Scientific and cultural dissemination feasibility assessment method and system based on knowledge graph

Abstract

This invention discloses a feasibility assessment method and system for science and culture dissemination based on knowledge graphs, specifically relating to the fields of computer data processing and knowledge graph technology. It addresses the problem that existing assessment methods struggle to effectively model and quantify the dynamic and evolving state of group cognition. The method involves acquiring temporal interactive text data from dissemination activities, constructing a dynamic group knowledge graph, identifying cognitive sub-groups, calculating the sub-group cross-domain and cohesion of concept nodes to identify core consensus and sets of questionable and divergent concepts, calculating the group cognitive convergence by analyzing the overlap of reasoning paths in different sub-groups, and calculating the group cognitive conflict by analyzing the density of controversial sub-graphs. This comprehensive approach generates a feasibility assessment result for the dissemination process, achieving a dynamic and refined evaluation of the effectiveness of science and culture dissemination.

Description

Technical Field

[0001] This invention relates to the fields of computer data processing and knowledge graph technology, and more specifically, to a method and system for feasibility assessment of scientific and cultural dissemination based on knowledge graphs. Background Technology

[0002] In the field of science and culture communication, especially in open discussions or long-term public participation projects addressing major public issues, evaluating the effectiveness and feasibility of communication activities is crucial for optimizing strategies. Existing technologies utilize knowledge graphs for structured analysis of communication content. By constructing domain knowledge graphs containing scientific concepts, entities, and relationships, static assessments of the completeness, logical coherence, and knowledge coverage of pre-set communication materials can be performed. Furthermore, combined with social network analysis, knowledge graphs can also be used to identify potential key communication nodes or information dissemination paths.

[0003] However, current knowledge graph-based assessment methods mainly analyze static, predefined content and one-way communication paths. Their fundamental limitation lies in the difficulty of effectively modeling and quantifying the dynamic and evolving state of group cognition generated in long-term, multi-round interactions. When communication activities involve extensive public participation and real-time interaction, this static snapshot-based analysis model cannot depict how group understanding evolves over time, how consensus is formed or differentiated, and how new collective cognitions emerge. This results in a serious disconnect between the assessment results and the real, complex dynamics of communication, making it difficult to support real-time insight into the progress of activities and adaptive strategy adjustments. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method and system for assessing the feasibility of science and culture dissemination based on knowledge graphs to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: Feasibility assessment methods for science and culture dissemination based on knowledge graphs include: S1. Obtain the time-series interactive text data generated during the target science and culture dissemination activities; S2. Based on knowledge graph technology, entities and relations are extracted from temporal interactive text data to construct a dynamic group knowledge graph composed of concept nodes and semantic relation edges; S3. Community discovery is performed on dynamic group knowledge graphs to identify multiple cognitive sub-communities; S4. Based on the cognitive sub-communities, calculate the sub-community crossover and sub-community cohesion of each concept node in the dynamic group knowledge graph, and identify the core consensus concept set and the questionable and divergent concept set. S5. Starting from each concept node in the core consensus concept set, perform a path traversal with a finite step length in the dynamic group knowledge graph to obtain the inference path set corresponding to each cognitive subgroup. Calculate the group cognitive convergence based on the overlap between the inference path sets of different cognitive subgroups. Induce a disputed subgraph from the set of questionable and divergent concepts, and calculate the group cognitive conflict degree based on the density of the disputed subgraph. S6. Based on the convergence degree of group cognition and the degree of conflict of group cognition, generate a feasibility assessment result for the progress of the target scientific and cultural dissemination activities.

[0006] Furthermore, acquire the temporal interactive text data generated during the target science and culture dissemination activities, including: Identify at least one interactive platform where the target science and culture dissemination activities take place; collect original interaction records from at least one interactive platform, the original interaction records including text content and timestamps; The text content in the original interaction record is cleaned to remove irrelevant characters and formatting noise, while retaining the timestamps. The cleaned and timestamp-retained text content is then organized into chronological interactive text data according to the timestamp order.

[0007] Furthermore, based on knowledge graph technology, entities and relationships are extracted from temporally sequential interactive text data to construct a dynamic group knowledge graph composed of concept nodes and semantic relationship edges, including: Named entity recognition is performed on temporally interactive text data to extract multiple conceptual entities as candidate conceptual nodes; Relation extraction is performed on temporal interactive text data to identify semantic relationships between multiple conceptual entities as candidate semantic relationship edges; Based on multiple candidate concept nodes and candidate semantic relation edges, a dynamic group knowledge graph composed of concept nodes and semantic relation edges is generated, where each concept node corresponds to a concept entity and each semantic relation edge corresponds to a semantic relation.

[0008] Furthermore, community discovery was performed on the dynamic group knowledge graph, identifying multiple cognitive sub-communities, including: Based on the weights of the semantic relationship edges between concept nodes in the dynamic group knowledge graph, the association strength of each pair of concept nodes is calculated. Based on the correlation strength between all concept nodes, the concept nodes are divided into different candidate sub-groups, such that the correlation strength between concept nodes within the same candidate sub-group is higher than the correlation strength between concept nodes in different candidate sub-groups. The candidate sub-communities obtained from the division are verified, and candidate sub-communities with overly sparse associations are merged or candidate sub-communities with overly loose internal structures are split to obtain multiple cognitive sub-communities.

[0009] Furthermore, based on the sub-community crossover and sub-community cohesion of each concept node in the cognitive sub-community calculation dynamic group knowledge graph, the core consensus concept set and the questionable and divergent concept set are identified, including: For each concept node in the dynamic group knowledge graph, the number of cognitive sub-communities to which it belongs is counted as the sub-community span of the concept node; For each concept node, calculate the sum of the weights of all semantic relation edges between the corresponding concept node and other concept nodes in each cognitive subgroup to which it belongs, and use this sum as the internal association strength of the corresponding concept node in the corresponding cognitive subgroup. The maximum internal association strength is used as the subgroup cohesion of the concept node. Concept nodes whose sub-community crossover exceeds the first threshold and whose sub-community cohesion exceeds the second threshold are included in the core consensus concept set; concept nodes whose sub-community crossover falls below the third threshold and whose sub-community cohesion exceeds the fourth threshold are included in the questionable and divergent concept set.

[0010] Furthermore, the first threshold is greater than the third threshold, and the second threshold is greater than or equal to the fourth threshold.

[0011] Furthermore, the convergence of group cognition is calculated, including: For each concept node in the core consensus concept set, starting from the corresponding concept node, a breadth-first traversal with a preset step size is performed in the dynamic group knowledge graph. The sequence of concept nodes and semantic relationship edges passed during the traversal is recorded as a reasoning path. All reasoning paths obtained by the traversal are taken as the complete set of reasoning paths corresponding to the corresponding concept node. For each cognitive subgroup, inference paths whose starting point and all concept nodes in the path belong to the corresponding cognitive subgroup are selected from the complete set of inference paths, thus forming the inference path set of the corresponding cognitive subgroup. Calculate the Jaccard similarity coefficient between the sets of reasoning paths of any two different cognitive subgroups, and take the average of all Jaccard similarity coefficients to obtain the group cognitive convergence.

[0012] Furthermore, the degree of group cognitive conflict is calculated, including: Extract a subgraph from the dynamic group knowledge graph that contains all concept nodes and all direct semantic relationship edges between them in the set of questionable and divergent concepts as the disputed subgraph; calculate the ratio of the number of semantic relationship edges to the number of concept nodes in the disputed subgraph to obtain the density of the disputed subgraph as the group cognitive conflict degree.

