Deep geographic learning system based on interdisciplinary knowledge collaboration

Abstract

The invention relates to the technical field of intelligent education, in particular to a deep geographic learning system based on interdisciplinary knowledge collaboration, which comprises a geographic teaching unit, a multi-disciplinary knowledge collaboration module, a multi-disciplinary integrated deep mining module and an intelligent interaction unit, wherein the geography teaching unit is used for providing geography knowledge of the system; the multi-subject knowledge collaboration module is used for constructing a dynamic subject routing controller, and carrying out collaborative association on a target subject selected by a student and a geographic subject in combination with a deep learning model; the multidisciplinary integrated deep mining module is used for performing attribution analysis on geographic problems according to a PC algorithm and a GNN reasoning algorithm, extracting a causal relationship, quantifying causal strength and generating a visual causal graph; and the intelligent interaction unit is used for constructing an immersive and personalized learning scene and improving the interestingness of learning. Therefore, the problems of insufficient interdisciplinary fusion of geographical education, lack of autonomy, shallow geographical phenomenon mining and the like in the prior art are solved.

Description

Technical Field

[0001] This application relates to the field of intelligent education technology, and particularly to a deep geography learning system based on interdisciplinary knowledge collaboration. Background Art

[0002] In today's education system, geography education occupies an indispensable position. As a discipline that studies natural phenomena, human phenomena on the earth's surface and their interrelationships, it is of great significance to cultivate students' geographical literacy. Although the geography education in schools around the world shows a diverse development trend, generally speaking, there are some common current situation characteristics and deficiencies, especially in the depth and breadth of interdisciplinary knowledge integration, which urgently need to be improved and enhanced.

[0003] In related technologies, geography teaching realizes the connection of subject knowledge through teachers' lectures. In this way, the interdisciplinary knowledge connection depends on the subjective design of teachers. If students want to know the connection between specific geographical phenomena and a specific subject, they can only consult relevant materials, lacking tools for independent exploration of the internal connection between disciplines; the existing geography teaching lacks in-depth analysis of the interdisciplinary integration of geographical phenomena, currently only staying at the level of knowledge point splicing, lacking systematic causal relationship explanation and in-depth excavation of the causal mechanism of complex non-linear relationships; the interaction mode is mainly a combination of text and static charts, lacking immersive scenario simulation and real-time feedback, and students cannot dynamically adjust multi-disciplinary variables through operations, resulting in weak knowledge application ability. Summary of the Invention

[0004] This application provides a deep geography learning system based on interdisciplinary knowledge collaboration to solve the problems of insufficient interdisciplinary integration, lack of autonomy, and shallow excavation of geographical phenomena in the prior art.

[0005] The first aspect embodiment of the present application provides a deep geographical learning system based on interdisciplinary knowledge collaboration, including: a geographical teaching unit, a multi-disciplinary knowledge collaboration module, a multi-disciplinary integrated in-depth mining module, and an intelligent interaction unit. Among them, the geographical teaching unit is used to provide the geographical knowledge of the system to help students build a foundation in the geographical discipline; the multi-disciplinary knowledge collaboration module is used to construct a dynamic discipline routing controller according to the discipline dynamic matching mechanism, and in combination with the deep learning model, enable students to select one or more disciplines from multiple disciplines to be associated with the geographical discipline, and enhance students' understanding of geographical knowledge from the perspective of the selected target discipline. Among them, the multiple disciplines include biology, physics, history, agriculture, economy, and human sociology; the multi-disciplinary integrated in-depth mining module is used to conduct attribution analysis on geographical problems from a multi-disciplinary perspective according to the PC algorithm and the GNN inference algorithm, extract causal relationships, quantify causal strengths, and generate a visual causal map according to the causal relationships and causal strengths; the intelligent interaction unit is used to construct an immersive and personalized learning scenario according to the interaction methods of AI intelligent assistance, interdisciplinary narrative, and interdisciplinary decision-making, and enhance the fun of learning.

[0006] Optionally, the geographical teaching unit includes: a physical geography teaching module and a human geography teaching module. Among them, the physical geography teaching module is used to provide the physical geography knowledge of the system, reveal the structure, function, dynamics, and their interaction laws of the natural system on the earth's surface, and help students build a cognitive framework for the natural environment; the human geography teaching module is used to provide the human geography knowledge of the system, analyze the mutual relationship between human activities and the geographical environment, and reveal the spatial distribution and evolution laws of human elements.

[0007] Optionally, the multi-disciplinary knowledge collaboration module includes: a multi-disciplinary knowledge base module, a dynamic discipline routing controller module, and a knowledge fusion engine module. Among them, the multi-disciplinary knowledge base module is used to collect the knowledge of geography and other disciplines and construct a knowledge graph according to the knowledge; the dynamic discipline routing controller module is used to evaluate the association strength between the geographical discipline and the target discipline through the interdisciplinary attention mechanism combined with the dynamic gating mechanism, judge whether it can be associated with the target discipline. If it can be associated, plan the target path from the knowledge graph according to the A-Star algorithm, and locate the target discipline knowledge according to the target path; the knowledge fusion engine module is used to input the geographical knowledge and the target discipline knowledge into the trained deep learning model, and output a natural language text with the association and collaboration relationship between the geographical knowledge and the target discipline knowledge.

[0008] Optionally, the dynamic subject routing controller module includes: a feature extraction module, an interdisciplinary attention module, a dynamic gating decision module, and a path planner module. Among them, the feature extraction module is used to extract features from geographical knowledge and the target subject to obtain a geographical feature vector G and a target subject feature vector T = { }, when it is a single-subject association, n is 1; the interdisciplinary attention module is used to calculate the association weight between geographical knowledge points and the target subject based on the attention mechanism according to the geographical feature vector G and the target subject feature vector T; the dynamic gating decision module is used to calculate a gating signal according to the association weight and real-time data, and determine whether the association with the target subject can be activated according to the gating signal. If it cannot be activated, the user is prompted to change the subject for association, where the real-time data includes historical association success rate, resource availability, etc.; if the target subject is a single subject, it is judged whether the value of the gating signal is greater than a first target value. If the value of the gating signal is greater than the first target value, the association is activated, otherwise, the user is prompted to change the subject; if the target subject is a multi-subject, the subjects with gating signal values greater than the first target value are screened to activate the association; the path planner module is used to use the A-Star algorithm to obtain the target association path between geographical knowledge and target subject knowledge points, and obtain specific knowledge and resources according to the target association path.

