Intelligent question answering method and system for low-altitude economic education for interactive learning platform

Abstract

The invention provides a low-altitude economic education intelligent question answering method and system for an interactive learning platform, and relates to the technical field of natural language process.The low-altitude economic education intelligent question answering method comprises the steps that firstly, a low-altitude economic education question data set containing multiple pieces of user interaction data is obtained, and a question feature set of the user interaction data is extracted; comprising semantic vector features, domain association features and context association features, retrieving a domain knowledge feature set in a low-altitude economic education knowledge base based on a question feature set, performing cross-dimensional feature fusion processing on the question feature set and the domain knowledge feature set, generating an optimized response statement of the user question statement, and sending the optimized response statement to the low-altitude economic education knowledge base. Finally, according to the difference characteristics of the optimized response statement and the historical response statement, the response service strategy of the interactive learning platform is dynamically adjusted, and the question and answer accuracy and the platform adaptability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of natural language processing, and in particular, to an intelligent question-answering method and system for low-altitude economy education for an interactive learning platform. Background Art

[0002] In existing interactive learning platforms, there are many deficiencies in the intelligent question-answering function for low-altitude economy education. On the one hand, most traditional question-answering systems simply match based on the surface text of user questions, lacking a deep understanding of question semantics, domain relevance, and context, resulting in poor accuracy and pertinence of answers. For example, for some questions with similar expressions but different meanings, the question-answering system is difficult to accurately distinguish and give precise answers. On the other hand, these question-answering systems often search for answers in the knowledge base in isolation, without effectively integrating question features with domain knowledge, making the answer content relatively single and one-sided. Moreover, existing question-answering systems lack a mechanism to dynamically adjust the response service strategy, cannot be optimized according to the user's historical response situation, and cannot adapt to the diverse needs of different users, reducing the learning experience and effect of users on the interactive learning platform. Summary of the Invention

[0003] In view of the problems mentioned above, in combination with the first aspect of the present invention, embodiments of the present invention provide an intelligent question-answering method for low-altitude economy education for an interactive learning platform, and the method includes: Obtain a set of low-altitude economy education question data submitted by a user on an interactive learning platform, where the set of low-altitude economy education question data includes a plurality of user interaction data, and each user interaction data is composed of at least one user question statement and a corresponding historical response statement; Extract a set of question features of the user interaction data, where the set of question features includes semantic vector features of the user question statement, domain association features between the user question statement and the low-altitude economy education field, and context association features of the user question statement; Retrieve a set of domain knowledge features in the low-altitude economy education knowledge base based on the set of question features; Perform cross-dimensional feature fusion processing on the set of question features and the set of domain knowledge features to generate an optimized response statement for the user question statement; Dynamically adjust the response service strategy of the interactive learning platform based on the difference features between the optimized response statement and the historical response statements in the user interaction data.

[0004] On the other hand, an embodiment of the present invention further provides an intelligent question-answering system for low-altitude economy education for an interactive learning platform, including a processor and a machine-readable storage medium. The machine-readable storage medium is connected to the processor. The machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0005] Based on the above aspects, the embodiment of the present invention can fully combine the question features and domain knowledge to improve the accuracy and comprehensiveness of the answer by obtaining a set of low-altitude economy education question data containing multiple user interaction data, extracting question feature sets such as semantic vector features, domain association features, and context association features of the user interaction data therein, and then retrieving the domain knowledge feature set in the low-altitude economy education knowledge base, and generating an optimized response statement through cross-dimensional feature fusion. At the same time, by dynamically adjusting the response service strategy based on the difference features between the optimized response statement and the historical response statement, the interactive learning platform can continuously optimize the response service according to the actual situation of users, better meet the diverse needs of different users in low-altitude economy education learning, improve the user learning experience and learning effect, and enhance the intelligence and adaptability of the platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] Figure 1 is a schematic flowchart of the execution process of the intelligent question-answering method for low-altitude economy education for an interactive learning platform provided by an embodiment of the present invention.

[0007] Figure 2 is a schematic diagram of exemplary hardware and software components of the intelligent question-answering system for low-altitude economy education for an interactive learning platform provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0008] The present invention will be specifically described below with reference to the accompanying drawings of the specification. Figure 1 is a schematic flowchart of the intelligent question-answering method for low-altitude economy education for an interactive learning platform provided by an embodiment of the present invention. The intelligent question-answering method for low-altitude economy education for an interactive learning platform will be introduced in detail below.

[0009] Step S110: Obtain a set of low-altitude economy education question data submitted by a user on an interactive learning platform. The set of low-altitude economy education question data includes multiple user interaction data, and each user interaction data is composed of at least one user question statement and a corresponding historical response statement.

[0010] In this embodiment, during the daily operation of the interactive learning platform, many users log in and initiate questions related to low-altitude economy education. The interactive learning platform has a perfect data reception and storage mechanism. When a user enters a question in the question interface of the interactive learning platform, the user's question statement can be quickly captured. At the same time, if there is a corresponding response record for this user's question statement in the past, the corresponding historical response statement can be retrieved from the historical data storage area. Each group of user question statements and historical response statements forms a user interaction data. As time goes by and many users participate, a large amount of user interaction data is accumulated, jointly constituting the low-altitude economy education question data set.

[0011] To illustrate more intuitively, the interactive learning platform uses a distributed storage system to manage this data. When a user submits a question about the efficiency evaluation of a new transportation mode in the low-altitude economy, the question statement can be encoded in a set format, and at the same time, it is checked whether there is a response to a similar question in history. If so, the corresponding historical response statement is associated with the current user question statement and stored in a data node of the distributed storage system. As more users submit different types of questions, each data node will be continuously updated and expanded, and finally a complete low-altitude economy education question data set is formed.

[0012] Step S120: Extract the question feature set of the user interaction data, where the question feature set includes the semantic vector feature of the user question statement, the domain association feature between the user question statement and the low-altitude economy education field, and the context association feature of the user question statement.

[0013] Step S121: Perform semantic encoding processing on the user question statement to generate the semantic vector feature of the user question statement, where the semantic vector feature includes the keyword distribution weight of the core concepts in the user question statement.

[0014] Step S1211: Invoke the pre-trained semantic encoder in the low-altitude economy field to perform word segmentation processing on the user question statement to obtain multiple semantic word units.

[0015] The pre-trained semantic encoder for the low-altitude economy field is trained based on a large amount of text data related to the low-altitude economy. This semantic encoder for the low-altitude economy field is designed to accurately identify professional vocabulary, terms, and common language expressions in the low-altitude economy field. When receiving a user question statement, the semantic encoder for the low-altitude economy field will perform a word segmentation operation on it. For example, if the user's question statement is "Under the trend of the development of the low-altitude economy, what are the significant differences between the emerging low-altitude logistics distribution model and the traditional model in terms of cost control and timeliness", the encoder will split this question statement into multiple semantic word units such as "low-altitude economy", "development trend", "emerging", "low-altitude logistics distribution model", "traditional model", "cost control", "timeliness", and "significant differences". This process is achieved based on the lexical analysis and language model inside the encoder, which can accurately divide each semantic word unit according to the semantic and grammatical rules of the vocabulary.

[0016] Step S1212: Analyze the context dependence relationship of the semantic word units to generate the local context weight of each semantic word unit, and the local context weight is used to characterize the semantic contribution degree of the semantic word unit in the user question statement.

[0017] After obtaining the semantic word units, it is necessary to analyze their dependence relationship in the context of the user question statement. This is because in a statement, each semantic word unit is interrelated, and they have different contributions to expressing the semantics of the entire statement. For example, in the above question statement, "low-altitude logistics distribution model" and "traditional model" are the core comparison objects and play a key role in semantic expression, so their local context weights are relatively high. Expressions like "under..." mainly provide background information, and their local context weights are relatively low.