[0013] Furthermore, based on the convergence and conflict of group cognition, a feasibility assessment of the progress of the target science and culture dissemination activities is generated, including: Map the convergence of group cognition to a preset convergence level range, and map the conflict level of group cognition to a preset conflict level range; Based on the combination of convergence level intervals and conflict level intervals, a preset propagation state matrix is ​​matched to obtain the corresponding target scientific and cultural dissemination activity process status identifier. Based on the status identifiers of the target scientific and cultural dissemination activities, a feasibility assessment result containing status descriptions and quantitative indicators is generated.

[0014] On the other hand, this invention provides a feasibility assessment system for the dissemination of science and culture based on knowledge graphs, including: The data acquisition module is used to acquire time-series interactive text data generated during the target science and culture dissemination activities; The knowledge graph construction module is used to extract entities and relationships from temporally interactive text data based on knowledge graph technology, and to construct a dynamic group knowledge graph composed of concept nodes and semantic relationship edges. The subgroup discovery module is used to discover communities in dynamic group knowledge graphs and identify multiple cognitive subgroups. The concept identification module is used to calculate the sub-community crossover and sub-community cohesion of each concept node in the dynamic group knowledge graph based on the cognitive sub-communities, and to identify the core consensus concept set and the questionable and divergent concept set. The measurement and calculation module is used to start from each concept node in the core consensus concept set, perform path traversal with a finite step length in the dynamic group knowledge graph to obtain the inference path set corresponding to each cognitive subgroup, calculate the group cognitive convergence based on the overlap between the inference path sets of different cognitive subgroups; induce the disputed subgraph from the questionable and divergent concept set, and calculate the group cognitive conflict degree based on the density of the disputed subgraph. The evaluation generation module is used to generate a feasibility assessment result for the progress of the target science and culture dissemination activities based on the degree of convergence and conflict of group cognition.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. By constructing a dynamic group knowledge graph and conducting in-depth quantitative analysis of the group's cognitive structure, a dynamic and refined assessment of the progress of science and culture dissemination activities is achieved. Traditional methods rely on static knowledge bases and one-way communication analysis, making it difficult to capture the evolution of group cognition in multiple rounds of interaction. This solution directly constructs and continuously updates the knowledge graph from temporal interactive texts, enabling it to truly reflect the migration of discussion hotspots and the evolution of conceptual associations. Through the cognitive sub-clusters identified through community discovery, the solution accurately depicts the naturally formed cluster structure of viewpoints within the group, providing a structured foundation for understanding the distribution of consensus and disagreement.

[0016] 2. Based on the sub-community cross-cutting and cohesion indicators defined by the dynamic group knowledge graph, and the derived group cognitive convergence and conflict degree, a collaborative quantitative assessment of the depth of consensus formation and the complexity of disagreements during the dissemination process is achieved. This transforms the assessment conclusion from a superficial statistical analysis of content coverage into a profound insight into the evolution trend and internal contradictions of the group's cognitive state. Based on this quantitative assessment result, dissemination bottlenecks can be identified in real time, key points of contention can be located, and the effectiveness of intervention measures can be evaluated. This provides reliable data-driven decision support for dynamically adjusting dissemination strategies, optimizing resource allocation, and improving the final effect of scientific and cultural dissemination. Attached Figure Description

[0017] Figure 1 This is a flowchart of the feasibility assessment method for science and culture dissemination based on knowledge graphs according to the present invention; Figure 2 This is a schematic diagram of the structure of the science and culture dissemination feasibility assessment system based on knowledge graphs of this invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1: Figure 1 This invention presents a feasibility assessment method for science and culture dissemination based on knowledge graphs, including: S1. Obtain the time-series interactive text data generated during the target science and culture dissemination activities; S2. Based on knowledge graph technology, entities and relations are extracted from temporal interactive text data to construct a dynamic group knowledge graph composed of concept nodes and semantic relation edges; S3. Community discovery is performed on dynamic group knowledge graphs to identify multiple cognitive sub-communities; S4. Based on the cognitive sub-communities, calculate the sub-community crossover and sub-community cohesion of each concept node in the dynamic group knowledge graph, and identify the core consensus concept set and the questionable and divergent concept set. S5. Starting from each concept node in the core consensus concept set, perform a path traversal with a finite step length in the dynamic group knowledge graph to obtain the inference path set corresponding to each cognitive subgroup. Calculate the group cognitive convergence based on the overlap between the inference path sets of different cognitive subgroups. Induce a disputed subgraph from the set of questionable and divergent concepts, and calculate the group cognitive conflict degree based on the density of the disputed subgraph. S6. Based on the convergence degree of group cognition and the degree of conflict of group cognition, generate a feasibility assessment result for the progress of the target scientific and cultural dissemination activities.

[0020] Step S1 involves acquiring the temporal interactive text data generated during the target science and culture dissemination activities, which is achieved through the following process: Identify at least one interactive platform where the target science and culture dissemination activity takes place. Identification of interactive platforms is accomplished by analyzing publicly available information about the target science and culture dissemination activity, including activity announcements, promotional materials, or participation rules. This publicly available information will indicate the primary online venues where the activity takes place. For example, for an online debate on the ethics of artificial intelligence, interactive platforms might include a specific social media hashtag or a dedicated section of an online forum. In practice, operators or automated monitoring systems identify one or more interactive platforms as data sources based on activity information. When the activity is conducted simultaneously on multiple platforms, all relevant platforms are listed as data collection targets.

[0021] Raw interaction records are collected from at least one identified interactive platform. Raw interaction records refer to user participation data directly obtained from the interactive platform without substantial content modification. The collection process utilizes publicly available application programming interfaces (APIs) provided by the corresponding interactive platform or web crawling techniques that follow its bot access protocols. For example, for platforms providing APIs, data is collected by calling API functions to retrieve a list of comments or posts and parsing the returned JSON data. For platforms primarily accessed via web pages, a crawler is written to simulate a browser initiating HTTP requests, receiving HTML responses, and using a Document Object Model (DOM) parser or specific selectors to extract user-posted text blocks and associated time information from the page. Each successfully collected raw interaction record must contain two core elements: text content and a timestamp. The text content is the substantial textual information posted by the user. The timestamp is the point in time when the record was generated and recorded on the interactive platform's server; the timestamp format may be the number of seconds since January 1, 1970, or a string in the format of year-month-day hour:minute:second. The collection program acquires and temporarily stores the text content and timestamp as inseparable data pairs.

[0022] Data cleaning is performed on the text content of the collected raw interaction records. The purpose of data cleaning is to remove irrelevant characters and formatting noise from the text content, while strictly preserving the timestamp associated with each raw interaction record. Irrelevant characters refer to characters that do not carry substantial semantic meaning and may interfere with subsequent analysis. Irrelevant characters include Hypertext Markup Language (HTML) tags, Cascading Style Sheet (CSS) code, multiple consecutive spaces, tabs, and newlines obtained during the collection process. Formatting noise includes garbled character sequences caused by network transmission or encoding / parsing errors. Data cleaning is performed by applying a series of predefined string processing rules. For example, a regular expression pattern matching the content between less than and greater than signs is used to identify and remove all HTML tags. Another regular expression pattern is used to replace more than one consecutive whitespace character with a single space character. Character encoding validation is performed on the text content. When a byte sequence that does not conform to the UTF-8 encoding rules is found, the sequence is replaced with a predefined placeholder or directly removed. The cleaning process is performed in memory or temporary storage area to ensure that the timestamp of each original interaction record maintains an accurate correspondence with the cleaned text content. The timestamp itself does not participate in any string transformation operations.

[0023] The cleaned text content, retaining its timestamps, is organized into chronological interactive text data according to the order of its timestamps. The organization process begins with timestamp parsing. Timestamp parsing converts various raw timestamp formats into an internal representation usable for numerical comparison. For example, a string timestamp in the format of year-month-day hour:minute:second is parsed and converted into an integer value representing the number of seconds elapsed from a standard reference time point to that time point. Then, chronological sorting is performed. Based on the unified internal time representation, all data entries are arranged in ascending order of value, using either quicksort or mergesort algorithms. After sorting, an ordered data list is formed, where each element contains the cleaned text content and its corresponding internal time representation value. This ordered data list is the chronological interactive text data output from step S1. The chronological interactive text data reflects the sequential relationship of the accumulation and evolution of text content in group interactions over time during the target scientific and cultural dissemination activities.

