A big data digital art event management method and system

Abstract

The application discloses a kind of big data digital art event management method and system, belong to big data processing and digital art event management technical field.The method includes: in response to event start instruction, the digital fingerprint of the works of the competition is read and organized as chain block, and the registration of evidence is completed by distributed node consensus verification;Read the work description text and art style label to construct the work characteristic code, and match the matching degree matrix with the evaluation characteristic file;Dynamic evaluation weight is obtained by using multi-level polling mechanism to distribute initial evaluation task, monitoring behavior feedback data periodically corrects evaluation weight;Dimensional view is generated by dimension classification aggregation according to whole-process state data, and pushed to the management end display.The present application realizes the event work credible evidence, evaluation task accurate matching and dynamic adjustment and event situation visualization.

Description

Technical Field

[0001] This invention relates to the fields of big data processing and digital art competition management technology, specifically a big data-based digital art competition management method and system. Background Technology

[0002] Digital art competitions involve a large number of entries and numerous judges, placing high demands on the reliability of work verification and the rationality of judge allocation in competition management. Currently, work registration typically relies on centralized server storage, with digital fingerprints and time information maintained by a single entity, posing risks of data tampering or loss, and making it difficult to prove the originality of works. In terms of judge allocation, random assignment or fixed category division is often used, failing to fully consider the artistic style preferences and professional differences of judges, leading to a mismatch between judging styles and expertise, affecting the objectivity and fairness of evaluation results. Regarding competition progress management, the lack of an effective dynamic feedback mechanism for judging behavior makes it impossible to adjust the distribution of judging resources in a timely manner based on actual performance during the judging process. Inefficient judging behavior can delay the overall competition progress, and administrators struggle to monitor the status of each stage of the competition in real time and quickly identify bottlenecks. Combining reliable work verification with precise judge allocation has become a challenge in the field of digital art competition management. On the one hand, a tamper-proof and multi-verifiable verification mechanism is needed to ensure that the ownership time and content integrity of entries are free from human interference. On the other hand, it is necessary to accurately extract artistic style characteristics from the descriptions of the works and quantitatively correlate them with the professional background of the judges and precedents in judging, forming a style-oriented basis for task assignment. Furthermore, the competition management must also have the ability to continuously perceive the behavior of the judges, and be able to adjust the judges' discourse power based on real-time behavioral data, eliminate inefficient links, and ensure that the allocation of judging resources always dynamically adapts to the actual progress of the competition. Summary of the Invention

[0003] The purpose of this invention is to provide a big data digital art competition management method and system to solve the problems of easy tampering of work evidence, lack of style matching in judging assignment and lack of dynamic behavior correction in the existing technology, and to realize credible on-chain evidence of competition works, automatic adaptation of judging tasks and judging styles, and visual monitoring of the competition situation.

[0004] To achieve the above objectives, this invention provides the following technical solution: This invention provides a big data-based digital art competition management method, comprising: responding to a competition launch command, reading the digital fingerprints of the entries, organizing the digital fingerprints into chain blocks in chronological order, and verifying the chain blocks through distributed nodes to complete the registration and certification of the entries. By utilizing the chain block structure combined with distributed consensus, the work certification records possess the characteristics of immutability and traceability, ensuring the credibility of the competition data from the source. Based on the registered entries, reading the work description text and associated art style tags of each entry, constructing a work feature code based on the mapping relationship between the work description text and the art style tags, and performing a matching calculation based on the work feature code and the reviewer's feature profile to obtain a matching degree matrix between the reviewer and the entries. The generation of the matching degree matrix can accurately align the work style characteristics with the reviewer's expertise and preferences, improving the rationality and professionalism of the review allocation. Based on the matching degree matrix, a multi-level polling mechanism is used to assign corresponding initial review tasks to each reviewer. During the competition, behavioral feedback data from the reviewers is monitored, and the review weights of each reviewer are periodically adjusted based on this data, resulting in dynamic review weights. This dynamic weight adjustment allows for adaptive adjustment of review task allocation based on the actual response of each reviewer, preventing the overall review process from being blocked due to insufficient response from individual reviewers and maintaining the efficiency of the competition. Full-process status data is collected during the competition and categorized and aggregated according to preset data dimensions to generate a dimensional view recording the competition's progress. This dimensional view is then pushed to the management terminal for display. Managers can use this multi-dimensional view to gain real-time insight into evidence preservation progress, review load, and process bottlenecks, enabling centralized monitoring and agile scheduling throughout the entire competition cycle.

[0005] As a preferred technical solution of the present invention, the generation of digital fingerprints includes: performing block hashing on the file content of the entry, then performing hashing on the concatenated result of the block hash values ​​to obtain a digital fingerprint, and attaching a timestamp and participant identifier. The assembly and verification of the block includes: combining the digital fingerprint, timestamp, and participant identifier into the block body; obtaining the hash value of the previous block as the block header; assembling the block body and block header into the current block; broadcasting the current block to a distributed network; having distributed nodes verify the order of the timestamps and the consistency of the hash value of the previous block; and appending the current block to the end of the chain structure after more than half of the nodes confirm its validity. This method of evidence preservation further strengthens the irreversibility of the registration sequence of the work and its anti-counterfeiting capabilities.