[0009] Optionally, the multi-subject integrated deep mining module includes: a causal discovery unit, a causal reasoning unit, and a visualization unit. Among them, the causal discovery unit is used to use the PC algorithm to mine the direct causal relationship between geography and multi-subjects, and construct a causal graph skeleton. Among them, a complete undirected graph is used to initialize the causal graph skeleton, irrelevant edges are excluded through conditional independence testing, and finally the undirected edges are oriented to generate a directed acyclic graph DAG; the causal reasoning unit is used to supplement the indirect or non-linear causal relationships not captured by the PC algorithm. Among them, the DAG graph output by the PC algorithm is used as the initial graph, and the non-linear mapping of the causal relationship between geography and multi-subjects is learned through a message passing mechanism to obtain the causal relationship and causal strength; the visualization unit is used to use a graphical tool to convert the abstract causal logic into an intuitive graph, where the visualization elements include nodes and edges.

[0010] Optionally, the calculation formula for the association weight is:

[0011] Among them, is the association weight, G is the geographical feature vector, T is the target subject feature vector, is the scaling factor; The calculation formula for the gating signal is

[0012] Among them, i is the target subject index, is the attention correlation weight for the i-th target subject, is the historical correlation success rate for the i-th target subject, is the resource effectiveness for the i-th target subject, where W and b are learnable parameters, is the gating signal value for the i-th target subject, ∈[0,1]; The formula of the A-Star algorithm is

[0013]

[0014] where f(v) is the comprehensive evaluation value of the knowledge graph node v, g(v) is the sum of the path weights from the starting point of the knowledge graph to the node v, e is the edge of the knowledge graph, w(e) is the edge weight, and H(v) is the heuristic function.

[0015] Optionally, the formula of the message passing mechanism is:

[0016] where l represents the number of GNN layers, N(i) represents the neighbor set of node i, is the non-linear activation function, and are trainable parameter matrices, represents the adjacency matrix element, indicating whether there is a causal edge between

[0017] Optionally, the intelligent interaction module includes: an AI intelligent assistance unit, an interdisciplinary narrative unit, and an interdisciplinary decision-making unit. Among them, the AI intelligent assistance unit is used to answer students' interdisciplinary questions in real time and dynamically recommend associated learning resources; the interdisciplinary narrative unit is used to use neural radiance field technology to reconstruct historical geographical scenes based on the historical event timeline and geographical information system, and integrate multi-disciplinary narrative layers to improve students' interest and enthusiasm in learning geography; the interdisciplinary decision-making unit is used to model the causal relationship between events, generate an interactive dynamic sand table, and support learners to trigger corresponding result changes by dragging and adding or subtracting sand table nodes.

[0018] The second aspect of the present application provides a deep geographical learning method based on interdisciplinary knowledge collaboration, including the following steps: obtaining geographical knowledge, preliminary learning results, and learning questions; obtaining associated geographical knowledge based on the preliminary learning results, selecting a target discipline from multiple disciplines to associate with the geographical knowledge, calculating the association weight between the target discipline and the geographical knowledge, calculating a gating signal value based on the association weight, determining whether to activate the association according to the gating signal value, if the association is activated, planning a target path from the knowledge graph according to the A-Star algorithm, obtaining target discipline knowledge according to the target path, and inputting the target discipline knowledge and the geographical knowledge into a deep learning model to obtain a natural language text; performing in-depth mining according to the learning questions, using the PC algorithm and the GNN inference algorithm to perform attribution analysis on geographical problems, extracting causal relationships, quantifying causal strengths, and generating a visual causal map according to the causal relationships and causal strengths.

[0019] The third aspect of the present application provides an electronic device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor executes the program to implement the deep geographical learning method based on interdisciplinary knowledge collaboration as described in the above embodiments.

[0020] The beneficial effects obtained by the above-mentioned present invention are as follows: The embodiments of the present application help students establish a complete knowledge framework from physical geography to human geography by providing systematic geographical knowledge, and lay a solid foundation for the discipline; by constructing a multi-disciplinary knowledge graph and a dynamic discipline routing controller, it supports students to independently select associated disciplines according to their interests or needs, automatically obtain the associated paths and knowledge points of the selected disciplines, and generate natural language texts for multi-disciplinary collaboration in combination with a deep learning model, which is conducive to cultivating students' ability to actively explore interdisciplinary knowledge, breaking through the limitations of teacher-led associations in traditional teaching, and the optional disciplines are also beneficial for students with weak foundations to gradually cultivate interdisciplinary abilities in a progressive manner; using the PC algorithm and the GNN inference algorithm to mine the interdisciplinary causal logic of geographical phenomena and generate a visual causal map can intuitively display the causal network between geographical knowledge and other disciplines, helping students understand the deep mechanisms of complex human-earth relationships; by constructing a three-dimensional learning scenario through an intelligent interaction method, it can enhance the relevance and memory depth of knowledge. Thus, the problems of insufficient interdisciplinary integration, lack of autonomy, and shallow mining of geographical phenomena in the prior art are solved.

[0021] The additional aspects and advantages of the present application will be partially given in the following description, partially become obvious from the following description, or be understood through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The above and / or additional aspects and advantages of the present application will become apparent and be readily understood from the following description of embodiments in conjunction with the accompanying drawings, where: Figure 1 FIG. 3 is a schematic structural diagram of a deep geography learning system based on interdisciplinary knowledge collaboration according to an embodiment of the present application; Figure 2 FIG. 6 is a schematic diagram of a geography teaching unit according to an embodiment of the present application; Figure 3 FIG. 9 is a schematic diagram of a multi-disciplinary knowledge collaboration module according to an embodiment of the present application; Figure 4 FIG. 12 is a schematic diagram of a dynamic discipline routing controller according to an embodiment of the present application; Figure 5 FIG. 15 is a schematic diagram of a multi-disciplinary integrated deep mining module according to an embodiment of the present application; Figure 6 FIG. 18 is a visual causal diagram of the urban heat island effect according to an embodiment of the present application; Figure 7 FIG. 21 is a schematic diagram of an intelligent interaction unit according to an embodiment of the present application; Figure 8 FIG. 24 is a flowchart of a deep geography learning method based on interdisciplinary knowledge collaboration according to an embodiment of the present application; Figure 9 FIG. 27 is a flowchart of an interdisciplinary knowledge collaboration method according to an embodiment of the present application.

