document d i Contains calligraphy-related knowledge points, and the representation of knowledge points K i Can be converted into a vector K through word embedding i =WordEmbedding(d i ).

[0063] Step S404: Acquire user interaction information.

[0064] Step S406: Analyze the interaction information based on a preset language chain model to obtain context-related information of the interaction information.

[0065] Specifically, the preset language chain model (i.e., the improved LangChain model) adopts a more powerful context encoder Transformer-XL, T XL , enhances the sensitivity to long-term conversation history, thereby improving the ability to understand long conversations. The context-awareness of the improved LangChain model can be expressed as follows:

[0066] C=T XL (H)

[0067] Among them, H represents the historical dialogue sequence. In addition, the improved LangChain model realizes the processing of text T, image I through multimodal input processing mechanism. p And the understanding of various data formats such as audio A, expressed as:

[0068] M=F multi (T, I p , A)

[0069] Among them, M represents the integrated multimodal information, and F multi It is the fusion function of multimodal information.

[0070] This system, which integrates the improved LangChain model and the ChatGLM model, not only provides structured logical reasoning, but also optimizes the naturalness of conversations and emotional matching, ensuring that the generated conversations are more natural and more in line with user expectations, while also having a high degree of adaptability and customized service capabilities.

[0071] Step S408, using the calligraphy knowledge base, the interaction information and the context-related information of the interaction information as inputs of a preset dialogue generation model to generate corresponding response information.

[0072] Among them, calligraphy knowledge base K calligraphy , the interaction information I and its context-related information C are provided as input to the preset dialogue generation model M ChatGLM , to generate the response information R. The preset dialogue generation model adjusts the style and content of the reply by integrating emotional sensitivity E and user personalized features P:

[0073] R=M ChatGLM (K calligraphy , I, C, E, P)

[0074] Among them, E represents emotional state, and P represents user portrait and interaction data. At the same time, the model introduces an enhanced error handling mechanism to handle D error , optimize response generation to ensure reasonable responses are provided even when receiving partially ambiguous or incorrect input:

[0075] ;

[0076] Even when the model receives partially ambiguous or erroneous input, This ensures that the generated responses remain reasonable.

[0077] Step S410: display response information to the user.

[0078] An embodiment of the present invention further provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method provided in the embodiment of the present invention when executing the computer program.

[0079] The embodiment of the present invention further provides a computer-readable storage medium, which stores computer instructions. When the computer instructions are executed by a processor, the method provided by the embodiment of the present invention is implemented.

[0080] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and range of equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0081] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A calligraphy teaching auxiliary system based on language chain and dialogue generation model, characterized in that: include: Database module, knowledge matching engine and interaction module; among them, The database module is used to store a calligraphy knowledge base; The interaction module is used to obtain the user's interaction information; The knowledge matching engine includes a language chain model and a dialogue generation model; The language chain model is used to analyze the interaction information and provide context-related information for the interaction information; The dialogue generation model is used to generate corresponding response information based on the calligraphy knowledge base, the interaction information and the context-related information of the interaction information; let D = {d1, d2, ..., d n } is the document collection of the calligraphy knowledge base, each document d i Contains calligraphy-related knowledge points, and the representation of knowledge points K i Can be converted into a vector K through word embedding i =WordEmbedding(d i ); Among them, calligraphy knowledge base K calligraphy , the interaction information I and its context-related information C are provided as input to the preset dialogue generation model M ChatGLM , to generate the response information R, the preset dialogue generation model adjusts the style and content of the reply by integrating the emotional sensitivity E and the user's personalized features P: R=M ChatGLM (K calligraphy ,I,C,E,P); At the same time, the model introduces an enhanced error handling mechanism to handle D error , optimize response generation to ensure reasonable responses are provided even when receiving partially ambiguous or incorrect input: ； Even when the model receives partially ambiguous or erroneous input, Ensure that the generated responses remain reasonable; The interaction module is further used to display the response information to the user; The calligraphy knowledge base includes: basic calligraphy knowledge, calligraphy works information and calligraphy teaching information; The language chain model includes an improved LangChain model; the context awareness capability of the improved LangChain model is expressed by the following formula: C=T XL (H); Among them, H represents the historical dialogue sequence, T XL is the context encoder; The improved LangChain model realizes the processing of text T, image I through a multimodal input processing mechanism. p And the understanding of multiple data formats of audio A is expressed as: M=F multi (T,I p ,A); Among them, M represents the integrated multimodal information, and F multi is the fusion function of multimodal information; It also includes an information collection module and an update module; wherein, The information collection module is used to collect calligraphy knowledge information uploaded by the user; The updating module is used to update the calligraphy knowledge base and the knowledge matching engine based on the calligraphy knowledge information; Using the given text input sequence X=(x1,x2,…,x n ), where the text input sequence X is the sample data after text segmentation. The self-attention mechanism first converts the text input sequence X into three vector sets: queries (Queries, Q), keys (Keys, K) and values ​​(Values, V). Then, the similarity between each query and all keys is calculated and converted into weight parameters. Finally, the weighted values ​​are generated and the output sequence Y=(y1,y2,…,y n ), where sequence Y corresponds to the previously inputted slice text of calligraphy materials, which is the knowledge point, question, answer or explanation of a specific topic related to the user query. In the transformer model, both sequence X and sequence Y are embedded in the high-dimensional space.