[0006] As another preferred technical solution of the present invention, the acquisition of the matching degree matrix specifically includes: segmenting the work description text to obtain a keyword sequence, mapping and matching it with a preset art style tag library, extracting the successfully matched art style tags and arranging them in the order of their appearance in the work description text; quantifying and encoding the arranged art style tag sequence to generate work feature codes; reading the review feature files of each reviewer, which contain preferred art style types and historical review records, converting them into review feature vectors, calculating the similarity between the work feature code and the review feature vector as the matching degree, and summarizing the matching degrees of all entries with all reviewers to form a matching degree matrix. This solution transforms abstract art styles into computable coded features, significantly reducing human intervention and subjective bias in review matching.

[0007] As a further preferred technical solution of the present invention, obtaining dynamic review weights includes: sorting the reviewers from high to low according to the matching degree values ​​in the matching degree matrix, assigning initial review tasks to the reviewers, and removing the assigned entries from subsequent polling; collecting the review time and review interaction frequency of the reviewers at fixed time intervals during the competition process, and calculating the response activity; marking reviewers with response activity below a preset activity threshold as reviewers to be corrected, reducing their review weights, and reallocating the unfinished review tasks of these reviewers according to the reduced review weights. This achieves timely correction of review weights, ensuring that review resources are always efficiently allocated throughout the entire competition process.

[0008] Preferably, the generation of the dimensional view includes: collecting work registration status data, task allocation status data, and review weight status data from the registration module, task allocation module, and weight correction module respectively, and uniformly formatting them into status records with timestamps; classifying and aggregating the status records according to work dimension, reviewer dimension, and time dimension, counting the number of completed and in-progress items under each dimension, generating a work progress view, a reviewer load view, and a competition timeline view, combining them into a competition situation view, and pushing it to the management terminal for display in the form of visual charts. Through the multi-dimensional situation view, managers can clearly grasp the overall progress of the competition and the details of each stage, and make timely resource allocation and anomaly handling.

[0009] This invention also provides a big data digital art competition management system, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the aforementioned big data digital art competition management method. This system integrates functions such as secure storage of works, intelligent matching of works with judges, dynamic adjustment of judging weights, and visualized monitoring of the entire process, significantly improving the standardization, fairness, and operational efficiency of digital art competition management.

[0010] The technical effects and advantages provided by the present invention in the above technical solution are as follows: The content of the submitted entries is divided into multiple data blocks. A hash operation is performed on each block to obtain a block hash value. These block hash values ​​are then concatenated in sequence and hashed again to generate a digital fingerprint of the entry. A timestamp and participant identifier are added to this digital fingerprint. This digital fingerprint, along with the timestamp and participant identifier, is combined to form the block body. The hash value of the previous block is used as the block header. After assembling the current block, it is broadcast to the distributed network. Each online node extracts the timestamp of the current block and the hash value of the previous block, compares them with its latest stored chain structure, verifies the order of the timestamps and the consistency of the previous block's hash, and returns a confirmation signal to the broadcast source upon successful verification. When the number of confirmation signals exceeds half of the total number of online nodes, the current block is written to the end of the chain structure and notified to the entire network. By using embedded timestamps and hash associations within a chain of blocks, any modification to the registered data will break the chain and be rejected by other nodes. The artwork registration record possesses the characteristics of immutability and full traceability. The consensus process among distributed nodes avoids the risk of manipulation by a single node, providing self-verifiable time-series evidence for the confirmation of ownership of digital artworks. The artwork description text is obtained from the registered entries, segmented, and mapped and matched against a pre-defined art style tag library. Successfully matched art style tags are extracted and quantified according to their position in the text to generate artwork feature codes. The judging feature profile, containing the judges' preferred art style types and historical judging records, is converted into judging feature vectors. The similarity between the artwork feature code and the judging feature vector is calculated, and a matching degree matrix is ​​formed. Initial judging tasks are assigned to judges based on matching degree from high to low, and the assigned artworks are removed from subsequent rounds. During the competition, judging time and interaction frequency are collected at fixed time intervals to calculate response activity. Judges with activity levels below a threshold are marked as needing correction, and their judging weight is reduced. Their unfinished judging tasks are then reassigned based on the adjusted weight. By calculating the similarity between the artwork's feature code and the reviewer's feature vector, precise stylistic targeting is achieved, reducing review bias caused by random assignment. Dynamic weight adjustment promptly weakens the influence of negative reviewers, transferring their backlogged tasks to active reviewers and ensuring that the overall pace of the competition is not hindered by individual events. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0012] Figure 1This is a flowchart of a big data-driven digital art competition management methodology. Figure 2 This is a flowchart of the registration and certification process for submitted entries; Figure 3 This is a flowchart of the process for generating the art style matching matrix of the submitted works; Figure 4 This is a flowchart of dynamic task allocation based on matching degree ranking and review activity. Figure 5 This is a flowchart of the workflow of a digital art competition management system. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0014] See Figure 1 This invention provides a big data-based digital art competition management method and system. The method includes: responding to a competition launch command, reading the digital fingerprints of the entries, organizing the digital fingerprints into chain blocks in chronological order, verifying the chain blocks through distributed nodes to complete the registration of the entries; based on the registered entries, reading the work description text and associated art style tags of each entry, constructing a work feature code based on the mapping relationship between the work description text and the art style tags, performing a matching calculation between the work feature code and the review feature files of the reviewing end to obtain a matching degree matrix between the reviewing end and the entries; based on the matching degree matrix, assigning corresponding initial review tasks to each reviewing end using a multi-level polling mechanism, monitoring the behavioral feedback data of the reviewing end during the competition process, periodically adjusting the review weights of the reviewing end based on the behavioral feedback data to obtain dynamic review weights; collecting full-process status data during the competition, classifying and aggregating the full-process status data according to preset data dimensions, generating a dimension view recording the progress status of the competition, and pushing the dimension view to the management end for display.

