Computer teaching resource management system based on cloud computing

Abstract

The invention relates to the technical field of resource management, in particular to a computer teaching resource management system based on cloud computing, which comprises a pair group mismatch screening module, a semantic boundary recognition module, a keyword semantic judgment module, a keyword classification management module and a keyword optimization updating module. According to the method, the verb and noun pairing combination in the courseware text is subjected to structured classification, and the affiliation relationship is compared by combining the teaching task vocabulary set, so that the semantic mismatched content combination can be accurately recognized, and the integrating degree of the courseware content and the teaching target is improved. After the distribution trend of the mismatch combination is subjected to numerical calculation, incremental change sections of the mismatch combination are recognized through clustering, the boundary position of semantic jump in the content can be effectively perceived, and stable reference is provided for subsequent semantic analysis. By means of judgment of the position of the keyword in the statement structure and the relation between the keyword and the trunk verb, the logical coherence of semantic evolution of the keyword is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of resource management, and in particular to a computer teaching resource management system based on cloud computing. Background Art

[0002] The field of resource management encompasses technologies for the orderly configuration and scheduling of information, hardware, and software resources in computer systems. Core areas within this area include data storage management, access control, load balancing, resource allocation, and sharing strategies, aiming to improve resource utilization efficiency and optimize overall system performance.

[0003] Among them, the computer teaching resource management system based on cloud computing refers to a system that uses cloud computing architecture to uniformly manage various resources required for computer teaching, including the storage structure of teaching resources, classification standards of teaching data, establishment of user access policies and setting of resource distribution paths.

[0004] Existing cloud computing-based teaching resource management methods mainly focus on the unified storage, access control and distribution path setting of resources, but lack in-depth capabilities in the intelligent analysis and semantic level control of the internal structure of teaching content. Teaching resources are mostly text courseware. Traditional methods do not make semantic attribution judgments on the verb-noun collocation relationship in the text content, which easily leads to unrecognized semantic deviations between teaching task objectives and courseware content, affecting the precise matching of teaching resources. In addition, there is a lack of recognition mechanism for structural changes or semantic transitions in teaching content, and it is impossible to clearly distinguish the logical boundaries between different teaching content segments, which makes keyword classification management face significant ambiguity. Keyword recognition often relies on word frequency and rule matching, ignoring contextual relationships and syntactic dependency information, resulting in confusion between functional keywords and modifying content, reducing the accuracy of semantic analysis. For example, when polysemous words appear repeatedly in courseware, they are often mistakenly classified as the same type of function, disrupting the teaching logic. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and propose a computer teaching resource management system based on cloud computing.

[0006] In order to achieve the above objectives, the present invention adopts the following technical solution: a computer teaching resource management system based on cloud computing, the system comprising: The pair mismatch screening module extracts verb and noun combinations from the courseware text content from the cloud and classifies them. It then screens the pair mismatch rate sequence by detecting whether each combination appears in the target vocabulary set of the specified teaching task. The semantic boundary recognition module performs increasing trend clustering recognition on the pair mismatch rate sequence by using a fuzzy C-means clustering algorithm, and marks a set of semantic boundary transition points of the courseware; The keyword semantic determination module identifies the occurrence position of keyword identifiers in the courseware text content through cloud semantic analysis, and determines the semantic conversion state of keywords based on the courseware semantic boundary transition point set, and organizes them into a keyword function switching list; The keyword classification management module performs boundary classification judgment on the keyword context structure feature value vector based on the keyword function switching list through the support vector machine algorithm to obtain the courseware keyword management result.

[0007] The present invention has the following improvements: the group mismatch rate sequence includes a semantic inconsistency group number, a paragraph index number, and a mismatch rate change value; the courseware semantic boundary transition point set specifically includes a position index, a clustering label type, and a mismatch trend direction; the keyword function switching list includes a function switching identifier, a semantic role comparison item, and a structural position mapping group; and the courseware keyword management result includes a boundary classification result mark, a valid keyword set index, and a context structure feature vector number.

[0008] The present invention is improved in that the pair mismatch screening module includes: The phrase extraction submodule obtains all sentences in the courseware text, identifies and extracts verbs with action characteristics and corresponding noun combinations in the sentences, forms paired items according to syntactic dependency rules, and generates action-object group constituent sets based on the semantic combination of the extracted paired items; The attribution detection submodule detects whether each object belongs to the preset teaching task target vocabulary set based on the action-object group set, and uses the coverage ratio and correlation calculation method of the object content in the combination and the target vocabulary set to perform attribution judgment, thereby obtaining the object attribution relationship judgment set; The sequence construction submodule counts the number of unattributed combinations and the total number of combinations according to the object attribution relationship determination value set, calculates the ratio of the combinations and arranges them in structural order to generate a pair mismatch rate sequence.

[0009] The present invention is improved in that the semantic boundary recognition module includes: The trend calculation submodule extracts the mismatch rate value of each of three consecutive content segments in the mismatch rate sequence of the pair group, calculates the mismatch rate increase between the two segments before and after, and determines whether the increase directions of the three segments are in the same direction. The set of segments with the same increase direction is selected as the trend area input item to generate a set of continuous increasing trend segments; The fuzzy classification submodule extracts the mismatch rate sequence corresponding to each segment in the continuous increasing trend segment set, sets the initial number of cluster centers and the fuzzy coefficient, and calls the fuzzy C-means clustering algorithm to perform membership division on the mismatch rate values to obtain the mismatch rate membership cluster label; The boundary annotation submodule determines whether the first position of the mismatch category segment where the cluster label appears continuously coincides with the position of the continuous increasing trend segment based on the mismatch rate affiliation cluster label. If so, the target position is added with a boundary marker and recorded as a courseware content segment index to obtain the courseware semantic boundary transition point set.