[0024] In step S2, entity and relation extraction is performed on the temporal interactive text data based on knowledge graph technology, and a dynamic group knowledge graph is constructed. This is achieved through the following process: Named entity recognition (NID) is performed on temporally interactive text data to extract multiple conceptual entities as candidate concept nodes. The input to NID is the text content portion of the temporally interactive text data output in step S1. The NID process employs a sequence labeling method combining a pre-trained language model and a Conditional Random Field (CRF) model. The pre-trained language model uses, for example, the BERT model, which maps each character in the input text sequence to a high-dimensional context-sensitive vector representation. Each piece of text content in the temporally interactive text data is input to the BERT model, which outputs a sequence of vector representations corresponding to each character in the text. The CRF model receives this sequence of vector representations as input features and predicts an entity label for each character in the sequence. The entity labeling system is predefined based on the characteristics of the science and culture dissemination field, and includes entity types such as scientific concept entities, technical method entities, and people / organization entities. The CRF model uses the Viterbi decoding algorithm to find the label sequence with the highest probability from all possible label sequences as the prediction result. After the model completes its predictions, consecutive character segments labeled as the same entity type in the text are merged to form a complete entity mention. For multiple different representations of the same entity that may exist in the text, a pre-built thesaurus is used to merge them, pointing to a standardized entity name. Each standardized entity name after merging is defined as a conceptual entity. The set of all conceptual entities identified and normalized from temporally interactive text data constitutes multiple candidate conceptual nodes for constructing a knowledge graph. For example, from a text discussing gene editing, the conceptual entities that might be identified and normalized include CRISPR-Cas9, off-target effects, and ethical review.

[0025] This study extracts relations from temporal interactive text data and identifies semantic relationships between multiple concept entities as candidate semantic relation edges. The input for relation extraction is the text content of the temporal interactive text data, and it also utilizes the positional information of concept entities in the text obtained from the named entity recognition step. The relation extraction process employs a sentence-level relation classification method based on a pre-trained language model. For each piece of text content in the temporal interactive text data, a relation classification sample is constructed for every two identified concept entity pairs. During sample construction, the text mentions of the two concept entities are labeled in the sentence by inserting special markers: a start marker and an end marker are inserted before and after the first entity mention, and another start marker and another end marker are inserted before and after the second entity mention. The labeled sentences are then input into the pre-trained language model. The pre-trained language model receives the sentences at the input end and outputs a vector representation corresponding to the position of the sentence classification marker as the relational semantic representation of the entire sentence. The semantic representation vector of the relation is fed into a fully connected layer classifier, which outputs the probability distribution of whether the two conceptual entities described in the sentence belong to a predefined relation type. The predefined relation types are pre-defined based on a scientific knowledge system and include those belonging to a relation type, causing a relation type, and applying to a relation type. The classifier predicts a relation type for each pair of conceptual entities. When the predicted probability exceeds a pre-set relation confidence threshold, the text is considered to express this semantic relationship between the two conceptual entities. The relation confidence threshold is determined through performance evaluation on a validation dataset containing sentence samples with correctly labeled relation types. By adjusting the value of the relation confidence threshold and observing the model's precision and recall on the validation set, a value that optimizes the overall evaluation metrics is selected as the relation confidence threshold. The adjustment range of the relation confidence threshold is typically between 0.6 and 0.8. Each pair of conceptual entities with semantic relation links identified from all texts, along with the relation type, collectively defines a candidate semantic relation edge. For example, a semantic relationship between the conceptual entity "high-sugar diet" and the conceptual entity "type 2 diabetes" may be identified from the text, resulting in a relation type.

[0026] A dynamic community knowledge graph composed of concept nodes and semantic relationship edges is generated based on multiple candidate concept nodes and candidate semantic relationship edges. The generation process includes graph structure initialization. An empty graph data structure is created, supporting the storage of nodes and edges and recording their attributes. All candidate concept nodes are traversed, and each unique concept entity is added as a concept node to the graph data structure. The attribute stored in a concept node includes its standardized name. All candidate semantic relationship edges are traversed. For each candidate semantic relationship edge, it is checked whether the two concept entities associated with the edge already exist as concept nodes in the graph data structure. If they exist, a link is created between the two corresponding concept nodes, and this link is added as a semantic relationship edge to the graph data structure. The attribute stored in a semantic relationship edge includes the relationship type. For cases where the same relationship type is identified multiple times between two concept nodes, the semantic relationship edge is reinforced by adding a weight attribute. The initial value of the weight attribute is set to the number of times the relationship type is observed. The weight attribute characterizes the frequency and intensity of the semantic relationship being mentioned and confirmed in community interactions. After constructing the basic graph structure, dynamic temporal dimension association is performed. The timestamp information of the temporally sequential interactive text data is linked to the construction of the knowledge graph. For each semantic relation edge, the earliest and most recently identified timestamps are recorded. The temporally sequential interactive text data is divided into fixed time windows, such as by day, and a subgraph snapshot is constructed for each candidate concept node and candidate semantic relation edge identified within each time window. All subgraph snapshots of consecutive time windows are arranged in chronological order, forming the temporal evolution sequence of the dynamic group knowledge graph. The final generated dynamic group knowledge graph is a graph network containing a set of concept nodes and a set of semantic relation edges. Each concept node corresponds to a concept entity extracted from the interactive text, and each semantic relation edge corresponds to a semantic relation identified from the interactive text. The semantic relation edges have weight attributes and time stamps, and the graph itself has multiple version snapshots that evolve over time.

[0027] Step S3 involves community discovery and identification of multiple cognitive sub-communities within the dynamic group knowledge graph, achieved through the following process: Based on the weights of semantic relationship edges between concept nodes in the dynamic group knowledge graph, the association strength of each pair of concept nodes is calculated. The input is the full graph structure data of the latest snapshot of the dynamic group knowledge graph generated in step S2, including the set of concept nodes, the set of semantic relationship edges, and the weight attribute of each semantic relationship edge. In step S2, the weight attribute of a semantic relationship edge is set to the number of times the relationship is observed in the text; therefore, its value is an integer greater than or equal to 1, and its physical meaning reflects the frequency with which the semantic relationship is commonly confirmed in group discussions. Association strength is a quantified value used to characterize the semantic connection between any two concept nodes in the graph. For any two different concept nodes in the dynamic group knowledge graph, it is checked whether there is a direct semantic relationship edge between them. If one or more direct semantic relationship edges connect the two concept nodes, the weight attribute values ​​of all semantic relationship edges connecting the two concept nodes are summed, and the sum is taken as the direct association strength component between the two concept nodes. If there is no direct semantic relationship edge between the two concept nodes, the indirect association strength component is calculated using a graph search algorithm. A breadth-first search algorithm is used to find all shortest paths connecting the two concept nodes. The shortest paths consist of a series of semantic relationship edges connected end-to-end. The indirect association strength component is calculated based on the path decay principle. For a shortest path of length L, where L is the number of semantic relationship edges, the path strength is defined as the product of the weights of all semantic relationship edges on that path, divided by a preset path decay coefficient γ raised to the power of L. The path decay coefficient γ is a real number greater than 1, and its specific value is set to ensure that the strength of the indirect association decays rapidly with increasing path length. The path decay coefficient γ can be determined by analyzing the graph diameter and average path length, selecting a value that makes the contribution of a path of length 3 negligible; for example, an experimental setting of γ to 2 is used. The path strength values ​​of all shortest paths connecting the two concept nodes are summed to obtain the indirect association strength component. If there are neither direct semantic relationship edges nor connected paths between the two concept nodes, the association strength of this pair of concept nodes is defined as 0. The final association strength between the concept node pairs equals the direct association strength component plus the indirect association strength component. For all possible concept node pairs in the dynamic group knowledge graph, repeat the above calculation process to obtain a complete association strength matrix. The rows and columns of the matrix correspond to each concept node, and the element values ​​in the matrix represent the association strength between the corresponding node pairs.