[0018] To determine the local context weight, an attention mechanism based on deep learning can be adopted. This mechanism will calculate the degree of association between each semantic word unit and other semantic word units, and assign a suitable weight value to each semantic word unit by learning the semantic structure and information flow in the statement. For example, for the semantic word unit "low-altitude logistics distribution model", the attention mechanism will consider its association with other semantic word units such as "cost control" and "timeliness", and determine its local context weight according to the tightness of the association.

[0019] Step S1213: Perform weighted aggregation processing on the semantic word units according to the local context weight to generate the initial semantic vector of the user question statement.

[0020] After obtaining the local context weights of each semantic word unit, weighted aggregation processing can be performed. Suppose the set of semantic word units is {u1, u2, …, un}, and the corresponding set of local context weights is {w1, w2, …, wn}. The process of weighted aggregation is to multiply each semantic word unit by its corresponding local context weight, and then combine these products. Specifically, for each semantic word unit ui, multiply it by the corresponding weight wi to get wi * ui, and then combine all wi * ui to form a vector representation, which is the initial semantic vector of the user's question sentence. This initial semantic vector comprehensively considers the semantic contribution degree of each semantic word unit and can more accurately express the semantic information of the user's question sentence.

[0021] Step S1214: Input the initial semantic vector into the semantic optimization model in the low-altitude economy field, and through knowledge graph embedding and term context association analysis, perform domain knowledge enhancement processing on the initial semantic vector to generate a semantic vector feature containing the distribution characteristics of low-altitude economy professional terms.

[0022] The semantic optimization model in the low-altitude economy field is a model designed for the low-altitude economy field, which combines knowledge graph and term context association analysis technologies. The knowledge graph is a knowledge base containing various concepts, entities, and their relationships in the low-altitude economy field. When the initial semantic vector is input into this semantic optimization model in the low-altitude economy field, the semantic optimization model in the low-altitude economy field will embed the relevant knowledge in the knowledge graph into the initial semantic vector. For example, if the concept of "low-altitude logistics distribution mode" is included in the initial semantic vector, the semantic optimization model in the low-altitude economy field will search for various attributes and relationships related to it in the knowledge graph, such as the operation process of this mode, related technologies, etc., and integrate this knowledge information into the initial semantic vector.

[0023] At the same time, term context association analysis will further analyze the meanings and associations of professional terms in the initial semantic vector in different contexts. In the low-altitude economy field, some professional terms may have multiple meanings, and through term context association analysis, their specific semantics in the current user's question sentence can be accurately determined. After being processed by knowledge graph embedding and term context association analysis, the initial semantic vector is enhanced with domain knowledge, and finally a semantic vector feature containing the distribution characteristics of low-altitude economy professional terms is generated. This semantic vector feature more comprehensively and accurately reflects the semantic information of the user's question sentence and contains the keyword distribution weights of core concepts.

[0024] Step S122: Extract the response semantic features of the historical response sentences in the user interaction data, perform temporal alignment processing on the response semantic features and the semantic vector features, and generate the context association features of the user's question sentence.

[0025] In this embodiment, a semantic extraction module can be used to process historical response statements to extract response semantic features. Based on natural language processing techniques, the semantic extraction module performs operations such as part-of-speech tagging and syntactic analysis on historical response statements, so as to identify key information and semantic structures in the statements. For example, for the historical response statement "The traditional low-altitude logistics distribution model mainly relies on economies of scale in cost control, while the emerging model pays more attention to the efficiency improvement brought by technological innovation", the semantic extraction module will identify key information such as "traditional model", "cost control", "economies of scale", "emerging model", "technological innovation", and "efficiency improvement", and convert this information into a feature vector, which is the response semantic feature.

[0026] After obtaining the response semantic features, it is necessary to perform temporal alignment processing on them with the semantic vector features of the previously generated user question statements. The purpose of temporal alignment is to make the two features match in the time dimension to reflect the context relationship between the user question and the historical response. The dynamic time warping (DTW) algorithm can be used to achieve temporal alignment. This algorithm will find an optimal matching path between the two feature sequences, so that the two sequences are as aligned as possible in time. In this way, the response semantic features and the semantic vector features are fused to generate context-related features of the user question statement. This context-related feature can reflect the information of the user question in the historical context and provide richer context information for subsequent processing.

[0027] Step S123: Determine the domain association feature of the user question statement according to the term coverage of the user question statement in the low-altitude economy education field, where the domain association feature is used to quantify the association strength between the user question statement and the core knowledge system of low-altitude economy education.

[0028] To determine the domain association feature of the user question statement, it is necessary to calculate the term coverage of the user question statement in the low-altitude economy education field. First, a term dictionary for the low-altitude economy education field needs to be established, which contains all the professional terms and common words in this field. Then, the user question statement is compared with this term dictionary. For each word in the user question statement, check whether it appears in the term dictionary. If it appears, then this word is considered a domain term.

[0029] For example, for the user question statement "Under the trend of the development of the low-altitude economy, what are the significant differences between the emerging low-altitude logistics distribution model and the traditional model in terms of cost control and timeliness?", by comparing with the term dictionary, it is found that terms such as "low-altitude economy", "low-altitude logistics distribution model", "cost control", and "timeliness" all belong to domain terms. Calculate the proportion of these domain terms in the user question statement, and this proportion is the term coverage.

[0030] The higher the term coverage, the greater the correlation strength between the user question statement and the core knowledge system of low-altitude economy education. Therefore, the term coverage can be directly used as a domain correlation feature, which can quantify the correlation strength between the user question statement and the core knowledge system of low-altitude economy education. This domain correlation feature plays an important role in the subsequent retrieval and matching processes, and it can help the system more accurately screen out domain knowledge related to the user's question.

[0031] Step S124: Concatenate the semantic vector feature, the context correlation feature, and the domain correlation feature to generate the question feature set.

[0032] After obtaining the semantic vector feature, the context correlation feature, and the domain correlation feature, they need to be concatenated. Feature concatenation is to connect different feature vectors in a set order to form a longer feature vector. For example, assume that the dimension of the semantic vector feature is m, the dimension of the context correlation feature is n, and the dimension of the domain correlation feature is p. Then the dimension of the question feature set obtained after concatenating them is m + n + p.

[0033] During the concatenation process, it is necessary to pay attention to maintaining the order and dimensional consistency of the feature vectors. The semantic vector feature can be placed in the front, followed by the context correlation feature, and finally the domain correlation feature. The question feature set obtained by such concatenation combines the semantic information, context information of the user question statement, and the correlation information with the low-altitude economy education field, providing a comprehensive and rich feature basis for subsequent retrieval and response generation.

[0034] Step S130: Retrieve the domain knowledge feature set in the low-altitude economy education knowledge base based on the question feature set.

[0035] Step S131: Perform multi-level knowledge matching in the low-altitude economy education knowledge base according to the question feature set. The multi-level knowledge matching includes a core concept matching layer, an application scenario matching layer, and a technical standard matching layer.

[0036] The low-altitude economy education knowledge base is a database containing a large amount of knowledge in the field of low-altitude economy, which is organized according to different categories and levels. When performing multi-level knowledge matching, it starts from the core concept matching layer. The core concept matching layer mainly focuses on the semantic vector features in the problem feature set, especially the keyword distribution weights among them. According to these keywords, directly related core knowledge points are searched in the low-altitude economy education knowledge base.

[0037] For example, for the keyword "low-altitude logistics distribution mode" in the semantic vector features of the problem feature set, all core knowledge points related to "low-altitude logistics distribution mode" are searched in the low-altitude economy education knowledge base, such as the definition and basic principles of this mode.

[0038] Next is the application scenario matching layer. This layer mainly searches for extended knowledge points associated with the implicit application scenario of the user's problem statement based on the context association features in the problem feature set. For example, if the user's problem statement mentions a comparison of low-altitude logistics distribution modes under a specific development trend, then extended knowledge points such as the application scenario of this mode under this development trend, as well as relevant cases and data, are searched in the low-altitude economy education knowledge base.