[0010] The present invention is improved in that the keyword semantic determination module includes: The position extraction submodule obtains all sentence structures in the courseware text content, screens the keyword content including multi-frequency information components, combines the part-of-speech tagging rules with the context grammatical structure to determine whether the keyword independently plays a semantic pointing role, locates the position index in the content fragment, and generates the keyword identification positioning result; The role identification submodule collects the first verb information and subject-verb combination items of the corresponding sentence based on the keyword identification and positioning results, and determines the semantic function of the keyword in the sentence based on the verb features, judging whether it serves as a functional instruction action or an annotation modification content, and calculates the semantic deviation direction based on the stability of the word order structure to obtain the keyword function role classification result; The function conversion submodule is based on the keyword function role classification results and the courseware semantic boundary transition point set, and screens whether the keyword crosses the boundary area and undergoes a semantic role change, and determines whether the function label classification in the previous and next areas is inconsistent. If the role category is inconsistent, it is marked as a conversion item and a keyword function switching list is established.

[0011] The present invention is improved in that the keyword classification management module includes: The feature construction submodule extracts the context sentences of each keyword in the courseware content in the keyword function switching list, calls syntactic dependency rules and position information to extract adjacent phrases, word order depth and sentence structure indicators, and constructs the indicator combination into a feature vector to generate a context structure feature value vector set; The boundary calculation submodule sets the boundary threshold and penalty coefficient according to the keyword function switching list and the context structure feature value vector set, classifies the samples by maximizing the interval using the support vector method, calculates the belonging boundary of each keyword between the switching areas, and the belonging boundary includes the continuous change class or the stable continuation class, and obtains the keyword boundary discrimination result; The classification and screening submodule identifies all keyword identifiers whose boundary categories are stable and continuous categories based on the keyword boundary discrimination results, relocates and collects the keywords according to their unique codes in the original text, and establishes courseware keyword management results.

[0012] The present invention is improved in that it further includes a keyword optimization and updating module, which determines the content consistency between the keywords in the courseware keyword management results and the corresponding semantic structures, adjusts the keyword identifiers according to the determination results, obtains the courseware keyword optimization results and stores them in the cloud database; The courseware keyword optimization result includes the corrected keyword identifier, the corresponding semantic structure identifier, and the cloud database storage path.

[0013] The present invention is improved in that the keyword optimization and updating module includes: The consistency determination submodule obtains the courseware keyword management results and the action-object group composition set, extracts the verb and object combination items corresponding to the keyword sentence, determines whether the keyword appears in the object item or forms a logical association with the verb, and generates a keyword pointing matching result; The identifier adjustment submodule filters out keyword identifiers that do not establish a verb or object semantic connection based on the keyword pointing matching results, reconstructs the position numbers and updates the semantic tags of the remaining keywords, and obtains an adjusted keyword identifier set; The data storage submodule extracts the keyword content, position index, and corresponding segment number fields based on the adjusted keyword identification set to form a structured record format, writes it into the cloud database under the specified path, and establishes the courseware keyword optimization results.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by structurally classifying the verb and noun pairing combinations in the courseware text and comparing their attribution relationships in combination with the teaching task vocabulary set, the content combination with semantic mismatch can be accurately identified, and the degree of fit between the courseware content and the teaching objectives can be improved. After the distribution trend of the mismatch combination is numerically calculated, its incremental change segment is identified by clustering, and the boundary position of the semantic jump in the content can be effectively perceived, providing a stable reference for subsequent semantic analysis. By judging the position of the keyword in the sentence structure and its relationship with the main verb, its semantic role is clarified, and combined with the changing trend of the semantic conversion behavior before and after the boundary, the function switching information is extracted, and the logical coherence of the keyword semantic evolution is strengthened. Further, feature vectors are extracted based on the context dependency structure, and the boundary attribution of the keyword in the semantic conversion area is numerically judged using a support vector machine, which can achieve an optimal balance between discrimination accuracy and generalization ability, and construct a keyword judgment mechanism that can be used for teaching resource management. Finally, by comparing the consistency of semantic structures, correcting keywords that deviate from the teaching semantic goals, and realizing structured storage processing, a dynamic update mechanism is formed, which not only improves the accuracy of keyword recognition, but also enhances the organization and retrieval efficiency of semantic data. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a system module diagram of the present invention; Figure 2 It is a system framework diagram of the present invention; Figure 3 Schematic diagram of a group mismatch screening module according to the present invention; Figure 4 Schematic diagram of the semantic boundary recognition module of the present invention; Figure 5 Schematic diagram of a keyword semantic determination module of the present invention; Figure 6 Schematic diagram of the keyword classification management module of the present invention; Figure 7 Schematic diagram of the keyword classification management module of the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0017] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings and are only for the convenience of describing the present invention and simplifying the description. They do not indicate or imply that the devices or elements referred to must have a specific direction, be constructed and operate in a specific direction, and therefore should not be understood as limiting the present invention. In addition, in the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0018] See also Figure 1 The present invention provides a technical solution: a computer teaching resource management system based on cloud computing, the system includes: The pair mismatch screening module extracts verb and noun combinations from the courseware text content from the cloud and classifies them. It then screens the pair mismatch rate sequence by detecting whether each combination appears in the target vocabulary set of the specified teaching task. The semantic boundary recognition module uses the fuzzy C-means clustering algorithm to perform increasing trend clustering recognition on the mismatch rate sequence of the group, marking the semantic boundary transition point set of the courseware; The keyword semantic determination module uses cloud-based semantic analysis to identify the occurrence locations of keyword identifiers in the courseware text content, and combines the courseware semantic boundary transition point set to determine the semantic conversion status of the keyword, and organizes it into a keyword function switching list; The keyword classification management module uses the support vector machine algorithm to perform boundary classification judgment on the keyword context structure feature value vector based on the keyword function switching list to obtain the courseware keyword management results; The group mismatch rate sequence includes the semantic inconsistency group number, paragraph index number, and mismatch rate change value. The courseware semantic boundary transition point set is specifically the position index, cluster label type, and mismatch trend direction. The keyword function switching list includes the function switching identifier, semantic role comparison item, and structural position mapping group. The courseware keyword management results include the boundary classification result mark, the effective keyword set index, and the context structure feature vector number.