[0028] Based on the association strength between all concept nodes, concept nodes are divided into different candidate sub-clusters, such that the association strength between concept nodes within the same candidate sub-cluster is higher than the association strength between concept nodes in different candidate sub-clusters. The partitioning process uses a community detection algorithm based on modularity optimization. Modularity Q is a scalar indicator for evaluating the quality of partitioning. The calculation of modularity Q is based on a comparison principle, namely, comparing the association strength within each candidate sub-cluster in the actual network with the expected association strength in a corresponding random network. First, initialization is performed, treating each concept node in the dynamic community knowledge graph as an independent candidate sub-cluster. The modularity Q value is calculated in the current partitioning state. The specific calculation method of modularity Q value is as follows: for the entire association strength matrix, traverse all concept node pairs. If a pair of concept nodes belongs to the same candidate sub-cluster, subtract the expected association strength value of the two concept nodes in the random network from the association strength value between the two concept nodes, and sum the differences of all node pairs belonging to the same candidate sub-cluster. Finally, normalize the sum by dividing the sum by twice the sum of all association strength values ​​in the network. The expected association strength of each concept node in the random network is proportional to its total association strength, which is the sum of the association strengths between that node and all other nodes in the network. Then, an iterative optimization process is performed to improve the modularity Q-value. Each concept node is traversed, and for the currently traversed concept node, an attempt is made to move it from its current candidate subgroup to another candidate subgroup belonging to each of its neighboring concept nodes. Neighboring concept nodes are those with an association strength greater than 0 with the currently traversed concept node. For each move attempt, the change in the modularity Q-value of the entire network after the move is calculated. The move attempt that maximizes the increase in modularity Q-value is selected and executed, i.e., changing the candidate subgroup belonging to the currently traversed concept node. If no move attempt increases the modularity Q-value, the currently traversed concept node retains its original candidate subgroup affiliation. This traversal and move attempt process is repeated on all concept nodes until a complete traversal is completed, and the modularity Q-value no longer increases, at which point the algorithm reaches a stable state. After the algorithm converges, the resulting concept node partitioning is divided into multiple candidate sub-communities, each of which is a set of concept nodes.

[0029] The candidate sub-communities obtained from the segmentation are validated. Candidate sub-communities with excessively sparse associations are merged, or those with excessively loose internal structures are split, resulting in multiple cognitive sub-communities. The validation process first evaluates the internal association density of each candidate sub-community. The internal association density of a candidate sub-community is defined as the ratio of the sum of the actual association strengths between all concept node pairs within that candidate sub-community to the sum of the theoretical maximum association strengths of all possible concept node pairs within that candidate sub-community. The theoretical maximum association strength is calculated using a normalized benchmark, which can be set as the sum of strengths when the association strength between all node pairs within the candidate sub-community is equal to the maximum association strength value in the entire association strength matrix. The internal association density value of each candidate sub-community is calculated. An internal association density threshold is set to determine whether a candidate sub-community has an excessively loose internal structure. The internal association density threshold is obtained by running the pre-process of step S3 on an interactive dataset of several representative example science and culture dissemination activities. This yields a series of candidate sub-community partitioning results. Domain experts, or those based on known dissemination structures, evaluate the quality of these partitioning results, identifying which sub-communities are reasonable and internally compact, and which are loosely structured. Then, the distribution range of the internal association density values ​​of these reasonably labeled sub-communities is statistically analyzed. A value at the lower quantile of this distribution is selected as the internal association density threshold; for example, selecting a value at the 20th quantile can filter out approximately 80% of loosely structured sub-communities. The internal association density threshold is typically adjusted between 0.2 and 0.4. If the internal association density of a candidate sub-community is lower than the internal association density threshold, a splitting operation is performed on that candidate sub-community. The splitting operation uses a method based on edge betweenness centrality. The edge betweenness of all semantic relationship edges within the candidate sub-community is calculated. Edge betweenness is defined as the number of paths passing through that edge in the shortest path between all concept node pairs within the candidate sub-community. Find the semantic relation edge with the highest betweenness value within the candidate subcommunity and temporarily remove it from the current subgraph structure. After removal, check the connectivity of the original candidate subcommunity. If removal causes the original candidate subcommunity to split into two or more disconnected node sets, define each connected node set as a new candidate subcommunity. If no split occurs, attempt to remove the semantic relation edge with the second highest betweenness value, repeating this process until the candidate subcommunity is successfully split. The verification process then evaluates the external association strength between different candidate subcommunities. For any two different candidate subcommunities, calculate their external association strength, defined as the average of the association strength values ​​between all concept node pairs belonging to each of the two candidate subcommunities. Set an external association strength threshold to determine whether the association between two candidate subcommunities is too sparse.The external association strength threshold is obtained similarly. By analyzing the example dataset, the distribution of external association strength values ​​between subcommunity pairs determined by experts or known structures to be independent and with clear boundaries is statistically analyzed. A value at the upper quantile of this distribution is selected as the external association strength threshold, for example, the value at the 80th quantile. This allows merging subcommunities with excessively close associations where the external association is higher than approximately 80%. The external association strength threshold is typically adjusted between 0.1 and 0.2. If the external association strength between two candidate subcommunities is higher than the external association strength threshold, these two candidate subcommunities are merged into a new, larger candidate subcommunity. After the merging operation, the internal association density of the newly generated candidate subcommunity needs to be recalculated to ensure it is not lower than the internal association density threshold. The above verification, splitting, and merging process is iteratively executed until all remaining candidate subcommunities simultaneously meet the following two conditions: Condition 1, the internal association density of each candidate subcommunity is not lower than the internal association density threshold; Condition 2, the external association strength between any two different candidate subcommunities is not higher than the external association strength threshold. The set of candidate sub-communities that finally stabilizes after this iterative optimization process is the set of multiple cognitive sub-communities to be identified in step S3.

[0030] In step S4, the sub-community crossover and sub-community cohesion of each concept node in the dynamic group knowledge graph are calculated based on the cognitive sub-communities, and the core consensus concept set and the questionable and divergent concept set are identified. This is achieved through the following process: For each concept node in the dynamic group knowledge graph, the number of cognitive sub-groups to which it belongs is counted as the sub-group span of the concept node. The input for this calculation is the multiple cognitive sub-groups identified in step S3 and the set of concept nodes in the dynamic group knowledge graph. Each cognitive sub-group is a set of concept nodes output in step S3. For each concept node in the dynamic group knowledge graph, a traversal operation is performed. The traversal operation involves sequentially checking each cognitive sub-group identified in step S3. The check determines whether the currently processed concept node belongs to the set of member nodes of the currently checked cognitive sub-group. If the concept node belongs to the currently checked cognitive sub-group, the cognitive sub-group is recorded as a cognitive sub-group to which the concept node belongs. After checking all the cognitive sub-groups identified in step S3, the total number of cognitive sub-groups recorded as belonging to the cognitive sub-group of the concept node is counted. This total number is the value of the sub-group span of the concept node. The sub-group span is a non-negative integer. The above process of traversing, checking, and counting is repeated for each concept node in the dynamic community knowledge graph to calculate a subcommunity span value for each concept node.