[0039] Finally, there is the technical standard matching layer. This layer filters out standard knowledge points that match the technical standard category to which the user's problem statement belongs based on the domain association features in the problem feature set. For example, if the user's problem involves cost control and timeliness of low-altitude logistics distribution modes, then standard knowledge points such as relevant industry technical standards and specifications are searched in the low-altitude economy education knowledge base.

[0040] Step S132: In the core concept matching layer, based on the keyword distribution weights in the semantic vector features, retrieve the core knowledge point features directly associated with the user's problem statement in the low-altitude economy education knowledge base.

[0041] In the core concept matching layer, the keyword distribution weights in the semantic vector features play a crucial role. These keyword distribution weights reflect the importance of different keywords in the user's problem statement. The knowledge in the knowledge base can be filtered and sorted according to these weights. For keywords with higher weights, key attention is paid to searching for directly related core knowledge points in the low-altitude economy education knowledge base.

[0042] For example, for the keyword "low-altitude logistics distribution mode", due to its relatively high weight in the semantic vector features, all core knowledge points such as the core definition, basic constituent elements, and main types of this mode can be searched in the low-altitude economy education knowledge base. These core knowledge points are extracted and characterized to form core knowledge point features, which contain core concept information directly related to the user's problem statement.

[0043] Step S133: In the application scenario matching layer, according to the context association features of the user question statement, retrieve the extended knowledge point features associated with the implied application scenario of the user question statement in the low-altitude economy education knowledge base.

[0044] The context association features reflect the information of the user question statement in the historical context and actual application. In the application scenario matching layer, the extended knowledge points related to the implied application scenario of the user question can be found in the low-altitude economy education knowledge base based on these context association features.

[0045] For example, if it is known from the context association features that the user question is to compare emerging and traditional low-altitude logistics distribution models under a specific trend of low-altitude economy development. Then, specific application cases of this model under this development trend can be found in the low-altitude economy education knowledge base, such as the cost control strategies and timeliness performance in actual operation of the emerging models adopted by different enterprises. Extract and characterize these extended knowledge points to form extended knowledge point features, which enrich the knowledge related to the user question and provide more practical application information.

[0046] Step S134: In the technical standard matching layer, based on the domain association features, filter out the standard knowledge point features in the low-altitude economy education knowledge base that match the technical standard category to which the user question statement belongs.

[0047] The domain association features quantify the association strength between the user question statement and the core knowledge system of low-altitude economy education, and at the same time reflect the technical standard category to which the user question belongs. In the technical standard matching layer, the standard knowledge points that match the technical standard category to which the user question belongs can be filtered out in the low-altitude economy education knowledge base based on these domain association features.

[0048] For example, if the domain association features indicate that the user question involves cost control and timeliness of low-altitude logistics distribution models, then relevant industry technical standards and specifications, such as cost accounting standards and timeliness evaluation indicators, can be found in the low-altitude economy education knowledge base. Extract and characterize these standard knowledge points to form standard knowledge point features, which ensure that the response can meet the requirements of industry technical standards and specifications.

[0049] Step S135: Perform hierarchical fusion processing on the core knowledge point features, the extended knowledge point features, and the standard knowledge point features according to a preset priority to generate the domain knowledge feature set, where the core knowledge point features default to the highest priority in subsequent dynamic weight assignment.

[0050] After obtaining the core knowledge point features, extended knowledge point features, and standard knowledge point features, it is necessary to perform hierarchical fusion processing on them. The preset priority stipulates the importance of different features in the fusion process. Since the core knowledge point features are the core concepts directly related to the user's question, they default to the highest priority in subsequent dynamic weight allocation.

[0051] The process of fusion processing can be to splice the three feature sets in the order of priority. First, place the core knowledge point features at the front, and then sequentially splice the extended knowledge point features and standard knowledge point features. The resulting domain knowledge feature set synthesizes knowledge information in multiple aspects such as core concepts, application scenarios, and technical standards, providing comprehensive and accurate knowledge support for subsequent response generation.

[0052] Step S140: Perform cross-dimensional feature fusion processing on the problem feature set and the domain knowledge feature set to generate an optimized response statement for the user question statement.

[0053] Step S141: Analyze the knowledge adaptability of the semantic vector features in the problem feature set to determine the semantic matching degree between the semantic vector features and each knowledge point in the domain knowledge feature set.

[0054] Knowledge adaptability analysis mainly focuses on the semantic similarity between the semantic vector features in the problem feature set and each knowledge point in the domain knowledge feature set. A method based on vector similarity calculation can be used to determine the semantic matching degree. For example, use the cosine similarity algorithm to compare the semantic vector features with the feature vectors of each knowledge point in the domain knowledge feature set.

[0055] For the semantic vector feature and the feature vector of a certain knowledge point, calculate the cosine value between them. The closer the cosine value is to 1, the more similar the directions of the two vectors are, that is, the higher the semantic matching degree; the closer the cosine value is to 0, the lower the semantic matching degree. In this way, a semantic matching degree value is calculated for each knowledge point in the domain knowledge feature set.

[0056] Step S142: Based on the semantic matching degree, construct a dynamic weight allocation mechanism to prioritize the knowledge points in the domain knowledge feature set and obtain a priority ranking result.

[0057] According to the calculated semantic matching degree, construct a dynamic weight allocation mechanism. The higher the semantic matching degree of a knowledge point, the higher the weight it will obtain in weight allocation. For example, the semantic matching degree can be used as an important parameter of the weight, and the semantic matching degree can be converted into a weight value through a linear function or a non-linear function.

[0058] Then, based on these weight values, the knowledge points in the domain knowledge feature set are sorted by priority. The knowledge points with higher weight values are ranked higher in the sorting. The obtained priority sorting result can reflect the degree of relevance of each knowledge point to the user's question, providing a basis for subsequent feature fusion and response generation.

[0059] Step S143: According to the priority sorting result, perform feature dimension projection processing on the top N priority knowledge point features before the priority sorting and the context association features in the question feature set respectively, generate projection knowledge features and projection question features with the same dimension, and perform fusion processing on the projection knowledge features and projection question features based on the bidirectional attention mechanism to generate knowledge-enhanced semantic features.

[0060] According to the priority sorting result, select the top N priority knowledge point features before the sorting. Perform feature dimension projection processing on these priority knowledge point features and the context association features in the question feature set respectively. The purpose of feature dimension projection processing is to convert feature vectors of different dimensions into the same dimension for subsequent fusion processing.

[0061] The feature dimension projection can be achieved by using methods such as linear transformation or non-linear mapping. For example, project the priority knowledge point features and context association features into the same feature space through a projection matrix to generate projection knowledge features and projection question features.

[0062] Then, perform fusion processing on the projection knowledge features and projection question features based on the bidirectional attention mechanism. The bidirectional attention mechanism can simultaneously focus on the mutual relationship between the projection knowledge features and projection question features, and fuse the information of the two by learning the attention weights between them. Specifically, the attention mechanism calculates the degree of association between each element in the projection knowledge features and projection question features, assigns attention weights according to the degree of association, and then combines the weighted features to generate knowledge-enhanced semantic features. This knowledge-enhanced semantic feature integrates the priority knowledge points and context information, enhancing the semantic expression ability.

[0063] Step S144: Input the knowledge-enhanced semantic features into the response generation model based on the multi-level attention mechanism to generate an initial response statement containing terms in the low-altitude economy education field and the source identifier of the knowledge point.

[0064] The response generation model based on the multi-level attention mechanism consists of multiple key modules and levels to effectively process knowledge-enhanced semantic features and generate response sentences. This model mainly includes an input layer, a multi-level attention layer, and an output layer. The input layer is responsible for receiving knowledge-enhanced semantic features and using them as the initial data for the response generation model to process. The multi-level attention layer is the core part of the response generation model, which is composed of multiple attention sub-layers, and each sub-layer has different focuses and functions.