[0019] See also Figure 2 and Figure 3 , the group mismatch screening module includes: The phrase extraction submodule obtains all sentences in the courseware text, identifies and extracts verbs with action characteristics and corresponding noun combinations in the sentences, forms paired items according to syntactic dependency rules, and generates action-object group constituent sets based on the semantic combination of the extracted paired items; Obtain all sentences in the courseware text, perform action feature recognition and matching item extraction for each sentence, first segment the courseware text according to the natural language processing sentence segmentation rules, for example, divide a page of teaching courseware into sentences such as "click the button to open the browser", "enter the URL in the browser and press Enter", and then identify the verbs. Among them, "click", "input", and "enter" all have action features. It is necessary to use part-of-speech tagging tools, such as Jieba or LTP, to identify the part of speech of the words, and combine the verb with its nearest noun into an action-object combination, such as "click-button", "input-URL", and "enter-browser". At this time, it is necessary to confirm the pairing of verbs and objects based on the dependency syntactic structure. The VOB (verb-object structure) in the dependency arc structure is used for filtering and matching. If the sentence structure is "verb-preposition-noun", the verb is extracted and the preposition is skipped to directly connect the noun to form a combination. Then all action-object combinations are semantically merged to avoid synonyms or near-synonyms affecting the consistency of the combination. For example, "start-application" and "open-software" are merged into "execute-program" according to the semantic similarity. After the processing is completed, all action-object combinations are formed into an action-object group set. Assuming that a courseware clip contains 20 sentences, 28 valid combinations are formed after processing. The action-object group set output by this submodule contains these 28 groups of combination information. The attribution detection submodule detects whether each object belongs to the preset teaching task target vocabulary set based on the action-object group set, and uses the coverage ratio and correlation calculation method of the object content in the combination and the target vocabulary set to make attribution judgments, thus obtaining the object attribution relationship judgment set; According to the action-object group composition set, the object noun is matched with the preset task target vocabulary set. The task target vocabulary set can be composed of keywords extracted from the teaching design objectives, such as "browser", "input box", "settings", and "search bar". After each object noun is extracted from the combination, it needs to be normalized, unified in capitalization, stemmed, and disambiguated. Then, the vocabulary coverage ratio of each noun to the target vocabulary set is calculated using the formula , where the parameters is the number of object words found in the target vocabulary in combination i, obtained by looking up the table word by word for the combination object and matching it with the target vocabulary entry. is the total number of object words in combination i, which is obtained by counting the original word counts at the object end in the combination. The value is calculated as the matching ratio, and the result is used to determine the attribution strength. A second-level screening is then performed based on the semantic relevance to calculate the average cosine similarity between the object word vector and the target set word vector. The parameter is the average dot product similarity between the object word vector and the target set word vector. The word vector is generated by a pre-trained model such as Word2Vec. If the value is greater than the similarity benchmark value of 0.6, it is judged to be an attributable item, otherwise it is an unattributed item. For example, if the object is "web page" and the average similarity with target words such as "browser", "website", and "page" is 0.72, it is classified as a hit item. The results are recorded and an object attribution relationship judgment set is constructed, which includes the combined index, matching ratio and attribution label.

[0020] The sequence construction submodule determines the value set based on the object ownership relationship, counts the number of unattributed combinations and the total number of combinations, calculates the ratio of the combinations and arranges them in structural order to generate a pair mismatch rate sequence; According to the object ownership relationship judgment value set, the number of unattributed combinations and the total number of combinations in each content segment are counted, and then the mismatch rate is calculated based on this. The formula , where the parameters represents the number of unattributed combinations in the jth segment. It is obtained by filtering the combinations with the attribution label of “no” in the decision set and counting them. Indicates the total number of combinations in the fragment, which is obtained by counting the corresponding number of combinations according to the fragment index. is the mismatch rate of the segment, which indicates the proportion of failed attribution. Assuming that the 4th segment contains 9 combinations, 4 of which are not attributable, then , and so on, traverse all the paragraphs in the courseware, and arrange the mismatch rate values of each paragraph in the order of the paragraphs to form a pair of mismatch rate sequences. If the courseware contains 12 paragraphs, then output a set of mismatch rate numerical sequences with a length of 12.