[0015] Example 1:

[0016] In specific implementation, please refer to Figure 2 In response to the competition launch command, the system executes the registration and documentation process for submitted works. The competition launch command is issued by the competition management platform during the open submission phase, triggering the registration and documentation module to run.

[0017] The original file data of the submitted entries is read and divided into multiple data blocks according to a preset fixed byte length. The preset fixed byte length is selected as 4096 bytes, which ensures that the size of each data block is consistent with the commonly used block size in a file system, facilitating parallel processing of hash operations. A hash operation is performed on each data block to obtain a block hash value corresponding to each block. The hash operation used is the SHA-256 algorithm, which can map data of any length to a 256-bit hash value. The hash values ​​of each block are concatenated end-to-end according to the original order of the data blocks to form a concatenated sequence. The SHA-256 hash operation is then performed on the concatenated sequence again to obtain the digital fingerprint of the submitted entry. The digital fingerprint is a unique digest of the overall content of the original file data.

[0018] The current system time is obtained as the timestamp, accurate to the millisecond, and expressed in Coordinated Universal Time (UTC). The participant's unique registration identifier is read as the participant identifier; this identifier is a globally unique string generated by the system during participant registration. The digital fingerprint, timestamp, and participant identifier are associated and stored as a single evidence record, which is then temporarily written to a local database.

[0019] The digital fingerprint, timestamp, and participant identifier are combined into the block body by concatenating them into a byte string in a preset order. The hash value of the previous block is used as the block header. The previous block refers to the last block in the existing chain structure. If the current block is the first block in the competition, the hash value of the previous block uses the preset genesis block hash value. The block body and block header are then assembled into the current block. During assembly, the version number and the hash value of the previous block are written into the block header, and the concatenated byte string is written into the block body. The block hash value of the current block is calculated, forming a complete block data structure.

[0020] The assembled current block is broadcast to all online nodes in the distributed network. The distributed network consists of multiple independent distributed nodes connected via a peer-to-peer protocol, with each node maintaining a complete copy of the chain structure. Online nodes are distributed nodes that are connected and able to send and receive messages at the time of broadcast.

[0021] Upon receiving the current block, each online node extracts the timestamp of the current block and the hash value of the previous block, and also extracts the hash value field of the previous block recorded within the current block. The online node retrieves the timestamp of the latest block from its own stored chain structure and calculates its hash value. The online node performs sequence verification, which includes two comparison operations: comparing whether the timestamp of the current block is later than the timestamp of the latest block, and comparing whether the hash value of the previous block recorded in the current block matches the calculated hash value of the latest block. The timestamp comparison uses a strict greater-than condition to ensure that the block timeline is strictly incremental.

[0022] If both comparisons in the sequential verification are true, the online nodes return confirmation signals to the broadcast source. The broadcast source is the node that initiated the broadcast of the current block, typically the server node where the evidence registration module resides. The broadcast source receives confirmation signals from each online node and counts the number of received confirmation signals. When the number of confirmation signals received by the broadcast source exceeds half of the total number of online nodes, the current block is written to the tail of the chain structure, that is, the current blockchain is appended after the latest block, updating the chain structure. Subsequently, the broadcast source sends an append completion notification to all online nodes, containing the block hash value of the appended current block. After receiving the append completion notification, the online nodes synchronously write the current block to their own stored chain structure copy, completing the evidence registration of the entry. If the number of confirmation signals received by the broadcast source within the preset timeout period does not exceed half of the total number of online nodes, the current block is abandoned, reassembled, and broadcast again until the consensus condition is met.

[0023] Example 2:

[0024] In specific implementation, please refer to Figure 3 The system retrieves the description text from the registered entries. This description text, filled in by the entrants when submitting their entries, is stored in the entry's metadata. The description text is then segmented to obtain a keyword sequence. The segmentation process uses a dictionary-based forward maximum matching method combined with a pre-defined stop word list. Continuous character sequences in the description text are matched against dictionary entries to extract meaningful words. Stop words appearing in the stop word list are removed, and the remaining words are arranged in their original order to form the keyword sequence.