2. The system according to claim 1, characterized in that: The language chain model is also used to clean and standardize the text data in the calligraphy knowledge base and convert it into vector form.

3. A calligraphy teaching auxiliary method based on language chain and dialogue generation model, characterized in that: The method comprises: Build a calligraphy knowledge base based on preset database resources; Obtain user interaction information; Analyze the interaction information based on a preset language chain model to obtain context-related information of the interaction information; Using the calligraphy knowledge base, the interaction information and context-related information of the interaction information as inputs of a preset dialogue generation model, generating corresponding response information; Let D = {d1, d2, …, d n } is the document collection of the calligraphy knowledge base, each document d i Contains calligraphy-related knowledge points, and the representation of knowledge points K i Can be converted into a vector K through word embedding i =WordEmbedding(d i ); Among them, calligraphy knowledge base K calligraphy , the interaction information I and its context-related information C are provided as input to the preset dialogue generation model M ChatGLM , to generate the response information R, the preset dialogue generation model adjusts the style and content of the reply by integrating the emotional sensitivity E and the user's personalized features P: R=M ChatGLM (K calligraphy ,I,C,E,P); At the same time, the model introduces an enhanced error handling mechanism to handle D error , optimize response generation to ensure reasonable responses are provided even when receiving partially ambiguous or incorrect input: ； Even when the model receives partially ambiguous or erroneous input, It is ensured that the generated reply still maintains rationality; the response information is displayed to the user; The calligraphy knowledge base includes: basic calligraphy knowledge, calligraphy works information and calligraphy teaching information; The language chain model includes an improved LangChain model; the context awareness capability of the improved LangChain model is expressed by the following formula: C=T XL (H); Among them, H represents the historical dialogue sequence, T XL is the context encoder; The improved LangChain model realizes the processing of text T, image I through a multimodal input processing mechanism. p And the understanding of multiple data formats of audio A is expressed as: M=F multi (T,I p ,A); Among them, M represents the integrated multimodal information, and F multi It is the fusion function of multimodal information; Using the given text input sequence X=(x1,x2,…,x n ), where the text input sequence X is the sample data after text segmentation. The self-attention mechanism first converts the text input sequence X into three vector sets: queries (Queries, Q), keys (Keys, K) and values ​​(Values, V). Then, the similarity between each query and all keys is calculated and converted into weight parameters. Finally, the weighted values ​​are generated and the output sequence Y=(y1,y2,…,y n ), where sequence Y corresponds to the previously inputted slice text of calligraphy materials, which is the knowledge point, question, answer or explanation of a specific topic related to the user query. In the transformer model, both sequence X and sequence Y are embedded in the high-dimensional space.

4. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to claim 3 when executing the computer program.

5. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the method according to claim 3 is implemented.

Images