[0023] Figure 10 FIG. 31 is a schematic structural diagram of an electronic device according to an embodiment of the present application. Detailed Embodiments

[0024] Embodiments of the present application will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to explain the present application and should not be construed as limiting the present application.

[0025] The following describes the deep geographical learning system based on interdisciplinary knowledge collaboration according to the embodiments of the present application. In view of the problems in the prior art mentioned in the above background art, such as insufficient interdisciplinary integration in geographical education, lack of autonomy, and shallow mining of geographical phenomena, the present application provides a deep geographical learning system based on interdisciplinary knowledge collaboration. In this system, systematic geographical knowledge is provided to help students establish a complete knowledge framework from physical geography to human geography and consolidate the subject foundation; by constructing a multi-disciplinary knowledge graph and a dynamic subject routing controller, students are supported to independently select related subjects according to their interests or needs, automatically obtain the associated paths and knowledge points of the selected subjects, and generate natural language texts with multi-disciplinary collaboration in combination with a deep learning model, which is conducive to cultivating students' ability to actively explore interdisciplinary knowledge, breaking through the limitations of teacher-led associations in traditional teaching. The optional subjects also help students with weak foundations gradually cultivate interdisciplinary capabilities in a progressive manner; the PC algorithm and the GNN inference algorithm are used to mine the interdisciplinary causal logic of geographical phenomena and generate a visual causal diagram, which can intuitively display the causal network between geographical knowledge and other disciplines and help students understand the deep mechanisms of complex human-earth relationships; by means of intelligent interaction, a three-dimensional learning scenario is constructed, which can enhance the knowledge relevance and memory depth. Thus, the problems in the prior art, such as insufficient interdisciplinary integration in geographical education, lack of autonomy, and shallow mining of geographical phenomena, are solved.

[0026] Specifically, Figure 1 is a schematic diagram of the composition of the deep geographical learning system based on interdisciplinary knowledge collaboration provided by the embodiments of the present application.

[0027] As Figure 1 shown, the deep geographical learning system 10 based on interdisciplinary knowledge collaboration includes: A geographical teaching unit 100, a multi-disciplinary knowledge collaboration module 200, a multi-disciplinary integrated deep mining module 300, and an intelligent interaction unit 400.

[0028] Among them, the geography teaching unit 100 is used to provide systematic geographical knowledge to help students build the foundation of the geography discipline; the multi-disciplinary knowledge collaboration module 200 is used to build a dynamic subject routing controller according to the subject dynamic matching mechanism, and combined with the deep learning model, so that students can select one or more subjects from multiple subjects to be associated with the geography discipline, and enhance students' understanding of geographical knowledge from the perspective of the selected target subject. Among them, the multiple subjects include biology, physics, history, agriculture, economy, and human sociology; the multi-disciplinary integrated in-depth mining module 300 is used to conduct attribution analysis on geographical problems from the perspective of multiple disciplines according to the PC algorithm and the GNN reasoning algorithm, extract causal relationships, quantify causal intensities, and generate a visual causal diagram according to the causal relationships and causal intensities; the intelligent interaction unit 400 is used to build an immersive and personalized learning scenario according to the interaction methods of AI intelligent assistance, interdisciplinary narrative, and interdisciplinary decision-making, and enhance the fun of learning.

[0029] It can be understood that the systematic knowledge framework constructed by the geography teaching unit in the embodiment of the present application ensures that students have a solid disciplinary foundation and avoids the problem of knowledge fragmentation that may occur in interdisciplinary learning. Through the dynamic subject routing mechanism introduced by the multi-disciplinary knowledge collaboration module, the intelligent matching of geography and knowledge in six major disciplinary fields is realized. Through the dynamic causal diagram generated by the multi-disciplinary integrated in-depth mining module using cutting-edge algorithms, the complex relationships that are difficult to present in traditional teaching are visualized, which can significantly improve students' systematic thinking ability. Through intelligent interaction, the personalization and fun of geography learning are enhanced. Overall, it is comprehensively upgraded from knowledge construction, interdisciplinary collaboration, in-depth analysis to interaction experience, effectively improving students' core geography literacy and complex problem-solving ability.

[0030] In the embodiment of the present application, the geography teaching unit 100 includes: as Figure 2 shown, the physical geography teaching module and the human geography teaching module.

[0031] Among them, the physical geography teaching module is used to provide systematic physical geography knowledge, reveal the structure, function, dynamics and their interaction laws of the natural system on the earth's surface, and help students build a cognitive framework of the natural environment; the human geography teaching module is used to provide systematic human geography knowledge, analyze the interaction between human activities and the geographical environment, and reveal the spatial distribution and evolution laws of human elements.

[0032] Specifically, physical geography mainly studies the formation, evolution and laws of the earth's natural environment, including the earth's structure, landform evolution, climate formation, hydrological cycle, ecosystem flow, etc. Human geography mainly studies the impact of human elements on the earth, and human elements include population migration, urban layout, economy, culture, etc.

[0033] It is understandable that the embodiments of the present application provide natural geography teaching to cultivate students' rational cognition and exploration ability of the natural world, provide human geography teaching for students to learn the impact of human activities on the geographical environment, and guide students to pay attention to the geographical logic and practical problems of human activities. Through the linkage of natural geography teaching and human geography teaching, a complete learning chain from understanding nature to understanding humanities and then to coordinating the relationship between humans and the environment is jointly constructed.

[0034] In the embodiments of the present application, the multi-disciplinary knowledge collaboration module 200 includes: as Figure 3 shown, a multi-disciplinary knowledge base module, a dynamic subject routing controller module, and a knowledge fusion engine module.

[0035] Among them, the multi-disciplinary knowledge base module is used to collect knowledge of geography and other disciplines and construct a knowledge graph according to the knowledge; the dynamic subject routing controller module is used to evaluate the association strength between the geography discipline and the target discipline through an interdisciplinary attention mechanism combined with a dynamic gating mechanism, determine whether it can be associated with the target discipline, and if it can be associated, plan a target path from the knowledge graph according to the A-Star algorithm, and locate the target discipline knowledge according to the target path; the knowledge fusion engine module is used to input geographical knowledge and target discipline knowledge into a trained deep learning model and output a natural language text with the association and collaboration relationship between geographical knowledge and target discipline knowledge.