[0021] See also Figure 2 and Figure 4 , the semantic boundary recognition module includes: The trend calculation submodule extracts the mismatch rate values of every three consecutive content segments in the mismatch rate sequence, calculates the mismatch rate increase between the two segments, and determines whether the increase directions of the three segments are in the same direction. It selects the set of segments with the same increase direction as the trend area input item to generate a set of continuous increasing trend segments. Extract the mismatch rate value of each three consecutive content segments in the mismatch rate sequence of the pair group. The acquisition method is to read the mismatch rate value generated by each segment in the previous module in the order of content arrangement. For example, form a combination for the first to third segments, and then slide the window downward in sequence. For the first two segments in each combination, extract their mismatch rate values and calculate their difference as the increase. If the mismatch rate of the previous segment is 0.22 and the next segment is 0.27, the corresponding increase is 0.05. To perform this operation, it is necessary to traverse the segments in the entire courseware that have generated mismatch rates. Use The formula is used to calculate the difference. For the The mismatch rate of the segment; For the Paragraph to The increase between the two segments is the difference between the mismatch rates of the two segments, and the unit is scalar. Continue to obtain the increase of the next two segments and judge whether the directions of the two increases are consistent. It can be determined that the three segments' mismatch rates show a continuously increasing numerical value, thus forming a segment with a valid trend. During execution, the positions of all segments meeting this logical condition in the original content should be recorded as the input source for the trend segment. The data recording format should retain the segment number and its corresponding increment value. After the traversal is complete, all segments that meet the conditions are constructed into a set, which will be used as the input source for the subsequent fuzzy partitioning to generate a set of continuously increasing trend segments.

[0022] The fuzzy classification submodule extracts the mismatch rate sequence corresponding to each segment in the continuous increasing trend segment set, sets the initial number of cluster centers and the fuzzy coefficient, and calls the fuzzy C-means clustering algorithm to perform membership division on the mismatch rate values to obtain the mismatch rate membership cluster label; Extract the mismatch rate value corresponding to each segment in the continuous increasing trend segment set. First, normalize all mismatch rate values in the segment using the minimum and maximum normalization method. The formula is: ; in, is the original mismatch rate value, is the calculated Value, for example, the original mismatch rate of the continuous segment is 、 、 ,but , , normalize each value to get , , .

[0023] Then set the initial number of cluster centers to , which means dividing the data into two categories: "slowly changing class" and "dramatically changing class", and the fuzzy coefficient is set to , as a soft clustering strength adjustment parameter.

[0024] The calculation formula of construction membership is as follows: ; in, Indicates the The sample points are The membership degree of the cluster center ranges from [0, 1]; It is Normalized mismatch rate sample values, such as , , ; Indicates the The value of the cluster center is initially set to , , take the boundary value in the normalized value; is the number of cluster categories, here it is 2; is the fuzziness coefficient, set to 2; represents the Euclidean distance.

[0025] Sample points As an example, calculate its membership to two centers: ; ; At this time, the sample point can be determined The membership of is more inclined to the second category (dramatic change category).

[0026] The cluster center update formula is: ; in: Indicates the updated Cluster center; is the membership degree calculated previously; is the normalized value of the sample point; is the number of sample points in the current segment, which is 3 here.

[0027] To update Category 2 centers For example, assuming that the square of the membership degree of the aforementioned sample point is , , , the normalized values are 0, 0.45, and 1 respectively, then: .

[0028] The process continues to iterate until the change in the two cluster centers is less than the set error threshold. After all samples are clustered, a set of category labels for all samples in the trend segment is generated as the output mismatch rate belonging cluster label.

[0029] Membership calculations are used to identify semantic boundaries where mismatch rates vary within courseware content. First, semantic attribution is determined for verb-noun pairings within the courseware text. Mismatch rate values are extracted for every three consecutive content segments, their changing trends analyzed, and increasing trend segments identified. The system normalizes the mismatch rate data within these trend segments, then sets initial cluster centers (typically the normalized results corresponding to the maximum and minimum values). The system also determines the number of clusters to be two: a "slowly changing" cluster and a "rapidly changing" cluster. A fuzzy coefficient is also set to adjust the softness or hardness of the clustering. The system then iteratively applies the fuzzy C-means clustering algorithm to the normalized mismatch rate data, calculating the membership of each data point to each cluster center, measuring its proximity to each category, and updating the cluster center accordingly. This process continues until the cluster centers converge. Finally, each mismatch rate value is assigned a membership label, indicating its corresponding change pattern within the trend segment. The distribution of these membership labels is then aligned with the starting position of the trend segment in the mismatch rate sequence. If there is consistency, the position is confirmed as the boundary point of the semantic transition.