[0031] For each concept node, the sum of the weights of all semantic relation edges between the corresponding concept node and other concept nodes within each cognitive sub-group is calculated. This sum is used as the internal association strength of the corresponding concept node within its corresponding cognitive sub-group, and the highest internal association strength is taken as the sub-group cohesion of the concept node. The inputs for this calculation include the dynamic group knowledge graph, the weight attributes of the semantic relation edges in the dynamic group knowledge graph, the multiple cognitive sub-groups identified in step S3, and the set of cognitive sub-groups to which each concept node belongs, calculated in the previous step. The weight attributes of the semantic relation edges are integer values ​​representing the frequency of observation of the relation, as defined in step S2. For a specific concept node, the set of cognitive sub-groups to which the concept node belongs is first obtained. Then, for each cognitive sub-group in this set, the internal association strength is calculated. The internal association strength calculation is performed on the current concept node and the currently processed cognitive sub-group. From the set of member nodes of the currently processed cognitive sub-group, the current concept node itself is excluded, resulting in the set of all other concept nodes within the current cognitive sub-group, called the neighbor concept node set. Initialize an accumulator variable, setting its value to 0. Iterate through each neighboring concept node in the neighboring concept node set. In the dynamic group knowledge graph, find all direct semantic relationship edges originating from the current concept node and ending at the currently traversed neighboring concept node, and then find all direct semantic relationship edges originating from the currently traversed neighboring concept node and ending at the current concept node. Sum the weight attribute values ​​of all these found semantic relationship edges to obtain a connection strength value. Add this connection strength value to the accumulator variable. After traversing all nodes in the neighboring concept node set, the final value stored in the accumulator variable is the internal association strength of the current concept node within the currently processed cognitive sub-group. When a concept node belongs to multiple cognitive sub-groups, repeat the above internal association strength calculation process for each cognitive sub-group to obtain multiple internal association strength values. The sub-group cohesion of a concept node is defined as the maximum value selected from all these calculated internal association strength values. For each concept node in the dynamic community knowledge graph, repeat the process of obtaining its set, calculating the strength of each internal association, and finding the maximum value to calculate a value of subcommunity cohesion for each concept node.

[0032] Concept nodes whose sub-group span exceeds a first threshold and whose sub-group cohesion exceeds a second threshold are included in the core consensus concept set. The core consensus concept set is initialized as an empty set of concept node identifiers. The first threshold is a numerical threshold used to filter concept nodes with high sub-group span. The first threshold is set based on the statistical distribution of the sub-group span values ​​of all concept nodes. It is obtained by calculating the average of the sub-group span values ​​of all concept nodes, then calculating the standard deviation of these values, and finally adding one standard deviation to the average as a candidate value for the first threshold. This candidate value is rounded up to the nearest integer to obtain the final first threshold. The second threshold is a numerical threshold used to filter concept nodes with high sub-group cohesion. It is set based on the statistical distribution of the sub-group cohesion values ​​of all concept nodes. It is obtained by calculating the upper quartile of the sub-group cohesion values ​​of all concept nodes. Sort all concept nodes by sub-cluster cohesion values ​​in ascending order, and find the value located at three-quarters of the total number; this value is the upper quartile and is used as the second threshold. During the inclusion operation, traverse every concept node in the dynamic group knowledge graph. For each concept node, read its sub-cluster spanning and sub-cluster cohesion values. Determine if the sub-cluster spanning value is greater than the first threshold and if the sub-cluster cohesion value is greater than the second threshold. If both conditions are met, add the concept node's unique identifier to the core consensus concept set.

[0033] Concept nodes with subcommunity scalability below the third threshold and subcommunity cohesion above the fourth threshold are categorized into a questionable divergent concept set. The questionable divergent concept set is initialized as an empty set of concept node identifiers. The third threshold is a numerical threshold used to filter concept nodes with low subcommunity scalability. The value of the third threshold must be less than the first threshold. The third threshold is set based on the statistical distribution of the subcommunity scalability values ​​of all concept nodes. The third threshold is obtained by calculating the lower quartile of the subcommunity scalability values ​​of all concept nodes. All concept nodes are sorted in ascending order of their subcommunity scalability values, and the value at the one-quarter mark is the lower quartile, which is used as the third threshold. The fourth threshold is a numerical threshold used to determine whether the subcommunity cohesion is sufficiently high when filtering questionable divergent concepts. The fourth threshold is set to be less than or equal to the second threshold. One way to obtain the fourth threshold is to set it directly to the same value as the second threshold. Another way is to set the fourth threshold as the second threshold minus a small positive offset, for example, 5% of the second threshold value. During the inclusion operation, each concept node in the dynamic group knowledge graph that has not yet been included in the core consensus concept set is traversed. For each such concept node, its sub-group spanning value and sub-group cohesion value are read. It is determined whether the sub-group spanning value of the concept node is less than a third threshold and whether the sub-group cohesion value of the concept node is greater than a fourth threshold. If both comparison conditions are met, the unique identifier of the concept node is added to the questionable divergent concept set. Through the above process, step S4 finally outputs the core consensus concept set and the questionable divergent concept set.

[0034] In step S5, starting from each concept node in the core consensus concept set, a path traversal with a finite step length is performed in the dynamic group knowledge graph to obtain the inference path set corresponding to each cognitive subgroup. The group cognitive convergence is calculated based on the overlap between the inference path sets of different cognitive subgroups. Furthermore, a disputed subgraph is induced from the set of questionable and divergent concepts, and the group cognitive conflict degree is calculated based on the density of the disputed subgraph. This is achieved through the following process: For each concept node in the core consensus concept set, a breadth-first traversal with a preset step size is performed in the dynamic group knowledge graph, starting from the corresponding concept node. The sequence of concept nodes and semantic relationship edges traversed during the traversal is recorded as a reasoning path, and all reasoning paths obtained are considered the complete set of reasoning paths corresponding to the concept node. The input for this operation includes the core consensus concept set output in step S4, the full set of nodes and edges in the dynamic group knowledge graph at the latest snapshot, and the cognitive sub-community information identified in step S3. The preset step size is a positive integer parameter used to control the depth of path exploration. The specific value of the preset step size is determined by analyzing the topological features of the dynamic group knowledge graph. The method is to calculate the shortest path length between all pairs of concept nodes with connected paths in the dynamic group knowledge graph, and then calculate the average of these shortest path lengths. The calculated average is rounded up to the nearest integer, and then 1 is added to this integer; the resulting value is used as the preset step size. For example, if the average shortest path length is 2.3, rounding up gives 3, and adding 1 gives 4, then the preset step size is set to 4. The preset step size is at least 2. For each concept node in the core consensus concept set, a complete breadth-first traversal is performed independently. At the start of the breadth-first traversal, an empty queue and an empty set for storing results are created. A path containing only the starting concept node is added to the queue as the initial state. This initial path is recorded as the identifier of the starting concept node. Entering the loop, when the queue is not empty, a current path is retrieved from the head of the queue. The depth of the current path is checked; path depth is defined as the number of semantic relation edges contained in the current path. If the depth of the current path equals a preset step size, the current path is stored as a complete inference path in the result set, and no further expansion is based on this path. If the depth of the current path is less than the preset step size, the identifier of the concept node at the end of the current path is obtained. In the edge index data structure of the dynamic group knowledge graph, all semantic relation edge records associated with the ending concept node are queried. For each associated semantic relation edge record, the identifier of the other endpoint concept node of that edge record is obtained. The current path is copied, and the identifier of the currently processed semantic relation edge is appended to the end of the copied path sequence, followed by the identifier of the other endpoint concept node, forming a new extended path. Check if a loop exists in the newly expanded path, i.e., whether the identifier of the newly added concept node already exists in the current path sequence. If the identifier of the newly added concept node has not appeared in the current path sequence, add the new expanded path to the tail of the queue. If the identifier of the newly added concept node already exists, discard the new expanded path to avoid a loop. Repeat this process until the queue is empty. At this point, the result set contains all the path sequences corresponding to the starting concept node, which is the complete set of inference paths.Each inference path sequence is a list consisting of alternating concept node identifiers and semantic relation edge identifiers, starting and ending with a concept node identifier. The above independent breadth-first traversal process is repeated for each concept node in the core consensus concept set to obtain the complete set of inference paths corresponding to each concept node in the set.