[0036] Among them, the knowledge graph G = (V, E), where V is a set of nodes representing target discipline knowledge points, and each node v ∈ V includes the following attributes: semantic embedding vector (v.embedding), discipline label to which it belongs (v.discipline), and basic cognitive cost (v.cost); E is a set of edges representing the relationships between knowledge points, and each edge e ∈ E includes the following attributes: initial relationship weight (e.weight), relationship type (e.type), where the relationship type includes causality, inclusion, dependence, etc.; Specifically, when studying the North Atlantic Current, a student selected physics, biology, and history as target disciplines for association. After the dynamic discipline routing controller evaluated the association strengths of physics, biology, and history disciplines respectively, it determined that all three disciplines could be associated. Using the A-Star algorithm, it planned the path "Cause of the Warm Current (Physics) → Influence of the Warm Current on Climate (Geography) → Abundant Fishery Resources Brought by the Warm Current (Biology) → Historical Process (History)" and located the corresponding disciplinary knowledge. The obtained knowledge was input into the deep learning model, and the integrated natural language text was output: "The North Atlantic Current is formed by the confluence of the Gulf Stream and the cold water flowing south from the Arctic Ocean. Its powerful warm water transportation makes the average annual temperature along the western coast of Europe about 10 °C higher than that of other regions at the same latitude, creating ice-free ports along the Norwegian fjords in winter and promoting the prosperity of shipping and fishery economies in Northern Europe. The warm and humid air currents brought by the warm current also contribute to the formation of the temperate maritime climate in Western Europe, which is suitable for the growth of juicy forage grasses and supports the development of the dairy industry. At the same time, the North Sea Fishing Ground formed by its confluence with the southward cold current is rich in salmon and cod resources and has become an important fishery base in Europe. The Hanseatic League monopolized the North Sea trade network by controlling the fishing ground, and cities such as Lübeck and Hamburg prospered due to cod processing and export, even affecting the political pattern of countries along the Baltic Sea." It can be understood that the embodiment of this application constructs a "Geography +" knowledge network, breaks the barriers of single disciplines, and based on the dynamic routing technology of the attention mechanism and the A-Star algorithm, can accurately locate high-value disciplinary association paths, integrate multidisciplinary knowledge through the deep learning model, generate natural language, and make complex knowledge more vivid and easy to understand. This solution enables geographical education to shift from one-way teaching to intelligent interaction, from disciplinary fragmentation to multidisciplinary integration, realizes the structured integration and automated association of multi-dimensional knowledge, and breaks through the bottleneck of fragmented interdisciplinary knowledge and low-efficiency association in traditional teaching.

[0037] In the embodiment of this application, the dynamic discipline routing controller module includes: as Figure 4 shown, a feature extraction module, an interdisciplinary attention module, a dynamic gating decision module, and a path planner module.

[0038] Among them, the feature extraction module is used to extract features from geographical knowledge and target disciplines, obtaining a geographical feature vector G and a target discipline feature vector T = { }, when it is a single-discipline association, n is 1; the cross-disciplinary attention module is used to calculate the association weight between geographical knowledge points and the target discipline based on the attention mechanism according to the geographical feature vector G and the target discipline feature vector T; the dynamic gating decision module is used to calculate the gating signal according to the association weight and real-time data, and judge whether the association with the target discipline can be activated according to the gating signal. If it cannot be activated, the user is prompted to change the discipline for association. Among them, the real-time data includes historical association success rate, resource effectiveness, etc.; if the target discipline is a single discipline, it is judged whether the gating signal value is greater than the first target value. If the gating signal value is greater than the first target value, the association is activated; otherwise, the user is prompted to change the discipline; if the target discipline is a multi-discipline, the disciplines with gating signal values greater than the first target value are screened to activate the association. Among them, in the embodiment of the present application, the first target value is 0.6; the path planner module is used to obtain the target association path between geographical knowledge and target discipline knowledge points using the A-Star algorithm, and obtain specific knowledge and resources according to the target association path.

[0039] Among them, the formula for calculating the association weight is:

[0040] Among them, is the association weight, G is the geographical feature vector, T is the target discipline feature vector, is the scaling factor; The formula for calculating the gating signal is

[0041] Among them, i is the target discipline index, is the attention association weight of the i-th target discipline, is the historical association success rate of the i-th target discipline, is the resource effectiveness of the i-th target discipline, W and b are learnable parameters, is the gating signal value of the i-th target discipline, ∈[0,1]; The A-Star algorithm formula is

[0042]

[0043] Among them, f(v) is the comprehensive evaluation value of the knowledge graph node v, g(v) is the sum of the path weights from the starting point of the knowledge graph to the node v, e is the edge of the knowledge graph, w(e) is the edge weight, and H(v) is the heuristic function.

[0044] Specifically, when studying the Mediterranean climate, a student selected Biology and History as target disciplines for association, extracted the characteristics of the Mediterranean climate and the target disciplines, and obtained the geographical feature vector G = {"Climate Type": "Mediterranean climate", "Main Characteristics": ["Hot and dry in summer", "Mild and rainy in winter"], "Geographical Distribution": ["West coast of the continents between 30-40 degrees north and south latitudes"], "Typical Regions": ["Coast of the Mediterranean Sea", "California", "Central Chile"], "Forming Reasons": ["Seasonal movement of the subtropical high pressure", "Influence of the westerly belt"]}, and the target discipline feature vector T = [{"Discipline": "Biology", "Feature Dimensions": ["Plant Adaptability", "Animal Behavior", "Ecosystem"], "Keywords": ["Drought-tolerant plants", "Hard-leaved forests", "Fire adaptation"]}, {"Discipline": "History", "Feature Dimensions": ["Civilization Development", "Agricultural History", "Trade Routes"], "Keywords": ["Ancient Greek civilization", "Roman Empire", "Olive cultivation"]}]. The association weights of the Biology discipline and the History discipline are calculated respectively through the attention mechanism. Among them, the association weight of the Biology discipline is 0.85, and the association weight of the History discipline is 0.72. The gating signal values of them are calculated respectively according to the association weights of the Biology discipline and the History discipline. Among them, the gating signal value of the Biology discipline is 0.76, and the gating signal value of the History discipline is 0.68. Since the gating signal values of both are greater than the first target value of 0.6, the association between Biology and History disciplines is activated simultaneously. The A-Star algorithm is used to plan the target path of "Mediterranean climate distribution → Duration of summer drought → Hard-leaved plant community → Suitable crops (olives, grapes) → Olive oil trade → City-state economic development → Development of maritime trade → Influence of climate on the rise and fall of civilizations" in the knowledge graph, and the knowledge points corresponding to the Biology and History disciplines are obtained according to the target path.