[0065] In the multi-level attention layer, first is the word-level attention sub-layer. This word-level attention sub-layer focuses on each word in the knowledge-enhanced semantic features and calculates the importance weights of each word when generating response sentences. Through the attention allocation at the word level, the response generation model can highlight key terms in the low-altitude economic education field, ensuring that these terms are accurately expressed in the response. For example, for the knowledge-enhanced semantic features related to the low-altitude logistics distribution mode, the word-level attention sub-layer will identify key terms such as "low-altitude logistics" and "distribution mode" and assign them higher weights.

[0066] Next is the sentence-level attention sub-layer, which considers the structure and semantic information of the knowledge-enhanced semantic features from the sentence level and calculates the importance of each sentence in response generation. This helps the model organize the logical order of the sentences, making the generated response sentences have good coherence and logic. For example, when describing the advantages and disadvantages of the low-altitude logistics distribution mode, the sentence-level attention sub-layer will reasonably arrange the order of the sentences describing the advantages and disadvantages, making the response easier to understand.

[0067] Then is the discourse-level attention sub-layer. This discourse-level attention sub-layer analyzes the knowledge-enhanced semantic features from the perspective of the overall discourse, considering the relevance of the response sentences to the entire low-altitude economic education field knowledge system. It can ensure that the generated response conforms to the logic and norms of the field at the macro level and can cover the necessary knowledge points. For example, when answering questions about the development trends of the low-altitude economy, the discourse-level attention sub-layer will comprehensively consider different aspects of the development trends to make the response comprehensive.

[0068] The multi-level attention layer integrates and filters the information in the knowledge-enhanced semantic features by continuously adjusting and allocating attention weights. In this process, the response generation model will perform weighted processing on the knowledge-enhanced semantic features according to the attention weights, highlighting important information and weakening unimportant information.

[0069] After being processed by multiple attention layers, the information is passed to the output layer. The main task of the output layer is to generate a specific response statement based on the processing results of the multiple attention layers. It combines the language expressions and grammar rules in the low-altitude economy education field to transform the processed information into natural and fluent text. At the same time, the output layer will embed the knowledge source identifier in the generated response statement. The embedding of the knowledge source identifier is achieved by extracting and annotating the knowledge-related information recorded during the processing of the multiple attention layers. For example, when the response generation model processes the knowledge-enhanced semantic features, it will record the source of each knowledge point. When outputting the response statement, these source information will be added to the statement in a set identifier form so that users can understand the knowledge origin of the response content.

[0070] Through the above steps, the response generation model based on the multi-level attention mechanism can transform the knowledge-enhanced semantic features into an initial response statement containing terms in the low-altitude economy education field and knowledge source identifiers.

[0071] Step S145: Perform an interpretability optimization process on the initial response statement, embed the knowledge source identifier corresponding to the user question statement in the initial response statement, and generate the optimized response statement.

[0072] The purpose of performing an interpretability optimization process on the initial response statement is to improve the transparency and credibility of the response, enabling users to clearly understand the basis of the response content. When performing the interpretability optimization process, it is first necessary to perform a structural analysis of the initial response statement. Through natural language processing techniques, each sentence and key information in the initial response statement are identified. For example, for an initial response statement about low-altitude logistics cost control strategies, the sentences describing different cost control strategies and the key terms involved are analyzed.

[0073] Then, according to the knowledge source information recorded in the response generation model before, the corresponding knowledge source identifier is embedded in the initial response statement. The embedding method can be to add a set identifier symbol after each key information or sentence and attach the corresponding knowledge source description. For example, add an identifier symbol, such as "[Source: Document Y in Knowledge Base X]", after the sentence describing a certain cost control strategy. In this way, when users view the response statement, they can intuitively understand the knowledge source of each part.

[0074] During the process of embedding the knowledge source identifier, attention needs to be paid to the accuracy and consistency of the identifier. Ensure that the knowledge source corresponding to each identifier is correct, and the formats and expressions of all identifiers are unified. At the same time, avoid the identifiers being too long or complex, so as not to affect the readability of the response statement.

[0075] Through the above interpretability optimization process, the knowledge source identifier is accurately embedded into the initial response statement, and finally an optimized response statement is generated. This optimized response statement not only provides an accurate answer but also enables the user to clearly understand the basis of the answer, improving the quality and credibility of the response.

[0076] Step S150: Dynamically adjust the response service strategy of the interactive learning platform based on the difference features between the optimized response statement and the historical response statements in the user interaction data.

[0077] Step S151: Perform in-depth semantic parsing on the optimized response statement and the historical response statements respectively to generate a first semantic parsing tree and a second semantic parsing tree, and extract a first set of core knowledge point nodes from the first semantic parsing tree and a second set of core knowledge point nodes from the second semantic parsing tree.

[0078] In-depth semantic parsing is a process of deeply analyzing the optimized response statement and the historical response statements, aiming to reveal the semantic structure and core knowledge points of the statements. For the optimized response statement, a special semantic parser is used for processing. The semantic parser will perform part-of-speech tagging, syntactic analysis, and semantic understanding on the statement to construct the first semantic parsing tree. The first semantic parsing tree is a tree structure, where the nodes represent different semantic units in the statement and the edges represent the semantic relationships between these units.

[0079] For example, for the optimized response statement "In the development of the low-altitude economy, the emerging low-altitude logistics and distribution model has improved timeliness through technological innovation, and cost control is also more effective [Source: Document B in Knowledge Base A]", the semantic parser will identify semantic units such as "low-altitude economy development", "emerging low-altitude logistics and distribution model", "technological innovation", "timeliness", and "cost control", and construct the relationships between them to form the first semantic parsing tree.

[0080] Similarly, perform in-depth semantic parsing on the historical response statement to generate the second semantic parsing tree. For example, for the historical response statement "The traditional low-altitude logistics and distribution model mainly relies on economies of scale to control costs", the second semantic parsing tree is constructed after semantic parsing.

[0081] After obtaining the first semantic parsing tree and the second semantic parsing tree, it is necessary to extract the set of core knowledge point nodes therein. A core knowledge point node refers to a node that represents key knowledge information in the semantic parsing tree. For the first semantic parsing tree, through traversing and analyzing the tree, core knowledge point nodes such as "emerging low-altitude logistics distribution model", "technological innovation", "timeliness", "cost control", etc. are identified and form the first set of core knowledge point nodes. For the second semantic parsing tree, core knowledge point nodes such as "traditional low-altitude logistics distribution model", "scale effect", "cost control", etc. are extracted and form the second set of core knowledge point nodes.

[0082] Step S152: Calculate the knowledge overlap degree and knowledge difference degree between the first set of core knowledge point nodes and the second set of core knowledge point nodes. The knowledge overlap degree is used to represent the proportion of knowledge points jointly covered in the response statements before and after optimization, and the knowledge difference degree is used to represent the distribution difference between the newly added knowledge points in the response statement after optimization and the missing knowledge points in the response statement before optimization.

[0083] The process of calculating the knowledge overlap degree and knowledge difference degree is achieved by comparing and analyzing the first set of core knowledge point nodes and the second set of core knowledge point nodes. First, find the same knowledge point nodes in the two sets. For example, there is a knowledge point node of "cost control" in both the first set of core knowledge point nodes and the second set of core knowledge point nodes.

[0084] Compare the number of the same knowledge point nodes with the total number of knowledge point nodes in the two sets to calculate the knowledge overlap degree. Specifically, let the number of elements in the first set of core knowledge point nodes be A, the number of elements in the second set of core knowledge point nodes be B, and the number of the same knowledge point nodes be C. The calculation of the knowledge overlap degree can be obtained by dividing C by the larger number of A and B. The higher the knowledge overlap degree, the greater the proportion of knowledge points jointly covered in the response statements before and after optimization.