[0035] For each cognitive subgroup, inference paths whose starting point and all concept nodes in the path belong to the corresponding cognitive subgroup are selected from the complete set of inference paths, forming the inference path set for the corresponding cognitive subgroup. The input for this operation includes the list of member node identifiers for each cognitive subgroup identified in step S3, and the complete set of inference paths generated for each concept node in the core consensus concept set in the previous step. For each cognitive subgroup identified in step S3, the selection operation is performed. An empty set is initialized as the inference path set for the current cognitive subgroup. The list of member node identifiers for the current cognitive subgroup is obtained. The complete set of inference paths corresponding to each concept node in the core consensus concept set is traversed. For each inference path in the currently traversed complete set of inference paths, the first element in the inference path sequence is read, i.e., the starting concept node identifier. It is checked whether the starting concept node identifier exists in the list of member node identifiers for the current cognitive subgroup. If it does not exist, the inference path is skipped. If it exists, all elements in the inference path sequence located at odd positions (i.e., the first, third, fifth, etc.) are read; these elements are all concept node identifiers. Check if each of these concept node identifiers exists in the list of member node identifiers of the current cognitive subgroup. This is done by comparing each concept node identifier to be checked with the list of member node identifiers. If all checked concept node identifiers belong to the current cognitive subgroup, the complete inference path sequence is added as an element to the inference path set of the current cognitive subgroup. If any checked concept node identifier does not belong to the current cognitive subgroup, the inference path is skipped. After traversing the complete set of all inference paths, the inference path set of the current cognitive subgroup is constructed. For each cognitive subgroup identified in step S3, repeat the above filtering operation to construct a corresponding inference path set for each cognitive subgroup.

[0036] Calculate the Jaccard similarity coefficient between the reasoning path sets of any two different cognitive sub-clusters, and average all Jaccard similarity coefficients to obtain the group cognitive convergence. The input for this calculation is the reasoning path set constructed for each cognitive sub-cluster in the previous step. For any two different cognitive sub-clusters identified in step S3, calculate the Jaccard similarity coefficient between the reasoning path sets of the first and second cognitive sub-clusters, as follows: First, calculate the intersection of the two reasoning path sets and count the number of reasoning paths contained in the intersection, denoted as X. Then, calculate the union of the two reasoning path sets and count the number of reasoning paths contained in the union, denoted as Y. The Jaccard similarity coefficient is calculated as: J = X / Y. Where J represents the Jaccard similarity coefficient, X represents the number of elements in the intersection, and Y represents the number of elements in the union. The Jaccard similarity coefficient ranges from 0 to 1. Traverse all cognitive sub-clusters identified in step S3 to generate all possible pairs of different cognitive sub-clusters. Assuming the total number of cognitive sub-communities is M, the total number of different cognitive sub-community pairs is calculated as: P = M × (M−1) / 2. For each different cognitive sub-community pair, its corresponding Jaccard similarity coefficient is calculated using the above method. The group cognitive convergence is calculated as: C = (J1 + J2 + … + Jp) / P. Where C represents the group cognitive convergence, J1 to Jp represent the calculated Jaccard similarity coefficients, P represents the total number of different cognitive sub-community pairs, and p represents the index of the cognitive sub-community pair. The group cognitive convergence is a real number between 0 and 1.

[0037] Extract a subgraph from the dynamic community knowledge graph that contains all concept nodes in the questionable and divergent concept set and all direct semantic relationship edges between them as the disputed subgraph. The input to this operation includes the questionable and divergent concept set output from step S4 and the full node and edge data of the latest snapshot in the dynamic community knowledge graph. First, create an empty node set and an empty edge set. Iterate through each concept node identifier in the questionable and divergent concept set. For each concept node identifier encountered, find the corresponding concept node record in the node data of the dynamic community knowledge graph. Add the found concept node record to the node set. Then, iterate through all semantic relationship edge records in the dynamic community knowledge graph. For each semantic relationship edge record, read the concept node identifiers of the two endpoints of the edge record. Check if both endpoint concept node identifiers exist in the node set. This is done by comparing each endpoint identifier with the identifiers of all nodes in the node set. If both endpoint concept node identifiers of a semantic relationship edge record belong to the node set, then add the semantic relationship edge record to the edge set. After traversing all edges, the graph structure formed by the set of nodes and the set of edges is the controversial subgraph induced from the dynamic collective knowledge graph. The controversial subgraph completely includes all concept nodes in the set of questionable and divergent concepts, as well as all direct connecting edges between these nodes that exist in the original graph.

[0038] The density of the disputed subgraph is calculated as the ratio of the number of semantic relation edges to the number of concept nodes, and is used as the degree of group cognitive conflict. The input for this calculation is the disputed subgraph obtained in the previous step. First, the number of concept nodes in the disputed subgraph is counted, denoted as N. Then, the number of semantic relation edges in the disputed subgraph is counted, denoted as E. The density of the disputed subgraph is calculated as: D = E / N. Where D represents the density of the disputed subgraph, E represents the number of semantic relation edges, and N represents the number of concept nodes. Density D is a real number greater than or equal to 0. When there is only one node in the node set, the edge set is empty, and the density D is 0. The value of density D serves as a measure of the degree of group cognitive conflict. Through the above process, step S5 finally outputs two quantitative indicators: group cognitive convergence and group cognitive conflict.

[0039] Step S6 generates a feasibility assessment result for the progress of the target science and culture dissemination activity based on the degree of convergence and conflict of group cognition. This is achieved through the following process: The group cognitive convergence is mapped to a preset convergence level interval. The input for this mapping operation is the group cognitive convergence value calculated in step S5. The preset convergence level interval is a set of numerical ranges predefined in the system configuration, used to classify continuous group cognitive convergence values ​​into discrete levels. The preset convergence level interval is determined by analyzing the theoretical range and historical actual distribution of group cognitive convergence. The theoretical range of group cognitive convergence is 0 to 1. The preset convergence level interval is set to three intervals: low convergence interval, medium convergence interval, and high convergence interval. The low convergence interval is defined as the range from 0 to a low convergence threshold, where 0 is included in the interval and the low convergence threshold is not included. The medium convergence interval is defined as the range from a low convergence threshold to a high convergence threshold, where the low convergence threshold is included in the interval and the high convergence threshold is not included. The high convergence interval is defined as ranging from the high convergence threshold to 1, where the high convergence threshold and 1 are both included within the interval. The specific values ​​of the low and high convergence thresholds are set based on the analysis results of historical science and culture dissemination activities. The analysis method involves collecting interaction data from multiple completed historical dissemination activities, independently running steps S1 to S5 for each activity to obtain the corresponding group cognitive convergence value. Domain experts are invited to qualitatively rate the consensus-forming effect of each historical activity based on its actual effects and records, with rating categories including poor consensus-forming, moderate consensus-forming, and good consensus-forming. Then, the group cognitive convergence values ​​corresponding to all historical activities rated as poor by experts are grouped into a numerical set. The upper quartile of this numerical set is calculated and used as a candidate value for the low convergence threshold. The group cognitive convergence values ​​corresponding to all historical activities rated as good by experts are grouped into another numerical set. The lower quartile of this numerical set is calculated and used as a candidate value for the high convergence threshold. In practical applications, these candidate values ​​can be directly used as low-convergence and high-convergence thresholds, or they can be fine-tuned based on experience around these candidate values. The mapping process compares the group cognitive convergence value calculated in step S5 with the determined low-convergence and high-convergence thresholds. If the group cognitive convergence value is less than the low-convergence threshold, it is mapped to the low-convergence level interval. If the group cognitive convergence value is greater than or equal to the low-convergence threshold and less than the high-convergence threshold, it is mapped to the medium-convergence level interval. If the group cognitive convergence value is greater than or equal to the high-convergence threshold, it is mapped to the high-convergence level interval. The mapping operation outputs a convergence level identifier, which is a string label used to identify the level interval to which it belongs.