[0045] It can be understood that in the embodiment of the present application, by converting geographical knowledge points and multiple disciplines into feature vectors, it helps students extract core elements from fragmented knowledge and reduces the cognitive load of interdisciplinary learning. The association weights between geography and target disciplines are calculated through the attention mechanism, and the dynamic gating decision module is combined to only activate highly certain associations to avoid cognitive overload caused by complex associations. For associations that do not meet the standards, guidance is provided to let students actively explore other possible associations, changing the traditional teaching mode of "teacher-led association" to an inquiry-based learning mode of "students' active trial and error - system intelligent guidance", which stimulates interdisciplinary curiosity. By planning the target path, a complete knowledge chain is automatically constructed to prevent students from getting stuck in irrelevant details, realizing efficient cognitive transfer across disciplines.

[0046] In the embodiment of the present application, the multi-disciplinary integrated in-depth mining module 300 includes: as Figure 5 shown, a causal discovery unit, a causal reasoning unit, and a visualization unit.

[0047] Among them, the causal discovery unit is used to mine the direct causal relationship between geography and multiple disciplines using the PC algorithm and construct the causal graph skeleton. Among them, the causal graph skeleton is initialized using a complete undirected graph, and the conditional independence test is used to judge whether the variable and are independent when given the set S. If there exists an S such that holds, then it is determined that is an irrelevant edge. The irrelevant edges are excluded in turn, and finally the undirected edges are oriented to generate a directed acyclic graph (DAG); the causal reasoning unit is used to supplement the indirect or non-linear causal relationships not captured by the PC algorithm. Among them, the DAG graph output by the PC algorithm is used as the initial graph, and the non-linear mapping of the causal relationship between geography and multiple disciplines is learned through the message passing mechanism to obtain the causal relationship and causal strength; the visualization unit is used to convert the abstract causal logic into an intuitive graph using a graphical tool, where the visualization elements include nodes and edges.

[0048] Among them, the formula of the message passing mechanism is:

[0049] Among them, l represents the number of GNN layers, N(i) represents the neighbor set of node i, is a non-linear activation function, and are trainable parameter matrices, represents the adjacency matrix element, indicating whether there is a causal edge between

[0050] Specifically, the visualization elements include nodes and edges. Among them, the edges include solid edges, dashed edges, and arrow directions. Among them, the solid edges represent the causal relationships determined by the PC algorithm, the dashed edges represent the potential causal relationships inferred by the GNN, and the arrow direction represents the causal flow direction.

[0051] Specifically, when deeply mining the urban heat island effect, according to the causal discovery algorithm, first assume that all discipline variables may affect each other, that is, construct a fully connected undirected graph, and then judge whether two variables are independent through the conditional independence test. For example, the two variables are urban building density and energy consumption. After giving "population density", the correlation between them disappears, then it is judged as an irrelevant edge and the irrelevant edge is deleted. Finally, the undirected edges are oriented to generate the DAG skeleton. Then, the non-linear mapping relationship is discovered through the GNN algorithm to supplement the indirect causal relationship. For example, there is a threshold for the inhibitory effect of vegetation coverage on the heat island effect. After training, the causal strength of each edge is output. Finally, as Figure 6 shown, the visual causal graph is output according to the causal relationship and causal strength.

[0052] It is understandable that the embodiment of the present application captures direct causal relationships through the PC algorithm, screens causal edges through conditional independence tests, ensures the rigor of direct causal relationships between geography and multiple disciplines, eliminates false associations, and constructs a reliable causal graph skeleton. In view of the indirect causal relationships and nonlinear relationships that are difficult to capture with the PC algorithm, the complex mapping of interdisciplinary variables is learned through the message passing mechanism, and the causal analysis is advanced from surface associations to mechanistic explanations, thus improving the causal closed loop. The GNN reasoning algorithm generates a causal strength value for each edge, clarifies the degree of influence of geography and multidisciplinary variables, avoids the ambiguity of traditional qualitative analysis, and transforms abstract causal relationships into visual causal graphs, realizing the quantification and interpretability of causal logic, which is conducive to students' deepening understanding of geographical laws.

[0053] In the embodiment of the present application, the intelligent interaction unit 400 includes: Figure 7 As shown, there are AI intelligent assistance unit, interdisciplinary narrative unit, and interdisciplinary decision-making unit.

[0054] Among them, the AI ​​intelligent assistance unit is used to answer students' interdisciplinary questions in real time and dynamically recommend related learning resources; the interdisciplinary narrative unit is used to use neural radiation field technology to reconstruct historical and geographical scenes based on the timeline of historical events and geographic information systems, integrate multidisciplinary narrative layers, and enhance students' interest and enthusiasm in learning geography; the interdisciplinary decision-making unit is used to model the causal relationship between events and generate an interactive dynamic sandbox, supporting learners to trigger corresponding result changes by dragging and adding and subtracting sandbox nodes.

[0055] Specifically, when studying the impact of the Yellow River's diversion on the civilization of the North China Plain, the interdisciplinary narrative function uses neural radiation field technology combined with geographic information systems and historical information to generate the Yellow River channel scene in 1128, superimposing the historical event marking layer, economic data flow layer, ecological environment evolution layer, etc. in the North China Plain area. Students can slide the timeline to view the changes in the river channel and the rise and fall of the city.

[0056] It is understandable that the embodiments of the present application inject intelligent, immersive, and interactive innovative kinetic energy into geography education through intelligent interaction. AI intelligent assistance can answer students' questions in real time, help students connect fragmented knowledge, and push relevant subject content in a targeted manner according to the content of students' questions and learning trajectories to meet personalized learning needs. Reconstructing historical geographical scenes with the help of neural radiation field technology and GIS can create an immersive and scenario-based learning experience for students, improve memory retention and learning interest. Interdisciplinary decision-making can cultivate high-level decision-making capabilities. Using dynamic sandboxes to model causal relationships and support interactive experiments of dragging and dropping, adding and subtracting nodes can cultivate students' thinking ability from hypothesis to verification to decision-making, and systematically improve teaching effectiveness and students' core literacy.