[0085] For the calculation of the knowledge difference degree, it is necessary to separately find the newly added knowledge point nodes that exist in the first set of core knowledge point nodes but do not exist in the second set of core knowledge point nodes, and the missing knowledge point nodes that exist in the second set of core knowledge point nodes but do not exist in the first set of core knowledge point nodes.

[0086] Then, analyze the distribution of the newly added knowledge point nodes and the missing knowledge point nodes. By classifying and counting the knowledge point nodes, the quantity differences of them in different knowledge fields or topics can be calculated. For example, there are multiple newly added knowledge point nodes in technological innovation, and there are some missing knowledge point nodes in the traditional operation mode. By comparing these quantity differences to determine the knowledge difference degree. The knowledge difference degree can reflect the changes in the knowledge coverage of the response statement after optimization compared with that before optimization.

[0087] Step S153: Construct a weighted difference scoring function based on the knowledge overlap degree and the knowledge difference degree to generate a semantic difference degree, which is used to quantify the knowledge coverage difference before and after response optimization.

[0088] Constructing a weighted difference scoring function requires comprehensive consideration of the impacts of the knowledge overlap degree and the knowledge difference degree. To determine the weights of the two in the scoring function, adjustments need to be made according to the actual application scenarios and requirements. For example, if more emphasis is placed on the new knowledge in the optimized response statement, a higher weight can be assigned to the knowledge difference degree; if it is desired to ensure the coherence of knowledge before and after optimization, a higher weight can be assigned to the knowledge overlap degree.

[0089] Let the knowledge overlap degree be K1, the knowledge difference degree be K2, the corresponding weights be W1 and W2 respectively, and W1 + W2 = 1. The weighted difference scoring function can be realized by weighted combination of K1 and K2. Specifically, the semantic difference degree D can be calculated by the formula D = W1 * (1 - K1) + W2 * K2. Among them, (1 - K1) represents the non-overlapping part of the knowledge, which is weighted and combined with the knowledge difference degree to obtain a comprehensive scoring value, that is, the semantic difference degree.

[0090] The semantic difference degree can quantify the knowledge coverage difference before and after response optimization. The larger the semantic difference degree, the greater the change in the knowledge content of the optimized response statement compared with that before optimization; the smaller the semantic difference degree, the more similar the knowledge coverage before and after optimization.

[0091] Step S154: Determine the response quality evaluation indicators of the interactive learning platform according to the semantic difference degree. The response quality evaluation indicators include the knowledge accuracy increment, the term coverage increment, and the user understanding increment.

[0092] The semantic difference degree reflects the knowledge coverage difference before and after response optimization. Based on this, the response quality evaluation indicators of the interactive learning platform can be determined. For the knowledge accuracy increment, when the semantic difference degree is large, it indicates that there are significant changes in the knowledge content of the optimized response statement, and it is necessary to evaluate whether these changes improve the knowledge accuracy. It can be compared with an authoritative low-altitude economy education knowledge base to check whether the knowledge points in the optimized response statement are accurate. If the knowledge points in the optimized response statement are more consistent with the content of the authoritative knowledge base, then the knowledge accuracy increment is positive; otherwise, it is negative.

[0093] The term coverage increment refers to the change in the coverage of terms in the low-altitude economy education field in the optimized response statement compared with that before optimization. It can be calculated by counting the number and types of domain terms in the response statements before and after optimization. If the optimized response statement covers more domain terms, then the term coverage increment is positive; otherwise, it is negative.

[0094] The increment of user understanding needs to be evaluated through user feedback or simulated user understanding tests. Some users can be invited to evaluate the response statements before and after optimization, or the complexity and clarity of the response statements can be analyzed through natural language processing techniques. If users think the response statements after optimization are more understandable, then the increment of user understanding is positive; otherwise, it is negative.

[0095] Step S155: Generate a service strategy adjustment instruction based on the response quality evaluation index, and the service strategy adjustment instruction includes a knowledge base update instruction, a semantic encoder optimization instruction, and a response generation adjustment instruction.

[0096] According to the response quality evaluation index, generate the corresponding service strategy adjustment instruction. When the increment of knowledge accuracy is low, it indicates that the knowledge in the knowledge base may be inaccurate or incomplete, and a knowledge base update instruction needs to be generated. The knowledge base update instruction will trigger the inspection and update operations of the low-altitude economy education knowledge base. Specifically, the knowledge points in the response statements after optimization will be compared with the content in the knowledge base, and the inaccurate or missing knowledge points will be corrected and supplemented. At the same time, the knowledge points in the knowledge base will be classified and sorted to improve the quality and usability of the knowledge base.

[0097] If the increment of term coverage is low, it indicates that the semantic encoder may be insufficient in identifying and processing terms in the low-altitude economy education field, and a semantic encoder optimization instruction needs to be generated. The semantic encoder optimization instruction will guide the optimization of the pre-trained semantic encoder in the low-altitude economy field. It can be trained by adding more domain term data, adjusting the parameters and model structure of the encoder, and improving its ability to identify and encode domain terms.

[0098] When the increment of user understanding is low, it indicates that the response generation model needs to be improved in terms of the readability and logic of the generated response statements, and a response generation adjustment instruction needs to be generated. The response generation adjustment instruction will be used to adjust the response generation model based on the multi-level attention mechanism. The output layer structure of the model can be optimized, and the algorithm for generating statements can be improved to make the generated response statements more natural, fluent, and easy to understand.

[0099] Step S156: Incrementally optimize the real-time response service process of the interactive learning platform according to the service strategy adjustment instruction, and the incremental optimization includes updating the distribution weights of the hot knowledge points in the low-altitude economy education knowledge base, adjusting the semantic optimization parameters of the semantic optimization model in the low-altitude economy field, and optimizing the response generation parameters of the response generation model.

[0100] Step S1561: Update the distribution weights of the hot knowledge points in the low-altitude economy education knowledge base.

[0101] Extract the set of core knowledge point identifiers associated with the optimized response statement based on the knowledge point source identifier embedded in the optimized response statement. First, parse the knowledge point source identifier in the optimized response statement to identify the identifier information of the core knowledge points contained therein. For example, for the content in the optimized response statement labeled as "[Source: Document Y in Knowledge Base X, Knowledge Point Z]", extract the identifier of Knowledge Point Z. Collect all such core knowledge point identifiers to form the set of core knowledge point identifiers.

[0102] Statistically analyze the user query frequency and associated response citation frequency of the set of core knowledge point identifiers within a preset time window. By analyzing the log data of the interactive learning platform, count the number of times each core knowledge point identifier is queried by users within the preset time window to obtain the user query frequency. At the same time, count the number of times this core knowledge point identifier is cited in the response statement to obtain the associated response citation frequency.

[0103] Generate a real-time popularity score for each core knowledge point through a weighted fusion algorithm based on the normalized user query frequency and associated response citation frequency. Since the numerical ranges of the user query frequency and associated response citation frequency may be different, they need to be normalized to bring them to the same scale. Let the normalized user query frequency be F1, the associated response citation frequency be F2, and the corresponding weights be W3 and W4, and W3 + W4 = 1. The real-time popularity score R can be calculated using the formula R = W3 * F1 + W4 * F2. Among them, the weights W3 and W4 are dynamically adjusted according to the preset teaching scenario. For example, in a teaching scenario that emphasizes user initiative in querying, a higher weight can be assigned to W3; in a teaching scenario that emphasizes response citation, a higher weight can be assigned to W4.

[0104] Adjust the retrieval priority weight of the corresponding knowledge point in the low-altitude economy education knowledge base based on the real-time popularity score. The higher the real-time popularity score of a knowledge point, the higher the feature matching weight it needs to obtain in the subsequent response generation process. Therefore, adjust the retrieval priority weight of the corresponding knowledge point in the knowledge base according to the real-time popularity score. For example, increase the retrieval priority weight of the knowledge points with high real-time popularity scores so that the system can find these hot knowledge points faster when retrieving the knowledge base, improving the accuracy and relevance of the response.