[0040] The system maps group cognitive conflict levels to preset conflict level intervals. The input for this mapping operation is the group cognitive conflict level value calculated in step S5. The preset conflict level intervals are another set of numerical ranges predefined in the system configuration, used to classify continuous group cognitive conflict level values ​​into discrete levels. Group cognitive conflict level is the density of the dispute subgraph. The preset conflict level intervals are set to three intervals: low conflict level, medium conflict level, and high conflict level. The low conflict level interval is defined as ranging from 0 to a low conflict level threshold, where 0 is included in the interval, and the low conflict level threshold is not included. The medium conflict level interval is defined as ranging from a low conflict level threshold to a high conflict level threshold, where the low conflict level threshold is included in the interval, and the high conflict level threshold is not included. The high conflict level interval is defined as ranging from a high conflict level threshold to a theoretical upper limit value, where both the high conflict level threshold and the theoretical upper limit value are included. The theoretical upper limit value can be set to a sufficiently large number, such as 10. The specific values ​​for the low-conflict threshold and the high-conflict threshold are also set based on the analysis results of historical cases. The analysis method involves collecting interaction data from multiple historical communication activities and calculating the group cognitive conflict value for each historical activity. Domain experts are invited to qualitatively rate the intensity of disagreements and controversies in each historical activity based on actual records. The rating categories include low, medium, and high levels of controversy. Then, a set of group cognitive conflict values ​​corresponding to all historical activities rated as low by experts is created. The upper quartile of this set is calculated and used as a candidate value for the low-conflict threshold. Another set of group cognitive conflict values ​​corresponding to all historical activities rated as high by experts is created. The lower quartile of this set is calculated and used as a candidate value for the high-conflict threshold. The mapping process compares the group cognitive conflict value calculated in step S5 with the determined low-conflict threshold and high-conflict threshold. If the group cognitive conflict value is less than the low-conflict threshold, it is mapped to the low-conflict level range. If the group's cognitive conflict level is greater than or equal to the low conflict level threshold but less than the high conflict level threshold, it is mapped to the medium conflict level range. If the group's cognitive conflict level is greater than or equal to the high conflict level threshold, it is mapped to the high conflict level range. The mapping operation outputs a conflict level identifier, which is a string label used to identify the level range to which it belongs.

[0041] Match the preset propagation state matrix according to the combination of the convergence degree level interval and the conflict degree level interval to obtain the corresponding target scientific and cultural communication activity process status identifier. The inputs for performing this matching operation are the convergence degree level identifier and the conflict degree level identifier mapped in the previous step. The preset propagation state matrix is a two-dimensional lookup table data structure predefined in the system configuration. The row index of the propagation state matrix corresponds to all possible convergence degree level identifiers, the column index corresponds to all possible conflict degree level identifiers, and each cell in the matrix stores a unique target scientific and cultural communication activity process status identifier string. The content of the propagation state matrix is predefined based on communication theory and the induction and summary of a large number of historical activity patterns. The method of constructing the propagation state matrix is to combine the three levels of convergence degree with the three levels of conflict degree to obtain nine possible combined states. Define a status identifier string with a clear semantics for each of these nine combined states. For example, define the status identifier string corresponding to the combination where the convergence degree level identifier is high and the conflict degree level identifier is low as "healthy consensus formation". Define the status identifier string corresponding to the combination where the convergence degree level identifier is low and the conflict degree level identifier is high as "highly differentiated viewpoints". Define the status identifier string corresponding to the combination where the convergence degree level identifier is medium and the conflict degree level identifier is medium as "dynamic game evolution". The corresponding status identifier strings for the other six combinations are defined in a similar manner. The matching process uses the convergence degree level identifier as the row key and the conflict degree level identifier as the column key to perform a lookup in the propagation state matrix. The lookup operation accesses the cell corresponding to the intersection of the row key and the column key in the propagation state matrix and reads the status identifier string stored in that cell. The read status identifier string is the target scientific and cultural communication activity process status identifier.

[0042] Based on the status identifiers of the target science and culture dissemination activities, a feasibility assessment result containing status descriptions and quantitative indicators is generated. The inputs for this generation operation are the status identifiers of the target science and culture dissemination activities obtained in the previous step, the original group cognitive convergence value, and the group cognitive conflict value calculated in step S5. The generation process first retrieves detailed text descriptions corresponding to the status identifiers of the target science and culture dissemination activities from a pre-defined status description template library. The status description template library is a set of mapping relationships predefined in the system configuration, mapping each possible status identifier string to a pre-written, more detailed natural language description text. For example, for the status identifier string "healthy consensus formation," the mapped detailed description text might be: "The current dissemination activities show a good trend of consensus formation, core concepts are widely and deeply recognized, controversial topics are effectively controlled, the activity process is healthy, and feasibility is high." For the status identifier string "highly divergent viewpoints," the mapped detailed description text might be: "The current dissemination activities show a highly divergent viewpoints, consensus around core issues is difficult to form, and there are significant differences between different subgroups, requiring key intervention to guide dialogue." Then, the group cognitive convergence and group cognitive conflict values ​​calculated in step S5 are formatted to generate quantitative indicator text. Formatting includes determining the number of decimal places to retain for each value. The number of decimal places is determined based on the magnitude and precision requirements of the value, typically retaining three decimal places. For example, the group cognitive convergence value 0.825123 is formatted as the string 0.825, and the group cognitive conflict value 0.420456 is formatted as the string 0.420. The formatted values ​​are combined with the indicator names to form the quantitative indicator text. Finally, the obtained detailed description text is merged with the generated quantitative indicator text to form a complete text paragraph. This complete text paragraph is the final feasibility assessment result. The feasibility assessment result can be directly output to a display device for user viewing, or stored in a database or written to a report file.

[0043] Example 2: Figure 2 A schematic diagram of the feasibility assessment system for science and culture dissemination based on knowledge graphs according to the present invention is provided. The feasibility assessment system for science and culture dissemination based on knowledge graphs includes: The data acquisition module is used to acquire time-series interactive text data generated during the target science and culture dissemination activities; The knowledge graph construction module is used to extract entities and relationships from temporally interactive text data based on knowledge graph technology, and to construct a dynamic group knowledge graph composed of concept nodes and semantic relationship edges. The subgroup discovery module is used to discover communities in dynamic group knowledge graphs and identify multiple cognitive subgroups. The concept identification module is used to calculate the sub-community crossover and sub-community cohesion of each concept node in the dynamic group knowledge graph based on the cognitive sub-communities, and to identify the core consensus concept set and the questionable and divergent concept set. The measurement and calculation module is used to start from each concept node in the core consensus concept set, perform path traversal with a finite step length in the dynamic group knowledge graph to obtain the inference path set corresponding to each cognitive subgroup, calculate the group cognitive convergence based on the overlap between the inference path sets of different cognitive subgroups; induce the disputed subgraph from the questionable and divergent concept set, and calculate the group cognitive conflict degree based on the density of the disputed subgraph. The evaluation generation module is used to generate a feasibility assessment result for the progress of the target science and culture dissemination activities based on the degree of convergence and conflict of group cognition.