[0025] The keyword sequence is mapped and matched against art style tags in a pre-defined art style tag library. This library contains a set of standardized tags representing different art styles; each tag is a string composed of Chinese characters or letters. The mapping and matching process is as follows: each art style tag in the library is read sequentially, and each tag is matched character-wise against each keyword in the keyword sequence. When the string of an art style tag is exactly the same as the string of a keyword, a complete match is considered successful, and the tag is marked as a successful match. Multiple successful matches may occur. The successful matches are then sorted according to their first appearance in the artwork description text, with the order determined by character offset; smaller offsets appear first, forming an ordered sequence of art style tags.

[0026] The extracted art style tags are arranged in the order of their appearance in the artwork description text. The arranged art style tags are then quantized and encoded to generate artwork feature codes. Before quantization and encoding, a pre-defined encoding table is used. Each pre-defined art style tag is assigned a unique binary code in the encoding table. The binary code for each art style tag has the same fixed length, and the encoding bit width is denoted as [bit width]. , The value is set to 16, which is chosen to ensure that the maximum capacity of the art style tag library is reached. The encoding table covers common art style categories with some redundancy. The generation method for each code in the encoding table is as follows: after sorting all art style tags in the art style tag library in lexicographical order, the sorting sequence number is represented in binary, with a minimum of [missing information - likely a missing section]. Pad the left side of each bit with zeros. Following the order of the art style tags in the art style tag sequence, read the binary code corresponding to each art style tag in the encoding table sequentially. Concatenate these binary codes end-to-end in the order they were read to generate a concatenated binary sequence, which is then used as the initial feature code.

[0027] The code length of the initial feature code is checked to see if it reaches a preset fixed code length, which is denoted as . , The value is set to 1024, a value chosen to balance the expressive power of the vector dimension with subsequent computational efficiency, providing sufficient encoding space to include multiple label information. If the code length of the initial feature code is less than the fixed code length... Then, padding bits are added to the end of the initial feature code, with each bit of the padding bit being 0, and the padding quantity is [number missing]. The difference between the initial feature code length and the actual code length is padded with filler bits to achieve a fixed code length. The padded sequence is used as the artwork's feature code. If the code length of the initial feature code is greater than the fixed code length... Then, a fixed code length is extracted starting from the first bit of the initial feature code. The fragments are used as the work's unique identifier. The work's unique identifier is a set of... A sequence of binary bits, where each bit can be either 0 or 1.

[0028] Read the judging feature files from each judging panel. These files contain the judges' preferred art style types and their historical judging records. The preferred art style types are stored as a set of tags, each tag coming from a pre-defined art style tag library. The judges' historical judging records include the identifiers of previously judged entries, the art style tags for each entry, and the scores given to each entry. Convert the judging feature files into judging feature vectors, where the dimension of the feature vector is the same as the fixed code length of the artwork's feature code. Similarly, each component of the review feature vector corresponds to the art style tag associated with the same binary bit in the artwork's feature code. The conversion process is as follows: determine which bit of which art style tag in the encoding table each component of the review feature vector corresponds to. Let the th element be... Each component corresponds to a specific art style tag ( It is an integer. The process involves: 1) determining the presence of the art style tag within the judges' preferred art style types. If the art style tag is present in the judges' preferred art style types, the preference contribution value is set to 1; otherwise, it is set to 0. 2) Extracting the scores of all entries tagged with the art style tag from the judges' historical review records, calculating the average of these scores, and normalizing this average to the interval [0,1] by dividing it by the maximum score. If no scores for the art style tag are found in the historical review records, the historical contribution value is set to 0. 3) Weighted summing of the preference contribution value and the historical contribution value to synthesize the _th_th element in the judges' feature vector. The values ​​of the first and second components are weighted by coefficients of 0.6 and 0.4, respectively. The weighting coefficients are chosen because preference information reflects the reviewer's subjective intentions better than historical rating information; therefore, the preference contribution value has a higher weight. The first component in the review feature vector... Each component is denoted as The calculation method is as follows:

[0029] in, For the first Whether the art style tag corresponding to each component belongs to the art style type preferred by the judges, and whether it belongs to the time period Not belonging to time ; For the review end to the first The historical normalized mean score of the art style tag corresponding to each component. The range of values ​​is When there is no historical rating data ; The value of falls within the interval Inside. Traverse all After the value is calculated, the complete review feature vector is obtained.

[0030] The work's unique code is considered as a... The vector composed of n components, the i-th component in the work's feature code The value of each component is denoted as , The value is either 0 or 1. The similarity between the artwork's feature code and the reviewer's feature vector is calculated as the matching score. Cosine similarity is used as the similarity metric, and its calculation method is as follows:

[0031] in, This indicates the degree of match between the submitted work and the judging criteria. The first of the work's feature codes One portion, For the review feature vector of the th One portion, A fixed code length for the work's feature code. This represents the sum of the products of all components of the work's feature code and the corresponding components of the review feature vector. This represents the magnitude of the feature vector of the work. This represents the magnitude of the evaluation feature vector. When the denominator is zero, The value is 0.