[0030] The boundary annotation submodule determines whether the first position of the mismatch category segment where the cluster label appears continuously coincides with the position of the segment with a continuous increasing trend according to the mismatch rate. If so, the target position is marked with a boundary marker and recorded as the courseware content segment index to obtain the courseware semantic boundary transition point set. Based on the mismatch rate affiliation cluster label, the label sequence of all samples in each trend segment is continuously searched. When executing the operation, it is necessary to first locate the starting position index of each segment in the courseware content. Then, based on the affiliation category corresponding to each normalized mismatch rate sample point, the sample index range of continuous clusters belonging to the same category (such as the drastic change category) is extracted from the label sequence. It is determined whether the cluster label appears in three or more consecutive segments in the current segment. If the number of consecutive segments meets this standard, it is further determined whether the position of its first segment overlaps at the starting position of the continuously increasing trend segment. The overlap determination method here is the consistency comparison of the segment number. That is, if the continuous starting position number of the cluster label is equal to the first number of the trend segment, it is considered to meet the conditions. For example, if the 5th, 6th, and 7th segments are determined to be of the same type of label, and the trend segment also starts from the 5th segment, then this area is determined to be the area where the semantic boundary jump occurs. On this basis, the beginning of the paragraph is used as the boundary point and a boundary marker is added. The corresponding position of the courseware content fragment needs to be recorded as a segment number and cluster category, and all boundary point positions that meet the conditions are collected and numbered. Finally, a structured record is formed and uniformly output as a courseware semantic boundary transition point set.

[0031] See also Figure 2 and Figure 5 , the keyword semantic determination module includes: The position extraction submodule obtains all sentence structures in the courseware text content, screens the keyword content including multi-frequency information components, combines the part-of-speech tagging rules with the context grammatical structure to determine whether the keyword independently plays a semantic pointing role, locates the position index in the content fragment, and generates the keyword identification positioning result; Obtain all sentence structures in the courseware text content. First, the text is divided into sentences, and a combination of punctuation separation and length window control is used to ensure that each sentence is a basic unit with semantic integrity. Then, the word frequency of all words in the sentences is counted and sorted, and the words that appear frequently in the courseware content are identified as candidate keywords. These words need to be further filtered according to the part-of-speech tagging rules. Only words with independent semantic pointing characteristics such as nouns, gerunds, and proper words are retained. Commonly used tools are jieba or LTP. According to the syntactic dependency structure, it is determined whether the word appears in the main component position, such as subject, object, predicate complement, etc., and whether it constitutes the semantic focus of the sentence. Then, combined with its contextual grammatical structure, it is determined whether it plays an independent semantic role. If the word appears in multiple sentence patterns and is positioned in a key semantic role, it is determined to be a keyword, and the index position of the content segment where it is located is located. It is recorded in a structured format consisting of number, word, period position and semantic position to generate a keyword identification and positioning result.

[0032] The role identification submodule collects the first verb information and subject-verb combination items of the corresponding sentence based on the keyword identification and positioning results. It then determines the semantic function of the keyword in the sentence based on the verb features, judging whether it serves as a functional instruction action or an annotation modification content. It also calculates the semantic deviation direction based on the stability of the word order structure to obtain the keyword function role classification result. Based on the keyword identification and positioning results, the sentence-initial verb and subject-verb combination of each keyword are extracted to determine whether the sentence structure containing the keyword is an action-driven expression. The extraction method is to perform dependency syntactic analysis on the target sentence, construct a verb dependency structure diagram, and screen the pairing relationship between the verb and the keyword in the subject-verb-object structure at the beginning of the sentence. For example, in "Click the button to open the browser", "click" is the action verb and "button" is the keyword. The two form an subject-verb pairing and are marked as functional instructions. Conversely, in "Browsers are usually used to open web pages", "browser" is a keyword, but it forms a non-direct action structure with "used for", which is identified as a modified explanatory expression. Then, the word order structure of the keyword and the surrounding syntactic nodes in the sentence is analyzed to extract the changes in the order of its semantic focus. The stability ratio of the keyword's position is calculated. If the keyword's sentence position in multiple sentences is shifted left and right, it is considered a semantic shift trend. The functional role classification labels and shift directions of all keywords in the current sentence are sorted and output as the keyword functional role classification results.

[0033] The function conversion submodule is based on the keyword function role classification results and the courseware semantic boundary transition point set. It screens whether the keyword crosses the boundary area and undergoes a semantic role change, and determines whether the function label classification in the previous and next areas is inconsistent. If the role category is inconsistent, it is marked as a conversion item and a keyword function switching list is established. Based on the keyword function role classification results and the courseware semantic boundary transition point set, all keyword annotation results and boundary point indexes are cross-checked, and the keywords appearing before and after the boundary are positioned and matched according to the period index and paragraph attribution to determine whether the role label switches before and after the boundary crossing. For example, the keyword "browser" is a functional instruction class in the 5th paragraph and a reference modification class in the 6th paragraph, and there is a boundary transition point between the two paragraphs, which is considered to have occurred a semantic role switch. Then, based on the role switching record, a set consisting of the keyword and its position segment index and role change mark is constructed to determine whether the role label changes from "instruction class" to "explanation class" or vice versa. If the role label is inconsistent before and after, the system marks it as a switching keyword, and numbers and classifies all such data. The final output structured result is a keyword function switching list.