[0057] The deep geographical learning system based on interdisciplinary knowledge collaboration proposed according to the embodiments of the present application helps students establish a complete knowledge framework from physical geography to human geography by providing systematic geographical knowledge, thereby consolidating the subject foundation. By constructing multi-disciplinary knowledge graphs and dynamic subject routing controllers, it supports students to independently select related disciplines according to their interests or needs, automatically obtain the associated paths and knowledge points of the selected disciplines, and generate natural language texts with multi-disciplinary collaboration in combination with deep learning models. This is conducive to cultivating students' ability to actively explore interdisciplinary knowledge, breaking through the limitations of teacher-led associations in traditional teaching. The optional disciplines also help students with weak foundations gradually cultivate interdisciplinary abilities in a progressive manner. Using the PC algorithm and the GNN inference algorithm to mine the interdisciplinary causal logic of geographical phenomena and generate visual causal diagrams can intuitively display the causal network between geographical knowledge and other disciplines, helping students understand the deep mechanisms of complex human-earth relationships. By constructing a three-dimensional learning scenario through an intelligent interaction method, it can enhance the relevance and depth of knowledge memory. Thus, the problems of insufficient interdisciplinary integration, lack of autonomy, and shallow mining of geographical phenomena in the prior art are solved.

[0058] Next, the deep geographical learning method based on interdisciplinary knowledge collaboration proposed according to the embodiments of the present application will be described with reference to the accompanying drawings.

[0059] Specifically, Figure 8 is a schematic flowchart of the deep geographical learning method based on interdisciplinary knowledge collaboration provided by the embodiments of the present application.

[0060] As Figure 8 shown, the deep geographical learning method based on interdisciplinary knowledge collaboration includes the following steps: In step S101, geographical knowledge, preliminary learning results, and learning questions are obtained.

[0061] Among them, the geographical knowledge includes physical geography and human geography.

[0062] It can be understood that by obtaining physical geography and human geography knowledge in the embodiments of the present application, it is conducive to building a complete geographical learning framework, consolidating the geographical learning foundation, having a preliminary understanding of geographical knowledge, and providing a basis for deeper interdisciplinary learning.

[0063] In step S102, as Figure 9As shown, relevant geographical knowledge is obtained based on the preliminary learning results. The target subject is selected from multiple disciplines and associated with the geographical knowledge, and the association weight between the target subject and the geographical knowledge is calculated. The gating signal value is calculated based on the association weight, and it is determined whether the association is activated according to the gating signal value. If the association is activated, the target path is planned from the knowledge graph according to the A-Star algorithm, the target subject knowledge is obtained according to the target path, and the target subject knowledge and the geographical knowledge are input into the deep learning model to obtain the natural language text.

[0064] Specifically, determining whether the association is activated according to the gating signal value includes: if the target subject is a single subject, determining whether the gating signal value is greater than the first target value. If the gating signal value is greater than the first target value, the association is activated; otherwise, the user is prompted to change the subject. If the target subject is a multi-subject, the subjects with the gating signal value greater than the first target value are screened to activate the association. If none of the target subjects are activated, the user is prompted to change the subject.

[0065] It can be understood that in the embodiment of the present application, by building a knowledge graph and a dynamic routing screening mechanism, a cross-disciplinary navigation system for geographical knowledge is realized, and the precision of associated learning is achieved. The gating signal value quantifies the association quality, automatically filters out low-value associations, ensures that students focus on high-value paths, and improves the learning efficiency of students. Taking the knowledge graph as the map, the dynamic routing as the navigation, and the deep learning as the translator, it helps students accurately locate high-value paths in complex subject associations and generate logical and in-depth cross-disciplinary explanations. Allowing students to independently select subjects gives students the decision-making power for cross-disciplinary exploration, which is conducive to stimulating students' initiative in learning and cultivating their decision-making thinking.

[0066] In step S103, in-depth mining is performed based on the learning questions, and the PC algorithm and the GNN inference algorithm are used to conduct attribution analysis on geographical problems from multiple disciplinary perspectives, extract causal relationships, quantify the causal intensity, and generate a visual causal graph according to the causal relationships and the causal intensity.

[0067] It can be understood that in the embodiment of the present application, the PC algorithm is used to mine direct causality, combined with the GNN inference algorithm to mine indirect causality and non-linear associations, deeply supplement the causal relationships, clarify the influence degree of each factor on geographical problems according to the quantified causal intensity, and present it to students in a visual way, which is conducive to students' in-depth understanding of geographical laws and more conducive to cultivating their ability to analyze real problems using the "geography +" thinking.

[0068] The deep geographical learning method based on interdisciplinary knowledge collaboration proposed in the embodiments of the present application helps students establish a complete knowledge framework from physical geography to human geography by providing systematic geographical knowledge, thus consolidating the subject foundation. By constructing a multi-disciplinary knowledge graph and a dynamic subject routing controller, it supports students to independently select associated disciplines according to their interests or needs, automatically obtain the associated paths and knowledge points of the selected disciplines, and generate natural language texts with multi-disciplinary collaboration in combination with deep learning models, which is conducive to cultivating students' ability to actively explore interdisciplinary knowledge, breaking through the limitations of teacher-led associations in traditional teaching. The optional disciplines also help students with weak foundations gradually cultivate interdisciplinary abilities in a progressive manner. Using the PC algorithm and the GNN inference algorithm to mine the interdisciplinary causal logic of geographical phenomena and generate visual causal diagrams can intuitively display the causal network between geographical knowledge and other disciplines, helping students understand the deep mechanisms of complex human-earth relationships. Thus, the problems of insufficient interdisciplinary integration, lack of autonomy, and shallow mining of geographical phenomena in the prior art are solved.

[0069] The following will specifically elaborate on the deep geographical learning method based on interdisciplinary knowledge collaboration, and the content is as follows: S1: Obtain geographical knowledge, preliminary learning results, and learning questions.

[0070] For example, when student A is learning the phenomenon of counter-urbanization in the process of urbanization, they learn the definition, characteristics, and possible causes of counter-urbanization, and at the same time have questions, such as "Why is it more likely to have counter-urbanization in economically developed regions?" and "How do economic factors affect counter-urbanization?" etc.

[0071] S2: Obtain associated geographical knowledge based on the preliminary learning results, select target disciplines from multiple disciplines to be associated with the geographical knowledge, calculate the association weight between the target discipline and the geographical knowledge, calculate the gating signal value according to the association weight, determine whether to activate the association according to the gating signal value. If the association is activated, plan the target path from the knowledge graph according to the A-Star algorithm, obtain the target discipline knowledge according to the target path, and input the target discipline knowledge and the geographical knowledge into the deep learning model to obtain a natural language text.