[0032] Perform the matching degree calculation as described above on all entries and all judges to obtain the matching degree value between each entry and each judge. Then, with all entries as rows and all judges as columns, fill the matching degree values ​​into the corresponding positions in the matrix and summarize them to form a matching degree matrix.

[0033] Example 3:

[0034] In specific implementation, please refer to Figure 4 Obtain the matching degree matrix generated by the matching degree calculation module. Each row in the matching degree matrix corresponds to an entry, and each column corresponds to a judge. The element values ​​in the matrix represent the matching degree between the corresponding entry and the corresponding judge. The matching degree values ​​range from [0,1]. The higher the value, the better the fit between the entry and the judge.

[0035] The entries and judges' combinations are sorted from highest to lowest according to the matching degree values ​​in the matching degree matrix, forming a sorted list. Each item in the sorted list is a combination item, which includes an entry identifier, a judge's identifier, and a matching degree value. The matching degree value is used as the sorting key, and the items are arranged in descending order. When multiple combinations have the same matching degree value, the lexicographical order of the entry identifiers is used to determine the order. The entry identifier is a unique number generated during the entry's registration and certification process.

[0036] A multi-level polling mechanism is used to assign initial judging tasks to each judging panel. The mechanism works as follows: starting with the combination with the highest matching score in the sorted list, each combination is processed sequentially. For the currently processed combination, it checks whether the judging panels in the combination have reached the preset task allocation limit and whether the entries in the combination have already been assigned. If the judging panels have not reached the task allocation limit and the entries have not been assigned, the entries in the combination are assigned to the judging panels in the combination and marked as assigned. The task allocation limit for each judging panel is set to a fixed value of 3. This value is based on the number of entries each judging panel can fully judge in one round of the competition, ensuring judging coverage while avoiding excessive judging load. Assigned entries are removed from subsequent polls, and are skipped when a combination containing such entries is encountered later. After processing one combination, the process continues with the next combination in the sorted list until all judging panels have reached their task allocation limit or all entries have been assigned.

[0037] After the initial judging tasks are assigned, the monitoring of the competition process is initiated. During the competition, judging behavior data on the assigned entries is collected at fixed time intervals. The fixed time interval is set to 60 seconds; this value allows the system to promptly capture changes in judging behavior while avoiding excessive data collection that could interfere with the judging terminals. Judging behavior data includes judging time and judging interaction frequency. Judging time is the length of time from when the judge opens the entry details page to when they submit their score, recorded in seconds. Judging interaction frequency is the ratio of the total number of swipes, zooms, clicks, etc., performed by the judge on the entry details page during the judging process to the judging time, calculated in times per minute.

[0038] The response activity level of the reviewing end is calculated based on the collected review time and review interaction frequency. Response activity level reflects the level of active participation of the reviewing end in the review task. The calculation method for response activity level is as follows: the review interaction frequency is mapped to a preset baseline interval by taking the ratio of the review interaction frequency to the preset baseline frequency, with the result limited to the range [0,1]. The preset baseline frequency is set to 10 times / minute, based on the statistical value of the average interaction frequency under normal review operations. The review time is converted into a time factor, which is the ratio of the preset standard review time to the actual review time, with the result limited to the range [0,1]. The preset standard review time is set to 300 seconds, representing the expected review time for a moderately complex entry. The calculation method for response activity level is as follows:

[0039] in, This indicates the level of activity on the review side. The value range is [0,1]; This indicates the frequency of review interactions on the review side, measured in times per minute. This indicates the preset reference frequency. ; This indicates the review time on the review side, in seconds. This indicates the preset standard review time. ; Contribution coefficient to interaction frequency The value is 0.5, which means that the frequency of review interactions and the time spent on review have equal weight in the calculation of response activity. This indicates that the smaller value is taken between 1 and the ratio of the review interaction frequency and the baseline frequency; This indicates that 1 is taken as the smaller of the ratio of standard review time to review time.

[0040] The response activity level is compared with a preset activity threshold of 0.3. This threshold is chosen because a low response activity level indicates that the reviewer may not have diligently fulfilled their review responsibilities. Reviewers with response activity levels below the preset threshold are marked as needing correction. The review weight of the reviewer to be corrected is reduced by multiplying its current review weight by a decay factor of 0.5. This decay factor is chosen to halve the review weight of the reviewer to reduce its contribution to the final score. The initial review weight before reduction is the default value of 1.0.

[0041] The incomplete review tasks of the reviewing end awaiting correction are redistributed based on the reduced review weights. Incomplete review tasks refer to entries that have been assigned to the reviewing end but for which review scores have not yet been submitted. During redistribution, the entries corresponding to the incomplete review tasks of the reviewing end awaiting correction are returned to the pool of available entries, and other reviewing ends with the highest matching degree are selected from the sorted list for assignment. These other reviewing ends are those not marked as awaiting correction and have not reached their task assignment limit. After redistribution, a dynamic review weight is obtained, which includes the default weight value of 1.0 maintained by normal reviewing ends and the reduced review weight value of the reviewing end awaiting correction, for use in the subsequent scoring and summarizing stage. Periodic corrections are performed according to a preset correction cycle of 600 seconds. This value ensures that weight corrections are performed at least once within a review task cycle. The correction cycle, combined with the fixed time interval, ensures timely handling of abnormal review behavior.