[0034] See also Figure 2 and Figure 6 , the keyword classification management module includes: The feature construction submodule extracts the context sentences of each keyword in the courseware content in the keyword function switching list, uses syntactic dependency rules and position information to extract adjacent phrases, word order depth and sentence structure indicators, and constructs the indicators into a feature vector to generate a context structure feature value vector set; To extract the contextual sentences for each keyword in the courseware content, the system first extracts the complete sentence containing the keyword from the original text based on the content segment number and sentence position index in the keyword tag. The context window is then expanded to one adjacent sentence forward and one adjacent sentence backward. For each term in the resulting contextual sentence group, a part-of-speech tagging model is used and syntactic dependency analysis is performed to identify dependency paths between the keyword and other terms. For example, in the sentence "browser is used to access web pages," the keyword "browser" forms a subject-verb dependency (SBV) with the verb "used for," while the verb forms a verb-object dependency (VOB) with "access web pages." The system records the dependency edge type as the first structural indicator. The system then analyzes the keyword's hierarchical depth in the syntactic tree, expanding the tree structure and counting the number of edges from the keyword node to the root node, i.e., the word order depth. If "browser" requires three nested levels from the root node, its word order depth is 3. The grammatical structure type is classified according to the overall structure of the sentence, and the sentence structure is judged to be imperative, interrogative, declarative or parallel. For example, "Please open the browser" is judged to be an imperative sentence, and "Do you need to set the homepage?" is judged to be an interrogative sentence. This structural classification result is converted into a coding form, such as 1 for imperative sentences, 2 for interrogative sentences, and 3 for declarative sentences, as the third type of feature. All structural features - dependency edge type, word order depth, grammatical structure category - are unified and integrated into a vector form to construct the feature items of a single keyword in the current context, such as the vector form is , , respectively representing the subject-verb dependency weight of 0.6, the word order depth of 3, and the sentence structure of an imperative sentence. After each keyword feature vector is formed, its keyword number and source sentence index number are recorded, and all keywords are processed in batches. The vector set generated is summarized to form the context structure feature value vector set. The boundary calculation submodule sets the boundary threshold and penalty coefficient based on the keyword function switching list and the context structure feature value vector set. It uses the support vector method to maximize the interval classification of samples and calculates the belonging boundary of each keyword between the switching areas. The belonging boundary includes continuous change class or stable continuation class, and obtains the keyword boundary discrimination result. According to the keyword function switching list and the context structure eigenvalue vector set, the support vector machine classification input matrix is constructed. First, the boundary discrimination threshold is set. This threshold is used to determine whether a keyword has switched its semantic role. The setting basis is the median offset of the structure vector score distribution in a large number of samples in the system. By statistically analyzing the scores of keywords marked as semantic transition points in the early training samples, it is found that the average boundary values are concentrated in the range of 0.08 to 0.12. Finally, the median of the interval is selected as the critical line, and the value is set to 0.1.

[0035] Weight parameters used for support vector machine classification The three feature dimensions corresponding to the structure vector are dependency feature, word order depth and sentence structure encoding. Set to the maximum, based on the fact that the dependency relationship between keywords and central verbs is the primary basis for judging semantic switching. Changes in the dependency structure will directly lead to changes in the semantic role of keywords, so it is given dominance; weight Used to measure the depth of a word in the syntactic tree. It has a moderate impact on semantic stability. Keywords at deeper positions in a sentence are more likely to be covered by modifying expressions. It is given a negative effect because over-coding of sentence structure may cause category dilution and have no positive impact on the judgment boundary. Therefore, a negative coefficient is introduced as an adjustment item to balance the fluctuation of the overall judgment model.

[0036] Constructing a linear boundary function as follows: ; in: For the Structural feature vector of keywords; The dependency weight between keywords and verbs ranges from 0 to 1 and is obtained by using a syntactic analysis tool to calculate dependency paths and assign weights. The depth of the keyword in the syntax tree, which represents the structural distance between it and the root node of the sentence, is obtained by traversing the syntax tree path; To code the sentence structure, the coding scheme was: declarative sentence = 1, imperative sentence = 2, interrogative sentence = 3, parallel sentence = 4; is the discriminant weight corresponding to each dimension; is a bias term used to adapt to the distribution offset between multi-class structure samples.

[0037] Assume that sample 1 is the keyword "browser", and the feature vector is , , ,but: .

[0038] Sample 2 is "page", the vector is , , ,but: .

[0039] Sample 3 is "URL", the vector is , , ,but: .

[0040] Based on the set threshold , the judgment rules are as follows: like , determined to be "continuous change type"; like , determined to be "stable and continuous"; Sample 1 has a boundary value of 0.1, which falls into the "continuously changing" category; sample 2 has a boundary value of 0.54, which is higher than the threshold and is also classified as "continuously changing"; and sample 3 has a boundary value of 0.14, also classified as "continuously changing." This calculation method is used to generate boundary scores and labels for all keyword samples, and the final output is the keyword boundary discrimination result.

[0041] The classification and screening submodule identifies the keyword identifiers of all boundary categories as stable and continuous categories based on the keyword boundary discrimination results, relocates and collects the keywords according to their unique codes in the original text, and establishes the courseware keyword management results; According to the boundary discrimination results of keywords, the boundary discrimination values and classification labels of all keyword samples are traversed. First, the corresponding keyword number, discrimination value and boundary category label in each record are read, and the keywords judged as "stable continuity type" are identified according to the set rules, that is, the boundary score Less than the boundary threshold After the recognition is completed, these keywords are accurately traced back and located based on their unique codes in the original text. The code contains the fragment number, sentence sequence number and word position index of the keyword in the text. The system re-extracts the original text position and the corresponding sentence pattern based on the code, organizes its structural information together with the keyword name into a structured data format, and uniformly marks it as a "confirmed" state. Then all keywords belonging to the stable category are uniformly collected and summarized, and a keyword object table is uniformly constructed for their structural vectors, position information and labels to ensure that each sample uniquely corresponds to a record, and the results are output in sequence as courseware keyword management results.