[0105] Step S1562: Adjust the semantic optimization parameters of the semantic optimization model in the low-altitude economy field.

[0106] Extract the feature matching deviation between the semantic vector features of the user question statement and the knowledge graph embedding vectors in the low-altitude economy education knowledge base. First, calculate the similarity between the semantic vector features of the user question statement and the knowledge graph embedding vectors. A vector similarity calculation method, such as the cosine similarity algorithm, can be used. Compare the calculated similarity with a preset ideal similarity to obtain the feature matching deviation. The feature matching deviation reflects the degree of difference between the semantic vector features and the knowledge graph embedding vectors.

[0107] Construct a dynamic attenuation coefficient based on the domain association features. The domain association features quantify the association strength between the user question statement and the core knowledge system of low-altitude economy education. According to the magnitude of the domain association features, construct a dynamic attenuation coefficient. The stronger the domain association features, the smaller the dynamic attenuation coefficient; the weaker the domain association features, the larger the dynamic attenuation coefficient.

[0108] Perform weighted fusion on the feature matching deviation to generate a context-aware feature correction amount. Multiply the feature matching deviation by the dynamic attenuation coefficient to obtain the context-aware feature correction amount. This feature correction amount takes into account the context information of the user question statement and the degree of association with domain knowledge.

[0109] Backpropagate the feature correction amount to the term context association analysis layer of the low-altitude economy domain semantic optimization model. Through the backpropagation algorithm, transmit the feature correction amount to the term context association analysis layer. In this layer, synchronously adjust the hierarchical attention weights during the knowledge graph embedding process and the span parameter of the term co-occurrence sliding window. Adjusting the hierarchical attention weights can make the model pay more attention to the knowledge graph information related to the user question, and adjusting the span parameter of the term co-occurrence sliding window can better capture the context relationship between terms.

[0110] Based on the knowledge accuracy increment in the response quality evaluation index, verify the term recognition error rate of the adjusted semantic optimization model. By comparing the results processed by the adjusted semantic optimization model with the authoritative term annotation data, calculate the term recognition error rate. When the term recognition error rate decreases and the strength of the domain association features increases, it indicates that the performance of the adjusted semantic optimization model has been improved. At this time, lock the iterative update amplitude of the current hierarchical attention weights to avoid over-adjustment resulting in unstable model performance.

[0111] Step S1563: Optimize the response generation parameters of the response generation model.

[0112] Extract the mapping deviation between the knowledge point source identifier of the optimized response statement and the hierarchical fusion priority of the domain knowledge feature set. First, encode the knowledge point source identifier of the optimized response statement to obtain an identifier embedding vector. At the same time, encode the hierarchical fusion priority of the domain knowledge feature set to obtain a hierarchical priority vector. Then, calculate the orthogonal projection distance between the identifier embedding vector and the hierarchical priority vector, and use this orthogonal projection distance as the mapping deviation. The mapping deviation reflects the degree of inconsistency between the knowledge point source identifier and the hierarchical fusion priority.

[0113] Construct a cross-level gradient constraint function based on the semantic difference degree. The semantic difference degree quantifies the knowledge coverage difference before and after response optimization. According to the magnitude of the semantic difference degree, construct a cross-level gradient constraint function. This function is used to limit the update amplitude of the bidirectional attention projection parameters in the multi-level attention mechanism to avoid excessive parameter updates leading to model instability.

[0114] Perform backpropagation optimization on the bidirectional attention projection parameters in the multi-level attention mechanism. Apply the cross-level gradient constraint function to the backpropagation algorithm to adjust the bidirectional attention projection parameters. Through continuous iterative updates, the parameters gradually converge to the optimal values, adjust the non-linear alignment coefficient between the knowledge feature dimension and the problem feature dimension, and improve the model's ability to fuse knowledge and problem features.

[0115] Reconstruct the decoder parameters of the response generation model based on the user understanding degree increment. The user understanding degree increment reflects the user's understanding of the optimized response statement. Adjust the decoder parameters according to the magnitude of the user understanding degree increment. The knowledge application scenario features corresponding to the knowledge point source identifier can be temporally coupled with the context association features of the historical response statements to dynamically enhance the feature fusion intensity of the scenario adaptation layer during the decoding process. In this way, the generated response statements are more in line with the user's understanding habits and improve user satisfaction.

[0116] Synchronously adjust the dynamic interface parameters of the standard verification layer in the response generation model through the real-time update status of the standard knowledge point features, so that the technical standard compliance of the generated response statements is synchronized with the latest industry standard changes in the knowledge base.

[0117] The standard knowledge point features reflect the latest technical standards and specifications in the low-altitude economy education field, and their real-time update status reflects the dynamic changes of industry standards. The standard verification layer in the response generation model is responsible for ensuring that the generated response statements comply with the current technical standards. In order to enable the standard verification layer to adapt to the changes in industry standards in a timely manner, it is necessary to adjust its dynamic interface parameters according to the real-time update status of the standard knowledge point features.

[0118] First, a monitoring mechanism for the real-time update status of standard knowledge point features needs to be established. This mechanism can regularly obtain the update information of standard knowledge point features from the low-altitude economy education knowledge base. For example, when the knowledge points regarding the safety standards of low-altitude aircraft in the knowledge base are updated, the monitoring mechanism will capture this change and record the updated content and time.

[0119] Then, based on the monitored update information, analyze the impact of these changes on the standard verification layer. Different updates of standard knowledge points may affect different interface parameters. For example, if the update involves new cost accounting standards, then the interface parameters related to cost accounting in the standard verification layer need to be adjusted.

[0120] Next, perform specific adjustment operations on the dynamic interface parameters of the standard verification layer. This process needs to determine the adjustment method and amplitude according to the updated standard knowledge point features and the structure of the response generation model. For some simple updates, the preset values of the interface parameters can be directly modified. For example, if the new standard stipulates the time limit index range of a certain low-altitude logistics distribution mode, then the interface parameters corresponding to the time limit verification in the standard verification layer can be directly updated to the new range value.

[0121] For more complex updates, it may be necessary to retrain or optimize the interface parameters. For example, when there are major changes in industry standards, involving adjustments to technical specifications in multiple aspects, it is necessary to use the new standard knowledge point feature data to retrain the interface parameters of the standard verification layer. This can be achieved by inputting the new data into the response generation model and using the backpropagation algorithm to adjust the interface parameters so that they can accurately verify whether the response statements comply with the new standards.

[0122] During the process of adjusting the dynamic interface parameters, it is also necessary to consider the impact of parameter adjustment on the entire response generation model. To avoid a decline in model performance or instability caused by parameter adjustment, sufficient testing and verification are required after parameter adjustment. A set of test data can be used to evaluate whether the response statements generated by the adjusted model comply with the new technical standards and whether the overall performance of the model remains stable.

[0123] At the same time, in order to ensure that the compliance of the generated response statements with technical standards is synchronized with the latest changes in industry standards in the knowledge base, a feedback mechanism also needs to be established. When the standard verification layer finds that the response statements do not comply with the latest standards during actual operation, it can promptly feedback this information to the monitoring mechanism and the parameter adjustment module. The monitoring mechanism will check the update status of the standard knowledge point features again, and the parameter adjustment module will further adjust the dynamic interface parameters according to the feedback information, forming a closed-loop synchronous adjustment system.

[0124] By following the above steps and continuously adjusting the dynamic interface parameters of the standard verification layer in the response generation model according to the real-time update status of the standard knowledge point features, the generated response statements can always meet the latest technical standards in the field of low-altitude economy education, improving the accuracy and professionalism of the responses.