[0044] All calculations involved in the embodiments are dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0045] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0046] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions according to the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wireless or wired transmission; wired transmission methods include optical fiber, twisted pair, coaxial cable, etc.; wireless transmission includes infrared, microwave, etc. Computer-readable storage media can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0047] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0048] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0049] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0050] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0051] If a function is implemented as a software module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0052] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Finally: The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A feasibility assessment method for science and culture dissemination based on knowledge graphs, characterized in that: include: S1. Obtain the time-series interactive text data generated during the target science and culture dissemination activities; S2. Based on knowledge graph technology, entities and relations are extracted from temporal interactive text data to construct a dynamic group knowledge graph composed of concept nodes and semantic relation edges; S3. Community discovery is performed on dynamic group knowledge graphs to identify multiple cognitive sub-communities; S4. Based on the cognitive sub-communities, calculate the sub-community crossover and sub-community cohesion of each concept node in the dynamic group knowledge graph, and identify the core consensus concept set and the questionable and divergent concept set. S5. Starting from each concept node in the core consensus concept set, perform a path traversal with a finite step length in the dynamic group knowledge graph to obtain the inference path set corresponding to each cognitive subgroup. Calculate the group cognitive convergence based on the overlap between the inference path sets of different cognitive subgroups. Induce a disputed subgraph from the set of questionable and divergent concepts, and calculate the group cognitive conflict degree based on the density of the disputed subgraph. S6. Based on the convergence degree of group cognition and the degree of conflict of group cognition, generate a feasibility assessment result for the progress of the target scientific and cultural dissemination activities.

2. The feasibility assessment method for science and culture dissemination based on knowledge graphs according to claim 1, characterized in that, Acquire time-series interactive text data generated during target science and culture dissemination activities, including: Identify at least one interactive platform where the target science and culture dissemination activities take place; collect original interaction records from at least one interactive platform, the original interaction records including text content and timestamps; The text content in the original interaction record is cleaned to remove irrelevant characters and formatting noise, while retaining the timestamps. The cleaned and timestamp-retained text content is then organized into chronological interactive text data according to the timestamp order.

3. The feasibility assessment method for science and culture dissemination based on knowledge graphs according to claim 1, characterized in that, Based on knowledge graph technology, entity and relation extraction is performed on temporally sequential interactive text data to construct a dynamic group knowledge graph composed of concept nodes and semantic relation edges, including: Named entity recognition is performed on temporally interactive text data to extract multiple conceptual entities as candidate conceptual nodes; Relation extraction is performed on temporal interactive text data to identify semantic relationships between multiple conceptual entities as candidate semantic relationship edges; Based on multiple candidate concept nodes and candidate semantic relation edges, a dynamic group knowledge graph composed of concept nodes and semantic relation edges is generated, where each concept node corresponds to a concept entity and each semantic relation edge corresponds to a semantic relation.

4. The feasibility assessment method for science and culture dissemination based on knowledge graphs according to claim 1, characterized in that, Community discovery was performed on the dynamic group knowledge graph, identifying multiple cognitive sub-communities, including: Based on the weights of the semantic relationship edges between concept nodes in the dynamic group knowledge graph, the association strength of each pair of concept nodes is calculated. Based on the correlation strength between all concept nodes, the concept nodes are divided into different candidate sub-groups, such that the correlation strength between concept nodes within the same candidate sub-group is higher than the correlation strength between concept nodes in different candidate sub-groups. The candidate sub-communities obtained from the division are verified, and candidate sub-communities with overly sparse associations are merged or candidate sub-communities with overly loose internal structures are split to obtain multiple cognitive sub-communities.

5. The feasibility assessment method for science and culture dissemination based on knowledge graphs according to claim 1, characterized in that, Based on the sub-community crossover and sub-community cohesion of each concept node in the dynamic group knowledge graph calculated from cognitive sub-communities, the core consensus concept set and the questionable and divergent concept set are identified, including: For each concept node in the dynamic group knowledge graph, the number of cognitive sub-communities to which it belongs is counted as the sub-community span of the concept node; For each concept node, calculate the sum of the weights of all semantic relation edges between the corresponding concept node and other concept nodes in each cognitive subgroup to which it belongs, and use this sum as the internal association strength of the corresponding concept node in the corresponding cognitive subgroup. The maximum internal association strength is used as the subgroup cohesion of the concept node. Concept nodes whose sub-community crossover exceeds the first threshold and whose sub-community cohesion exceeds the second threshold are included in the core consensus concept set; concept nodes whose sub-community crossover falls below the third threshold and whose sub-community cohesion exceeds the fourth threshold are included in the questionable and divergent concept set.

6. The feasibility assessment method for science and culture dissemination based on knowledge graphs according to claim 5, characterized in that, in, The first threshold is greater than the third threshold, and the second threshold is greater than or equal to the fourth threshold.

7. The feasibility assessment method for science and culture dissemination based on knowledge graphs according to claim 1, characterized in that, Calculating the convergence of group cognition includes: For each concept node in the core consensus concept set, starting from the corresponding concept node, a breadth-first traversal with a preset step size is performed in the dynamic group knowledge graph. The sequence of concept nodes and semantic relationship edges passed during the traversal is recorded as a reasoning path. All reasoning paths obtained by the traversal are taken as the complete set of reasoning paths corresponding to the corresponding concept node. For each cognitive subgroup, inference paths whose starting point and all concept nodes in the path belong to the corresponding cognitive subgroup are selected from the complete set of inference paths, thus forming the inference path set of the corresponding cognitive subgroup. Calculate the Jaccard similarity coefficient between the sets of reasoning paths of any two different cognitive subgroups, and take the average of all Jaccard similarity coefficients to obtain the group cognitive convergence.

8. The feasibility assessment method for science and culture dissemination based on knowledge graphs according to claim 1, characterized in that, Calculating the degree of group cognitive conflict includes: Extract a subgraph from the dynamic group knowledge graph that contains all concept nodes and all direct semantic relationship edges between them in the set of questionable and divergent concepts as the disputed subgraph; calculate the ratio of the number of semantic relationship edges to the number of concept nodes in the disputed subgraph to obtain the density of the disputed subgraph as the group cognitive conflict degree.

9. The feasibility assessment method for science and culture dissemination based on knowledge graphs according to claim 1, characterized in that, Based on the convergence and conflict of group cognition, a feasibility assessment of the progress of the target science and culture dissemination activities is generated, including: Map the convergence of group cognition to a preset convergence level range, and map the conflict level of group cognition to a preset conflict level range; Based on the combination of convergence level intervals and conflict level intervals, a preset propagation state matrix is ​​matched to obtain the corresponding target scientific and cultural dissemination activity process status identifier. Based on the status identifiers of the target scientific and cultural dissemination activities, a feasibility assessment result containing status descriptions and quantitative indicators is generated.

10. A knowledge graph-based feasibility assessment system for the dissemination of science and culture, used to implement the knowledge graph-based feasibility assessment method for the dissemination of science and culture as described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to acquire time-series interactive text data generated during the target science and culture dissemination activities; The knowledge graph construction module is used to extract entities and relationships from temporally interactive text data based on knowledge graph technology, and to construct a dynamic group knowledge graph composed of concept nodes and semantic relationship edges. The subgroup discovery module is used to discover communities in dynamic group knowledge graphs and identify multiple cognitive subgroups. The concept identification module is used to calculate the sub-community crossover and sub-community cohesion of each concept node in the dynamic group knowledge graph based on the cognitive sub-communities, and to identify the core consensus concept set and the questionable and divergent concept set. The measurement and calculation module is used to start from each concept node in the core consensus concept set, perform path traversal with a finite step length in the dynamic group knowledge graph to obtain the inference path set corresponding to each cognitive subgroup, calculate the group cognitive convergence based on the overlap between the inference path sets of different cognitive subgroups; induce the disputed subgraph from the questionable and divergent concept set, and calculate the group cognitive conflict degree based on the density of the disputed subgraph. The evaluation generation module is used to generate a feasibility assessment result for the progress of the target science and culture dissemination activities based on the degree of convergence and conflict of group cognition.

Images