[0042] See also Figure 2 and Figure 7 , further comprising a keyword optimization and updating module, which determines the content consistency between keywords and corresponding semantic structures in the courseware keyword management results, adjusts the keyword identifiers according to the determination results, obtains the courseware keyword optimization results and stores them in the cloud database; The courseware keyword optimization results include the corrected keyword identifier, the corresponding semantic structure identifier, and the cloud database storage path; Keyword optimization update module includes: The consistency judgment submodule obtains the courseware keyword management results and the action-object group composition set, extracts the corresponding verb and object combination items in the sentence where the keyword is located, determines whether the keyword appears in the object item or forms a logical association with the verb, and generates the keyword pointing matching result; The courseware keyword management results and action-object group structure are obtained. For each keyword sample, the sentence containing the keyword is extracted and the corresponding action-object combination structure is identified. Specifically, the system first locates the sentence fragment of the keyword in the courseware text based on the keyword encoding. Then, the verb and its associated noun component are searched within the same sentence, corresponding to the action and object. For example, in "Click the button to open the browser," "browser" is the keyword, "open" is the verb, and "browser" is the object. Based on the syntactic dependency relationship, the system confirms that the two are a verb-object structure. It then determines whether the keyword appears in the object item or forms a syntactic dependency path with the verb (e.g., VOB, SBV, IOB, etc.). If the keyword neither appears in the object part nor forms a direct structural connection with the verb, it is marked as having no semantic directionality. This operation traverses all keywords and performs structural comparison. The matching status and matching path are recorded one by one, forming a matching label ("hit" or "missing"), which is summarized as the keyword-directed matching result.

[0043] The identifier adjustment submodule filters out keyword identifiers that do not establish verb or object semantic connections based on keyword pointing matching results, reconstructs the position numbers and updates the semantic tags of the remaining keywords, and obtains an adjusted keyword identifier set; According to the keyword pointing matching results, first filter out all keyword entries marked as "missing", that is, samples that do not form a logical connection with actions or objects in the sentence structure, mark them as invalid and eliminate them, and then reconstruct and number the remaining keyword samples according to their position index. The specific processing process is to extract the text position, sentence number and word position index of each keyword, compare them with the vacancies caused by the eliminated items, renumber them in the order of the fragments, and add a semantic role identification field to its data field for subsequent functional differentiation. The semantic tag is assigned according to the matching path type identified in the pointing result, such as "action target", "object extension", etc. After completing the identification update of all retained items, all fields are summarized into a unified structure item, and the output is the adjusted keyword identification set.

[0044] The data storage submodule extracts the keyword content, position index, and segment number fields based on the adjusted keyword identifier set to form a structured record format, writes it into the cloud database under the specified path, and establishes the courseware keyword optimization results; Based on the adjusted keyword identifier set, the keyword content, fragment number, sentence position index, semantic label and other fields in each record are extracted and constructed into a standard structured data table. The field format includes four items: keyword name, starting position, sentence index, and semantic type. After generating the structured record, the system calls the data storage interface according to the configuration parameters, sets the data path parameters, and the path address is set through the cloud platform variables, such as " / resource / kwdata / opt_result / ". The structure table is then written to the cloud database table under the path in sequence, a unified number index is established, and field archiving is completed, finally generating the courseware keyword optimization results.

[0045] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A computer teaching resource management system based on cloud computing, characterized by: The system comprises: The pair mismatch screening module extracts verb and noun combinations from the courseware text content from the cloud and classifies them. It then screens the pair mismatch rate sequence by detecting whether each combination appears in the target vocabulary set of the specified teaching task. The semantic boundary recognition module performs increasing trend clustering recognition on the pair mismatch rate sequence by using a fuzzy C-means clustering algorithm, and marks a set of semantic boundary transition points of the courseware; The keyword semantic determination module identifies the occurrence position of keyword identifiers in the courseware text content through cloud semantic analysis, and determines the semantic conversion state of keywords based on the courseware semantic boundary transition point set, and organizes them into a keyword function switching list; The keyword classification management module performs boundary classification judgment on the keyword context structure feature value vector based on the keyword function switching list through the support vector machine algorithm to obtain the courseware keyword management result.

2. The computer teaching resource management system based on cloud computing according to claim 1, characterized in that: The group mismatch rate sequence includes the semantic inconsistency group number, paragraph index number, and mismatch rate change value; the courseware semantic boundary transition point set is specifically the position index, cluster label type, and mismatch trend direction; the keyword function switching list includes the function switching identifier, semantic role comparison item, and structural position mapping group; the courseware keyword management result includes the boundary classification result mark, the valid keyword set index, and the context structure feature vector number.

3. The computer teaching resource management system based on cloud computing according to claim 1, characterized in that: The pair mismatch screening module includes: The phrase extraction submodule obtains all sentences in the courseware text, identifies and extracts verbs with action characteristics and corresponding noun combinations in the sentences, forms paired items according to syntactic dependency rules, and generates action-object group constituent sets based on the semantic combination of the extracted paired items; The attribution detection submodule detects whether each object belongs to the preset teaching task target vocabulary set based on the action-object group set, and uses the coverage ratio and correlation calculation method of the object content in the combination and the target vocabulary set to perform attribution judgment, thereby obtaining the object attribution relationship judgment set; The sequence construction submodule counts the number of unattributed combinations and the total number of combinations according to the object attribution relationship determination value set, calculates the ratio of the combinations and arranges them in structural order to generate a pair mismatch rate sequence.