[0125] In the whole implementation process of the intelligent Q&A method for low-altitude economy education on the interactive learning platform, each step is closely linked, forming a complete closed-loop system. Starting from obtaining the user question data set, through question feature extraction, domain knowledge retrieval, feature fusion to generate optimized response statements, and then to adjusting the response service strategy according to the differences between the optimized response and the historical response, each link plays a key role in improving the performance and service quality of the Q&A system.

[0126] In terms of data collection, the platform ensures that the system can obtain the latest and most comprehensive user demand information by continuously collecting the low-altitude economy education question data set submitted by users on the interactive learning platform. In the feature extraction stage, in-depth semantic analysis and feature mining are carried out on the user question statements, including the extraction of semantic vector features, domain association features, and context association features, providing a rich and accurate feature basis for subsequent knowledge retrieval and response generation.

[0127] In the process of knowledge retrieval, through multi-level knowledge matching, a set of domain knowledge features related to the user question is screened out from the low-altitude economy education knowledge base, including core knowledge point features, extended knowledge point features, and standard knowledge point features. The fusion and priority ranking of these feature sets enable the system to more accurately locate and utilize relevant knowledge.

[0128] In the feature fusion and response generation stage, through cross-dimensional feature fusion processing, the question feature set and the domain knowledge feature set are organically combined to generate optimized response statements containing low-altitude economy education field terms and knowledge point source identifiers. At the same time, interpretability optimization processing is carried out on the optimized response statements, improving the transparency and credibility of the responses.

[0129] Finally, by comparing the difference features between the optimized response statements and the historical response statements, the response service strategy of the interactive learning platform is dynamically adjusted. This includes operations such as updating the distribution weights of hot knowledge points in the knowledge base, adjusting the semantic optimization parameters of the semantic optimization model, and optimizing the response generation parameters of the response generation model. These adjustment measures can enable the system to continuously adapt to the changes in industry standards and the diversity of user needs, improving the quality of responses and the effect of services.

[0130] Throughout the process, attention should also be paid to the protection of privacy-sensitive data. During the data collection phase, the problem data submitted by users may contain some sensitive information. To prevent the leakage of this information, data encryption technology can be used to encrypt the data. For example, symmetric encryption algorithms can be used to encrypt the user's question statements and historical response statements, and only authorized system modules can decrypt and process them. At the same time, corresponding security measures also need to be taken during data storage and transmission, such as using secure network protocols and storage devices to ensure the security and integrity of the data.

[0131] In addition, for the training and optimization of artificial intelligence models, certain principles and methods need to be followed. When training the model, a large amount of labeled data is required for supervised learning. These labeled data should be representative and accurate, and be able to reflect the actual situation in the field of low-altitude economy education. During the model training process, appropriate training parameters need to be set, such as the learning rate, the number of iterations, etc., to ensure that the model can converge to the optimal solution. At the same time, the model also needs to be evaluated and verified, and different evaluation metrics are used to measure the performance of the model, such as accuracy, recall rate, etc.

[0132] In terms of model optimization, the parameters and structure of the model need to be continuously adjusted according to the feedback information in actual applications. For example, when it is found that the model performs poorly on certain types of problems, the performance of the model can be improved by increasing the training data, adjusting the number of layers or neurons of the model, etc.

[0133] In summary, by implementing the above-mentioned intelligent question-answering method for low-altitude economy education on the interactive learning platform, more accurate, professional, and interpretable response services can be provided to users, while continuously optimizing the performance and service strategies of the system to adapt to the development of the low-altitude economy education field and the changes in user needs.

[0134] Figure 2 FIG. shows a schematic diagram of exemplary hardware and software components of an intelligent question-answering system 100 for low-altitude economy education on an interactive learning platform that can implement the ideas of the present application. For example, the processor 120 can be used on the intelligent question-answering system 100 for low-altitude economy education on the interactive learning platform and is used to execute the functions in the present application.

[0135] The intelligent question-answering system 100 for low-altitude economy education on the interactive learning platform can be a general-purpose server or a special-purpose server, both of which can be used to implement the intelligent question-answering method for low-altitude economy education on the interactive learning platform of the present application. Although only one server is shown in the present application, for convenience, the functions described in the present application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0136] For example, the low-altitude economy education intelligent Q&A system 100 for an interactive learning platform may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and different forms of storage media 140, such as disks, ROMs, or RAMs, or any combination thereof. Exemplarily, the low-altitude economy education intelligent Q&A system 100 for an interactive learning platform may further include program instructions stored in a ROM, a RAM, or other types of non-transitory storage media, or any combination thereof. The methods of the present application can be implemented according to these program instructions. The low-altitude economy education intelligent Q&A system 100 for an interactive learning platform further includes an I / O interface 150 between the computer and other input / output devices.

[0137] For ease of illustration, only one processor is described in the low-altitude economy education intelligent Q&A system 100 for an interactive learning platform. However, it should be noted that the low-altitude economy education intelligent Q&A system 100 in the present application may further include multiple processors. Therefore, the steps executed by one processor described in the present application may also be jointly executed or separately executed by multiple processors. For example, if the processor of the low-altitude economy education intelligent Q&A system 100 for an interactive learning platform executes step A and step B, it should be understood that step A and step B may also be jointly executed by two different processors or separately executed in one processor. For example, the first processor executes step A, the second processor executes step B, or the first processor and the second processor jointly execute steps A and B.

[0138] In addition, an embodiment of the present invention further provides a readable storage medium, in which computer-executable instructions are preset. When the processor executes the computer-executable instructions, the above-mentioned low-altitude economy education intelligent Q&A method for an interactive learning platform is implemented.

[0139] It should be noted that, in order to simplify the description of the present invention disclosure and thus help the understanding of one or more embodiments of the invention, in the previous description of the embodiments of the present invention, sometimes multiple features are merged into one embodiment, drawing, or description thereof.

Claims

1. An intelligent Q&A method for low-altitude economy education in an interactive learning platform, characterized in that, The method includes: Obtaining a set of low-altitude economy education problem data submitted by a user on an interactive learning platform, where the set of low-altitude economy education problem data includes multiple user interaction data, and each user interaction data is composed of at least one user question statement and a corresponding historical response statement; Extracting a set of problem features of the user interaction data, where the set of problem features includes semantic vector features of the user question statement, domain association features between the user question statement and the low-altitude economy education field, and context association features of the user question statement; Retrieving a set of domain knowledge features in a low-altitude economy education knowledge base based on the set of problem features; Performing cross-dimensional feature fusion processing on the set of problem features and the set of domain knowledge features to generate an optimized response statement for the user question statement; Dynamically adjusting the response service strategy of the interactive learning platform based on the difference features between the optimized response statement and the historical response statements in the user interaction data.

2. The intelligent Q&A method for low-altitude economy education for an interactive learning platform according to claim 1, wherein The extracting the set of problem features of the user interaction data includes: Performing semantic encoding processing on the user question statement to generate semantic vector features of the user question statement, where the semantic vector features include the keyword distribution weights of core concepts in the user question statement; Extracting response semantic features of the historical response statements in the user interaction data, and performing temporal alignment processing on the response semantic features and the semantic vector features to generate context association features of the user question statement; Determining the domain association features of the user question statement according to the term coverage of the user question statement in the low-altitude economy education field, where the domain association features are used to quantify the association strength between the user question statement and the core knowledge system of low-altitude economy education; Performing feature splicing on the semantic vector features, the context association features, and the domain association features to generate the set of problem features.

3. The intelligent question-answering method for low-altitude economy education for an interactive learning platform according to claim 2, wherein, The performing semantic encoding processing on the user question statement to generate semantic vector features of the user question statement includes: Invoking a pre-trained semantic encoder in the low-altitude economy field to perform word segmentation processing on the user question statement to obtain multiple semantic word units; Performing context dependence analysis on the semantic word units to generate local context weights for each semantic word unit, where the local context weights are used to represent the semantic contribution degree of the semantic word unit in the user question statement; Performing weighted aggregation processing on the semantic word units according to the local context weights to generate an initial semantic vector of the user question statement; Inputting the initial semantic vector into a semantic optimization model in the low-altitude economy field, and performing domain knowledge enhancement processing on the initial semantic vector through knowledge graph embedding and term context association analysis to generate semantic vector features including the distribution features of low-altitude economy professional terms.