[0042] Example 4:

[0043] In practice, during the competition, the work's registration and documentation status data is collected from the registration and documentation module. This data includes the registration and documentation status of each entry, with four possible values: "Pending Registration," "Registering," "Registered," and "Registration Failed." During data collection, the registration and documentation module records each entry's registration and documentation status change event in log form. Log entries include the entry's identifier, the status value before the change, the status value after the change, and the timestamp of the change. The work's registration and documentation status data collector polls the registration and documentation module's status recording interface at a fixed interval of 30 seconds to retrieve any new status change events since the last collection.

[0044] Task allocation status data is collected from the task allocation module. This data includes a list of entries assigned to each review panel and the status of each allocation relationship, which can be categorized into four values: "Pending Review," "Under Review," "Submitted," and "Overdue for Review." The task allocation module generates a task status record each time an entry is allocated or a review panel submits its review results. This record includes the review panel identifier, entry identifier, allocation timestamp, and current status value. The task allocation status data collector queries the task allocation module for task status records at a fixed interval of 30 seconds.

[0045] The review weight status data is collected from the weight correction module. This data includes the current review weight value for each reviewer and its weight change history. After each review weight correction, the weight correction module generates a weight change record, which includes the reviewer's identifier, the weight value before correction, the weight value after correction, the reason for correction, and the correction timestamp. The review weight status data collector reads the weight change records from the weight correction module at a fixed interval of 30 seconds.

[0046] The collected data on work registration status, task allocation status, and review weight status are uniformly formatted into timestamped status records. During formatting, the work registration status data is extracted from the entry identifier, registration status, and status change timestamp to construct a work registration status record. This record includes a record type field, an entity identifier field, a status value field, and a record time field, with the record type field set to "Work Registration". The task allocation status data is extracted from the reviewer identifier, entry identifier, allocation relationship status, and allocation timestamp to construct a task allocation status record, with the record type field set to "Task Allocation". The review weight status data is extracted from the reviewer identifier, current review weight value, and weight change timestamp to construct a review weight status record, with the record type field set to "Weight Correction". All constructed status records carry a unified timestamp with second-level precision.

[0047] The status records are categorized according to three dimensions: work, judge, and time. For the work dimension, the entry identifier is used as the classification key, grouping status records by entry identifier. For each entry, the current value of the registration status and the number of judges associated with the task assignment status record are calculated. For the judge dimension, the judge identifier is used as the classification key, grouping status records by judge identifier. For each judge, the total number of assigned entries, the number of entries reviewed, and the current review weight value are calculated. For the time dimension, the status records are categorized into corresponding time intervals based on the timestamp, with a preset time granularity of 300 seconds. The competition duration is divided into consecutive 300-second intervals, and the frequency of status value changes within each time interval is calculated.

[0048] The number of completed and ongoing entries are counted separately for each dimension. Under the entry dimension, entries with a registration status of "registered" are counted as completed, while entries with a registration status of "pending registration" or "registering" are counted as ongoing. Under the judging dimension, entries that have been judged and submitted for scoring are counted as completed, while entries that have been assigned but whose assignment status is "pending judging" or "judging" are counted as ongoing. Under the time dimension, within each time interval, the number of times the status changes to "completed" is counted as completed, and the number of times the status changes to "ongoing" is counted as ongoing.

[0049] Generate a progress view of the submitted works. The progress view is represented by a visual progress bar. The total length of the progress bar corresponds to the total number of all submitted works. The filled portion of the progress bar corresponds to the number of completed works, and the unfilled portion corresponds to the number of works in progress. The fill ratio of the progress bar is calculated as follows:

[0050] in, Indicates the progress percentage of the work. The value range is [0,1]; This indicates the number of completed works in the category of "works," specifically the number of entries with a "certificate registered" status. This indicates the total number of entries that have been registered and filed. The progress view of the entries is displayed in the entry registration area of ​​the competition management interface.

[0051] Generate a reviewer load view. The reviewer load view displays the load status of each reviewer in list format. Each row in the list corresponds to a reviewer and includes the reviewer's identifier, the number of entries assigned, the number of reviews completed, and the load percentage. The load percentage is calculated by dividing the number of reviews completed by the number of entries assigned; when the number of assigned entries is zero, the load percentage is 0. Reviewers are sorted from lowest to highest load percentage. Reviewers with a load percentage below a preset load threshold (0.3) are marked with a warning color, indicating that the corresponding reviewer's review progress is lagging.

[0052] Generate a timeline view of the event. The timeline view uses a time axis as its main axis, divided into continuous time intervals according to a preset time granularity. Within each time interval, the completed and ongoing quantities are overlaid and displayed. The completed quantities are represented by the height of a bar chart, while the ongoing quantities are represented by bars of different colors stacked on top of each other. The timeline starts at the moment the event start command is issued and ends at the current system time, extending to the right in real time as the event progresses.