[0072] For example, student A selects three subjects, namely history, economics, and humanities and sociology, to correlate with the phenomenon of counter-urbanization. The correlation weights and gating signal values of these three subjects are calculated respectively. The gating signal value of the history subject is 0.42, the gating signal value of the economics subject is 0.78, and the gating signal value of the humanities and sociology subject is 0.71. Since the gating signal value of the history subject is less than the first target value, only the economics and humanities and sociology subjects are activated. Through the A-Star algorithm, the target path of "rising rent in the city center (economics) → middle class pursuing living space (humanities), enterprises relocating outside the city (economics) → suburban construction (economics) → increased employment opportunities in the suburbs (economics) → population migration (geography)" is planned from the knowledge graph. According to the target path, the corresponding knowledge points are obtained, and the corresponding knowledge points and the phenomenon of counter-urbanization are input into the deep learning model, and the integrated natural language text is output: In the phenomenon of counter-urbanization in the process of urbanization, the economic factor of rising rent in the city center becomes the core driving force. On the one hand, it forces enterprises to relocate to the suburbs due to the rising operating costs, driving suburban construction and increasing local employment opportunities; on the other hand, it prompts the middle class to migrate to the suburbs in pursuit of a more comfortable living space. The dual effects of enterprise relocation and population migration, through the positive cycle of enhanced suburban economic vitality and improved public facilities, promote the diffusion of the population distribution from the city center to the suburbs.

[0073] S3: Conduct in-depth mining based on learning questions, use the PC algorithm and the GNN inference algorithm to conduct attribution analysis on geographical problems from multiple disciplinary perspectives, extract causal relationships, quantify causal intensities, and generate a visual causal graph according to the causal relationships and causal intensities.

[0074] For example, according to the learning question of student A, "Why is counter-urbanization more likely to occur in economically developed areas", the causal graph skeleton is constructed according to the PC algorithm. Weakly associated edges are removed through conditional independence tests, and strongly associated edges are retained to generate a DAG skeleton, such as economically developed → rising rent, agglomeration of middle and high-income groups → higher costs, lower living standards → industrial and population migration. The CNN algorithm is combined to supplement indirect and non-linear causal relationships and calculate causal intensities, among which indirect causal relationships. For example, economically developed → increased environmental pressure → shift of living preferences to the suburbs. Finally, a visual causal graph is drawn according to the causal relationships and causal intensities.

[0075] In summary, through the construction of a dynamic knowledge graph and a disciplinary routing system in the embodiments of the present application, complex geographical phenomena such as counter-urbanization are organically connected with multi-disciplinary knowledge such as economics and sociology. Based on the visual causal analysis of the PC algorithm and GNN inference, learners can more easily understand cross-dimensional action mechanisms such as "industrial upgrading driving up land prices → middle class suburban migration".

[0076] Figure 10 It is a schematic structural diagram of the electronic device provided by the embodiments of the present application. The electronic device may include: A memory 1001, a processor 1002, and a computer program stored on the memory 1001 and executable on the processor 1002.

[0077] When the processor 1002 executes the program, it implements the deep geographical learning method based on interdisciplinary knowledge collaboration provided in the above embodiments.

[0078] Furthermore, the electronic device further includes: A communication interface 1003 for communication between the memory 1001 and the processor 1002.

[0079] The memory 1001 is used to store a computer program executable on the processor 1002.

[0080] The memory 1001 may include high-speed RAM memory and may also include non-volatile memory, such as at least one disk memory.

[0081] If the memory 1001, the processor 1002, and the communication interface 1003 are implemented independently, the communication interface 1003, the memory 1001, and the processor 1002 can be interconnected through a bus to complete mutual communication. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the sake of representation, Figure 10 only a thick line is used to represent it in the figure, but it does not mean that there is only one bus or one type of bus.

[0082] Optionally, in a specific implementation, if the memory 1001, the processor 1002, and the communication interface 1003 are integrated on a chip, the memory 1001, the processor 1002, and the communication interface 1003 can complete mutual communication through an internal interface.

[0083] The processor 1002 may be a Central Processing Unit (CPU), or an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0084] In the description of this specification, the descriptions referring to terms such as "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of this application. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0085] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of such features. In the description of this application, the meaning of "N" is at least two, such as two, three, etc., unless otherwise specifically defined.

[0086] Any process or method description shown in the flowchart or described in other ways herein can be understood to represent a module, segment, or part of code including one or more N executable instructions for implementing a customized logical function or process, and the scope of the preferred embodiments of this application includes additional implementations, where the functions can be executed in a substantially simultaneous manner or in a reverse order according to the involved functions, rather than in the order shown or discussed, which should be understood by those skilled in the technical field to which the embodiments of this application belong.

[0087] It should be understood that each part of this application can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware as in another embodiment, any one or a combination of the following technologies well known in the art can be used: discrete logic circuits with logic gate circuits for implementing logical functions on data signals, application specific integrated circuits with appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0088] Those of ordinary skill in the technical field of this application can understand that all or part of the steps carried by the methods of the above embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

Claims

1. A deep geographical learning system based on interdisciplinary knowledge collaboration, characterized in that, Including: A geography teaching unit, a multi-disciplinary knowledge collaboration module, a multi-disciplinary integrated in-depth mining module, and an intelligent interaction unit. Among them, The geography teaching unit is used to provide systematic geography knowledge to help students build a foundation in the geography discipline; The multi-disciplinary knowledge collaboration module is used to build a dynamic discipline routing controller according to the discipline dynamic matching mechanism, and combined with a deep learning model, enabling students to select one or more disciplines from multiple disciplines to be associated with the geography discipline, and enhancing students' understanding of geography knowledge from the perspective of the selected target discipline. Among them, the multiple disciplines include biology, physics, history, agriculture, economy, and human sociology; The multi-disciplinary integrated in-depth mining module is used to conduct attribution analysis on geography problems from a multi-disciplinary perspective according to the PC algorithm and the GNN inference algorithm, extract causal relationships, quantify causal strengths, and generate a visual causal map according to the causal relationships and causal strengths; The intelligent interaction unit is used to build an immersive and personalized learning scenario according to the interaction methods of AI intelligent assistance, interdisciplinary narration, and interdisciplinary decision-making, enhancing the fun of learning.