4. The intelligent question-answering method for low-altitude economy education for an interactive learning platform according to claim 1, characterized in that, The retrieving a set of domain knowledge features in a low-altitude economy education knowledge base based on the set of problem features includes: Perform multi-level knowledge matching in the low-altitude economy education knowledge base according to the set of problem features, where the multi-level knowledge matching includes a core concept matching layer, an application scenario matching layer, and a technical standard matching layer; In the core concept matching layer, based on the keyword distribution weights in the semantic vector features, retrieve the core knowledge point features directly associated with the user's question statement in the low-altitude economy education knowledge base; In the application scenario matching layer, according to the context association features of the user's question statement, retrieve the extended knowledge point features associated with the implicit application scenario of the user's question statement in the low-altitude economy education knowledge base; In the technical standard matching layer, based on the domain association features, filter the standard knowledge point features in the low-altitude economy education knowledge base that match the technical standard category of the user's question statement; Perform hierarchical fusion processing on the core knowledge point features, the extended knowledge point features, and the standard knowledge point features according to a preset priority to generate the domain knowledge feature set, where the core knowledge point features default to the highest priority in subsequent dynamic weight assignment.

5. The intelligent question-answering method for low-altitude economy education for an interactive learning platform according to claim 1, characterized in that, The cross-dimensional feature fusion processing of the problem feature set and the domain knowledge feature set to generate the optimized response statement for the user's question statement includes: Perform a knowledge fitness analysis on the semantic vector features in the problem feature set to determine the semantic matching degree between the semantic vector features and each knowledge point in the domain knowledge feature set; Based on the semantic matching degree, construct a dynamic weight assignment mechanism to prioritize the knowledge points in the domain knowledge feature set to obtain a priority ranking result; According to the priority ranking result, perform feature dimension projection processing on the top N priority knowledge point features before the priority ranking and the context association features in the problem feature set respectively to generate projection knowledge features and projection problem features of the same dimension, and perform fusion processing on the projection knowledge features and projection problem features based on a bidirectional attention mechanism to generate knowledge-enhanced semantic features; Input the knowledge-enhanced semantic features into a response generation model based on a multi-level attention mechanism to generate an initial response statement containing terms in the low-altitude economy education field and knowledge point source identifiers; Perform an interpretability optimization process on the initial response statement, and embed the knowledge point source identifier corresponding to the user's question statement in the initial response statement to generate the optimized response statement.

6. The intelligent Q&A method for low-altitude economy education for an interactive learning platform according to claim 1, wherein The dynamic adjustment of the response service strategy of the interactive learning platform based on the difference features between the optimized response statement and the historical response statements in the user interaction data includes: Perform in-depth semantic parsing processing on the optimized response statement and the historical response statement respectively to generate a first semantic parsing tree and a second semantic parsing tree, and extract a first set of core knowledge point nodes in the first semantic parsing tree and a second set of core knowledge point nodes in the second semantic parsing tree; Calculate the knowledge overlap degree and knowledge difference degree between the first set of core knowledge point nodes and the second set of core knowledge point nodes. The knowledge overlap degree is used to represent the proportion of knowledge points jointly covered in the response statements before and after optimization, and the knowledge difference degree is used to represent the distribution difference between the newly added knowledge points in the optimized response statement and the missing knowledge points in the response statement before optimization; Construct a weighted difference scoring function based on the knowledge overlap degree and the knowledge difference degree to generate a semantic difference degree, which is used to quantify the knowledge coverage difference before and after response optimization; Determine the response quality evaluation index of the interactive learning platform according to the semantic difference degree. The response quality evaluation index includes the increment of knowledge accuracy, the increment of term coverage, and the increment of user understanding; Generate a service strategy adjustment instruction based on the response quality evaluation index. The service strategy adjustment instruction includes a knowledge base update instruction, a semantic encoder optimization instruction, and a response generation adjustment instruction; Incrementally optimize the real-time response service process of the interactive learning platform according to the service strategy adjustment instruction. The incremental optimization includes updating the distribution weight of hot knowledge points in the low-altitude economy education knowledge base, adjusting the semantic optimization parameters of the semantic optimization model in the low-altitude economy field, and optimizing the response generation parameters of the response generation model.

7. The intelligent question-answering method for low-altitude economy education for an interactive learning platform according to claim 6, wherein The update of the distribution weight of hot knowledge points in the low-altitude economy education knowledge base includes: Extract the set of core knowledge point identifiers associated with the optimized response statement according to the knowledge point source identifier embedded in the optimized response statement; Count the user query frequency and associated response citation frequency of the set of core knowledge point identifiers within a preset time window; Generate a real-time popularity score for each core knowledge point through a weighted fusion algorithm based on the normalized user query frequency and associated response citation frequency, where the weights of the user query frequency and the associated response citation frequency are dynamically adjusted according to the preset teaching scenario; Adjust the retrieval priority weight of the corresponding knowledge point in the low-altitude economy education knowledge base based on the real-time popularity score, so that high-frequency hot knowledge points obtain a higher feature matching weight in the subsequent response generation process.

8. The intelligent question-answering method for low-altitude economy education for an interactive learning platform according to claim 6, characterized in that, The adjustment of the semantic optimization parameters of the semantic optimization model in the low-altitude economy field includes: Extract the feature matching deviation between the semantic vector feature of the user question statement and the knowledge graph embedding vector in the low-altitude economy education knowledge base. The feature matching deviation is calculated by the normalized vector cosine similarity; Construct a dynamic decay coefficient according to the domain association feature, and perform weighted fusion on the feature matching deviation to generate a context-aware feature correction amount; Backpropagate the feature correction amount to the term context association analysis layer of the semantic optimization model in the low-altitude economy field, and synchronously adjust the hierarchical attention weight in the knowledge graph embedding process and the span parameter of the term co-occurrence sliding window; Based on the increment of knowledge accuracy in the response quality evaluation index, verify the term recognition error rate of the adjusted semantic optimization model. When the term recognition error rate decreases and the intensity of the domain association feature increases, lock the iterative update amplitude of the current hierarchical attention weight.

9. The intelligent question-answering method for low-altitude economy education for an interactive learning platform according to claim 6, characterized in that, The response generation parameters of the optimized response generation model include: extracting the mapping deviation between the knowledge point source identifier of the optimized response statement and the hierarchical fusion priority of the domain knowledge feature set, where the mapping deviation is calculated by the orthogonal projection distance between the identifier embedding vector and the hierarchical priority vector; constructing a cross-level gradient constraint function according to the semantic difference degree, performing backpropagation optimization on the bidirectional attention projection parameters in the multi-level attention mechanism, and adjusting the non-linear alignment coefficient between the knowledge feature dimension and the problem feature dimension; reconstructing the decoder parameters of the response generation model based on the user understanding degree increment, coupling the knowledge application scenario features corresponding to the knowledge point source identifier and the context association features of the historical response statement in time series, and dynamically enhancing the feature fusion intensity of the scenario adaptation layer in the decoding process; synchronously adjusting the dynamic interface parameters of the standard verification layer in the response generation model through the real-time update status of the standard knowledge point features, so that the technical standard compliance of the generated response statement is synchronized with the latest industry standard changes in the knowledge base.

10. An intelligent Q&A system for low-altitude economy education on an interactive learning platform, characterized in that, It includes a processor and a memory. The memory is connected to the processor. The memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the low-altitude economy education intelligent question-answering method for the interactive learning platform according to any one of claims 1-9 above.

Images