[0053] The project progress view, the judges' workload view, and the event timeline view are combined into a single event status view. The event status view uses a grid layout, with the project progress view at the top, the judges' workload view at the bottom left, and the event timeline view at the bottom right. All three views share the same event execution time coordinates. When the user adjusts the time position in the event timeline view, the project progress view and the judges' workload view simultaneously refresh to display the status at the corresponding moment.

[0054] The event status view is pushed to the management terminal in the form of visual charts. The push is server-initiated, with a persistent WebSocket connection established between the management terminal and the management server. When the overall status data is updated and the event status view is regenerated, the management server serializes the latest event status view into a JSON data packet and sends it to the management terminal via the WebSocket connection. Upon receiving the data packet, the management terminal parses and re-renders the various chart components in the event status view, using Canvas elements for graphic drawing. When there is no stateful data update, the management server sends a heartbeat signal to the management terminal at 30-second intervals to maintain the connection. Upon receiving this signal, the management terminal maintains the current display content.

[0055] Example 5:

[0056] See Figure 5 A big data-driven digital art competition management system includes a memory, a processor, and a computer program stored in the memory and running on the processor. The memory is a physical storage medium with persistent data capabilities, specifically non-volatile memory, used to store the operating system, computer program, digital fingerprints and evidence records of entries, artwork description text and artistic style tag data, matching degree matrix, judging task allocation data, judging weight records, and overall process status data. The processor is a central processing unit with instruction execution and computation capabilities. The processor exchanges data with the memory via the system bus, reads computer program instructions from the memory, and executes corresponding operations according to the instructions.

[0057] In practice, the processor executes the computer program to implement the evidence registration step. Responding to the competition start command, the processor calls the evidence registration module to read the original file data of the entries stored in memory. The original file data is divided into multiple data blocks of a preset fixed byte length. A hash operation is performed on each data block to obtain a block hash value. The hash values ​​of each block are then concatenated in the original order of the data blocks and hashed again to generate the digital fingerprint of the entry. The processor obtains the current system time as a timestamp, reads the participant's unique registration identifier from memory as the participant identifier, and associates the digital fingerprint, timestamp, and participant identifier into an evidence record, which is then written back to memory. The processor combines the digital fingerprint, timestamp, and participant identifier into the block body, reads the hash value of the latest block from the chain structure stored in memory as the hash value of the previous block, assembles the block body and the hash value of the previous block into the current block, broadcasts the current block to all online nodes in the distributed network through the network interface connected to the system, and receives confirmation signals returned by each online node. When the number of received confirmation signals exceeds half of the total number of online nodes, the current block is written to the tail of the chain structure, completing the registration of the entry, and the updated chain structure is persisted to memory.

[0058] When executing the computer program, the processor performs a feature matching step. The processor reads the artwork description text from the registered entries stored in memory, performs word segmentation on the description text to obtain a keyword sequence. The keyword sequence is generated using a forward maximum matching method combined with a pre-set stop word list. The processor reads a pre-set art style tag library from memory, performs a character-to-character match between the keyword sequence and the tags in the library, and sorts the successfully matched art style tags according to their first appearance in the artwork description text, forming an ordered art style tag sequence. The processor reads a pre-set encoding table, concatenates the binary codes corresponding to each art style tag according to the order of the art style tag sequence to generate an initial feature code, checks whether the length of the initial feature code reaches a preset fixed length, and performs padding or truncation operations based on the detection result to obtain the artwork feature code. The processor reads the judging feature files from each judging panel from the memory. The judging feature files contain the artistic style types preferred by the judging panel and historical judging records. The judging feature files are converted into judging feature vectors. The similarity between the work's feature code and the judging feature vector is calculated as the matching degree. The matching degree values ​​between all entries and all judging panels are summarized to form a matching degree matrix, which is then written into the memory.

[0059] When executing the computer program, the processor implements the task allocation and weight adjustment steps. The processor reads the matching degree matrix from memory, sorts the combinations of entries and judges according to their matching degree values ​​from high to low, iterates through the sorted list, and assigns an entry as an initial judging task to each judge. During allocation, it checks whether the judge has reached the preset task allocation limit and whether an entry has already been assigned. Successfully assigned entries are marked as assigned and written back to memory. The processor starts a timed task to collect judging behavior data of the judges on the assigned entries at preset fixed time intervals. This data includes judging time and interaction frequency, and is written to memory. The processor calculates the judge's response activity based on the judging time and interaction frequency, compares it with a preset activity threshold, and marks judges with response activity below the threshold as needing correction. The judge's weight is reduced, and the unfinished judging tasks of the needing correction are reassigned to other judges based on the reduced weight, resulting in dynamic judging weights. These dynamic judging weights are then updated in the judging weight record in memory.