2. The deep geographical learning system based on interdisciplinary knowledge collaboration according to claim 1, characterized in that The multi-disciplinary knowledge collaboration module includes: a multi-disciplinary knowledge base module, a dynamic discipline routing controller module, and a knowledge fusion engine module. Among them, The multi-disciplinary knowledge base module is used to collect knowledge of geography and other disciplines and build a knowledge graph according to the knowledge; The dynamic discipline routing controller module is used to evaluate the association strength between the geography discipline and the target discipline through an interdisciplinary attention mechanism combined with a dynamic gating mechanism, judge whether it is associated with the target discipline. If it is associated, then plan a target path from the knowledge graph according to the A-Star algorithm, and locate the target discipline knowledge according to the target path; The knowledge fusion engine module is used to input geography knowledge and target discipline knowledge into a trained deep learning model and output a natural language text with the association and collaboration relationship between geography knowledge and target discipline knowledge.

3. The deep geographical learning system based on interdisciplinary knowledge collaboration according to claim 2, wherein The dynamic discipline routing controller module includes: a feature extraction module, an interdisciplinary attention module, a dynamic gating decision module, and a path planner module. Among them, The feature extraction module is used to extract features from geographical knowledge and the target subject, obtaining a geographical feature vector G and a target subject feature vector T = { }, where n is 1 when it is a single-subject association; The interdisciplinary attention module is used to calculate the association weight between geography knowledge points and the target discipline based on the attention mechanism according to the geography feature vector G and the target discipline feature vector T; The dynamic gating decision module is used to calculate a gating signal according to the association weight and real-time data, and judge whether the association with the target discipline can be activated according to the gating signal. If it cannot be activated, it prompts the user to change the discipline for association. Among them, the real-time data includes historical association success rates, resource effectiveness, etc.; if the target discipline is a single discipline, it judges whether the gating signal value is greater than a first target value. If the gating signal value is greater than the first target value, the association is activated, otherwise, it prompts the user to change the discipline; if the target discipline is multiple disciplines, it screens the disciplines with gating signal values greater than the first target value to activate the association; The path planner module is used to obtain the target association path between geography knowledge and target discipline knowledge points using the A-Star algorithm, and obtain specific knowledge and resources according to the target association path.

4. The deep geographical learning system based on interdisciplinary knowledge collaboration according to claim 1, characterized in that The multi-disciplinary integrated in-depth mining module includes: a causal discovery unit, a causal reasoning unit, and a visualization unit. Among them, the causal discovery unit is used to use the PC algorithm to mine the direct causal relationship between geography and multiple disciplines, construct the skeleton of the causal graph. Among them, the complete undirected graph is used to initialize the skeleton of the causal graph, irrelevant edges are excluded through conditional independence testing, and finally the undirected edges are oriented to generate a directed acyclic graph (DAG); the causal reasoning unit is used to supplement the indirect or non-linear causal relationships not captured by the PC algorithm. Among them, the DAG graph output by the PC algorithm is used as the initial graph, and the non-linear mapping of the causal relationship between geography and multiple disciplines is learned through the message passing mechanism to obtain the causal relationship and causal strength; the visualization unit is used to use graphical tools to convert abstract causal logic into intuitive graphics, where the visualization elements include nodes and edges.

5. The in-depth geography learning system based on interdisciplinary knowledge collaboration according to claim 3, characterized in that, The calculation formula for the association weight is: ； Among them, is the correlation weight, G is the geographical feature vector, and T is the target subject feature vector. is the scaling factor; The calculation formula for the gating signal is ； where \(i\) is the index of the target subject, is the attention correlation weight of the \(i\)-th target subject, is the historical correlation success rate of the \(i\)-th target subject, is the resource effectiveness of the \(i\)-th target subject, \(W\) and \(b\) are learnable parameters, is the gating signal value of the \(i\)-th target subject, \(\in[0,1]\); The formula for the A-Star algorithm is ； ； Among them, f(v) is the comprehensive evaluation value of the knowledge graph node v, g(v) is the sum of the path weights from the starting point of the knowledge graph to node v, e is the edge of the knowledge graph, w(e) is the edge weight, and H(v) is the heuristic function.

6. The in-depth geography learning system based on interdisciplinary knowledge collaboration according to claim 4, wherein The formula for the message passing mechanism is: ； where \(l\) represents the number of GNN layers, and \(N(i)\) represents the neighbor set of node \(i\). is a non-linear activation function. and are trainable parameter matrices. represents the elements of the adjacency matrix, indicating whether there is a causal edge between 7. The in-depth geography learning system based on interdisciplinary knowledge collaboration according to claim 1, characterized in that The intelligent interaction module includes: an AI intelligent assistance unit, an interdisciplinary narrative unit, and an interdisciplinary decision-making unit. Among them, the AI intelligent assistance unit is used to answer students' interdisciplinary questions in real time and dynamically recommend associated learning resources; the interdisciplinary narrative unit is used to use the neural radiance field technology to reconstruct the historical geographical scene based on the historical event timeline and geographic information system, and integrate the interdisciplinary narrative layer to improve students' interest and enthusiasm in learning geography; the interdisciplinary decision-making unit is used to model the causal relationship between events, generate an interactive dynamic sand table, and support learners to trigger corresponding result changes by dragging, adding, or deleting sand table nodes.

8. The in-depth geography learning system based on interdisciplinary knowledge collaboration according to claim 1, characterized in that The geography teaching unit includes: a physical geography teaching module and a human geography teaching module. Among them, the physical geography teaching module is used to provide systematic physical geography knowledge, reveal the structure, function, dynamics and their interaction laws of the natural system on the earth's surface, and help students build a cognitive framework of the natural environment; the human geography teaching module is used to provide systematic human geography knowledge, analyze the mutual relationship between human activities and the geographical environment, and reveal the spatial distribution and evolution laws of human elements.

9. A deep geographical learning method based on interdisciplinary knowledge collaboration, characterized in that, It includes the following steps: Obtain geographical knowledge, preliminary learning results, and learning questions; Obtain the associated geographical knowledge according to the preliminary learning results, select the target discipline from multiple disciplines to associate with the geographical knowledge, calculate the association weight between the target discipline and the geographical knowledge, calculate the gating signal value according to the association weight, determine whether to activate the association according to the gating signal value. If the association is activated, plan the target path from the knowledge graph according to the A-Star algorithm, obtain the target discipline knowledge according to the target path, and input the target discipline knowledge and the geographical knowledge into the deep learning model to obtain natural language text; Deeply mine according to the learning questions, use the PC algorithm and the GNN inference algorithm to conduct attribution analysis on geographical problems, extract causal relationships, quantify causal strengths, and generate a visualized causal graph according to the causal relationships and causal strengths.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by a processor to implement the deep geographical learning method based on interdisciplinary knowledge collaboration as described in claim 9.

Images