4. The computer teaching resource management system based on cloud computing according to claim 1, characterized in that: The semantic boundary recognition module includes: The trend calculation submodule extracts the mismatch rate value of each of three consecutive content segments in the mismatch rate sequence of the pair group, calculates the mismatch rate increase between the two segments before and after, and determines whether the increase directions of the three segments are in the same direction. The set of segments with the same increase direction is selected as the trend area input item to generate a set of continuous increasing trend segments; The fuzzy classification submodule extracts the mismatch rate sequence corresponding to each segment in the continuous increasing trend segment set, sets the initial number of cluster centers and the fuzzy coefficient, and calls the fuzzy C-means clustering algorithm to perform membership division on the mismatch rate values to obtain the mismatch rate membership cluster label; The boundary annotation submodule determines whether the first position of the mismatch category segment where the cluster label appears continuously coincides with the position of the continuous increasing trend segment based on the mismatch rate affiliation cluster label. If so, the target position is added with a boundary marker and recorded as a courseware content segment index to obtain the courseware semantic boundary transition point set.

5. The computer teaching resource management system based on cloud computing according to claim 4, characterized in that: The mismatch rate value is divided into membership levels using the formula: ； in, Indicates the The sample points are The membership degree of the cluster center, It is Normalized mismatch rate sample values, Indicates the The value of the cluster center, is the number of cluster categories, is the fuzzy coefficient, represents the Euclidean distance.

6. The computer teaching resource management system based on cloud computing according to claim 1, characterized in that: The keyword semantic determination module includes: The position extraction submodule obtains all sentence structures in the courseware text content, screens the keyword content including multi-frequency information components, combines the part-of-speech tagging rules with the context grammatical structure to determine whether the keyword independently plays a semantic pointing role, locates the position index in the content fragment, and generates the keyword identification positioning result; The role identification submodule collects the first verb information and subject-verb combination items of the corresponding sentence based on the keyword identification and positioning results, and determines the semantic function of the keyword in the sentence based on the verb features, judging whether it serves as a functional instruction action or an annotation modification content, and calculates the semantic deviation direction based on the stability of the word order structure to obtain the keyword function role classification result; The function conversion submodule is based on the keyword function role classification results and the courseware semantic boundary transition point set, and screens whether the keyword crosses the boundary area and undergoes a semantic role change, and determines whether the function label classification in the previous and next areas is inconsistent. If the role category is inconsistent, it is marked as a conversion item and a keyword function switching list is established.

7. The computer teaching resource management system based on cloud computing according to claim 1, characterized in that: The keyword classification management module includes: The feature construction submodule extracts the context sentences of each keyword in the courseware content in the keyword function switching list, calls syntactic dependency rules and position information to extract adjacent phrases, word order depth and sentence structure indicators, and constructs the indicator combination into a feature vector to generate a context structure feature value vector set; The boundary calculation submodule sets the boundary threshold and penalty coefficient according to the keyword function switching list and the context structure feature value vector set, classifies the samples by maximizing the interval using the support vector method, calculates the belonging boundary of each keyword between the switching areas, and the belonging boundary includes the continuous change class or the stable continuation class, and obtains the keyword boundary discrimination result; The classification and screening submodule identifies all keyword identifiers whose boundary categories are stable and continuous categories based on the keyword boundary discrimination results, relocates and collects the keywords according to their unique codes in the original text, and establishes courseware keyword management results.

8. The computer teaching resource management system based on cloud computing according to claim 7, characterized in that: To calculate the belonging boundary of each keyword between switching areas, the formula is used: ； Constructing a linear attribution boundary function , used to map the structural features of each keyword into a continuous numerical score and compare it with the set boundary threshold to determine the attribution boundary; in, For the The structural feature vector of keywords, is the dependency weight between keywords and verbs, is the depth of the keyword in the syntax tree, indicating the structural distance from the root node of the sentence. Code the sentence structure as follows: declarative sentence = 1, imperative sentence = 2, interrogative sentence = 3, parallel sentence = 4, , , is the discriminant weight corresponding to each dimension; is a bias term used to adapt to the distribution offset between multi-class structure samples.

9. The computer teaching resource management system based on cloud computing according to claim 1, characterized in that: The module further includes a keyword optimization and updating module, which determines the consistency between the keywords in the courseware keyword management results and the corresponding semantic structures, adjusts the keyword identifiers according to the determination results, obtains the courseware keyword optimization results and stores them in the cloud database; The courseware keyword optimization result includes the corrected keyword identifier, the corresponding semantic structure identifier, and the cloud database storage path.

10. The computer teaching resource management system based on cloud computing according to claim 9, characterized in that: The keyword optimization and updating module includes: The consistency determination submodule obtains the courseware keyword management results and the action-object group composition set, extracts the verb and object combination items corresponding to the keyword sentence, determines whether the keyword appears in the object item or forms a logical association with the verb, and generates a keyword pointing matching result; The identifier adjustment submodule filters out keyword identifiers that do not establish a verb or object semantic connection based on the keyword pointing matching results, reconstructs the position numbers and updates the semantic tags of the remaining keywords, and obtains an adjusted keyword identifier set; The data storage submodule extracts the keyword content, position index, and corresponding segment number fields based on the adjusted keyword identification set to form a structured record format, writes it into the cloud database under the specified path, and establishes the courseware keyword optimization results.

Images