[0060] When executing the computer program, the processor performs the status acquisition and view generation steps. At fixed intervals, the processor collects work registration status data from the status log written to memory by the registration module, task allocation status data from the task status records written to memory by the task allocation module, and review weight status data from the weight change records written to memory by the weight correction module. The processor formats these three types of status data into timestamped status records and stores them in the full-process status dataset in memory. The processor categorizes the status records according to work dimension, reviewer dimension, and time dimension, and calculates the number of completed and ongoing works for each dimension, generating a work progress view, a reviewer load view, and a competition timeline view. These three views are then combined into a competition status view. The processor pushes the competition status view to the management terminal in the form of a visual chart through the system connection display interface. The push uses a WebSocket long connection. When status data is updated, the processor recalculates the competition status view, serializes it, and sends it to the management terminal via WebSocket. The management terminal receives the data and refreshes the chart rendering.

[0061] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A big data-driven digital art competition management method, characterized in that, The method includes: In response to the competition launch command, the digital fingerprint of the entries is read, and the digital fingerprints are organized into chain blocks in chronological order. The chain blocks are then verified through consensus by distributed nodes to complete the registration and storage of the entries. Based on the entries after registration and filing, the description text and associated art style tags of each entry are read. Based on the mapping relationship between the description text and the art style tags, the feature code of the entry is constructed. The feature code of the entry is matched and calculated with the review feature file of the reviewer to obtain the matching degree matrix between the reviewer and the entry. Based on the matching degree matrix, a multi-level polling mechanism is used to assign corresponding initial review tasks to each reviewer, and the behavioral feedback data of the reviewers is monitored during the competition. The review weights of the reviewers are periodically adjusted based on the behavioral feedback data to obtain dynamic review weights. Collect full-process status data during the event, classify and aggregate the full-process status data according to preset data dimensions, generate a dimensional view that records the progress status of the event, and push the dimensional view to the management terminal for display.

2. The method for managing digital art competitions using big data according to claim 1, characterized in that, In response to the competition launch command, the digital fingerprint of the submitted work is read, and the digital fingerprint is organized into a chain of blocks in chronological order. The chain of blocks is then verified through consensus by distributed nodes to complete the registration and notarization of the submitted work. Specifically, this includes: In response to the competition launch command, a hash operation is performed on the file content of the entries to generate a digital fingerprint, and a timestamp and participant identifier are added to the digital fingerprint; The digital fingerprint, timestamp, and participant identifier are combined into the block body, the hash value of the previous block is obtained as the block header, and the block body and block header are assembled into the current block. The current block is broadcast to the distributed network. The distributed nodes verify the current block sequentially based on the timestamp of the current block and the timestamp of the previous block. After more than half of the distributed nodes confirm the block, the current block is appended to the end of the chain structure, thus completing the registration and certification of the entry.

3. The method for managing digital art competitions using big data according to claim 1, characterized in that, The obtained matching matrix between the review panel and the submitted works specifically includes: Read the description text of the entries after they have been registered and filed. Perform word segmentation on the description text to obtain a keyword sequence. Map and match the keyword sequence with the tags in the preset art style tag library and extract the successfully matched art style tags. The extracted art style tags are arranged in the order of their appearance in the work description text, and the arranged art style tags are quantified and encoded to generate work feature codes. Read the judging feature files from each judging panel. The judging feature files contain the artistic style types preferred by the judging panel and the judging panel's historical judging records. Convert the judging feature files into judging feature vectors. Calculate the similarity between the work's feature code and the judging feature vector as the matching degree. Summarize the matching degrees of all entries with all judging panels to form a matching degree matrix.

4. The method for managing digital art competitions using big data according to claim 1, characterized in that, The obtained dynamic review weights specifically include: The entries and the judges are sorted from high to low according to the matching degree values ​​in the matching degree matrix. The entry with the highest matching degree is selected and assigned to the judge as the initial judging task. The assigned entry is removed from the subsequent rounds. During the competition, the judging behavior data of the judges on the assigned entries is collected at fixed time intervals. The judging behavior data includes the judging time and the frequency of judging interactions. The response activity of the review end is calculated based on the review time and the frequency of review interaction. Review ends with a response activity lower than the preset activity threshold are marked as ends to be corrected. The review weight of the ends to be corrected is reduced. The unfinished review tasks of the ends to be corrected are redistributed according to the reduced review weight to obtain dynamic review weight.

5. The method for managing digital art competitions using big data according to claim 1, characterized in that, The process involves collecting full-process status data during the event, classifying and aggregating the data according to preset data dimensions, generating a dimensional view recording the event's progress, and then pushing this dimensional view to the management terminal for display. Specifically, this includes: During the competition, data on the status of work evidence storage, task allocation, and review weight were collected from the evidence registration module, task allocation module, and weight correction module, respectively. The collected data was then uniformly formatted into status records with timestamps. The status records are categorized according to the dimensions of works, judges, and time. The number of completed and ongoing works under each dimension is counted to generate a progress view of works, a load view of judges, and a timeline view of the competition. The project progress view, the judges' workload view, and the competition timeline view are combined into a competition status view, which is then pushed to the management terminal in the form of a visual chart.

6. The method for managing digital art competitions using big data according to claim 2, characterized in that, In response to the competition launch command, the process involves performing a hash operation on the file content of the submitted entries to generate a digital fingerprint, and then attaching a timestamp and participant identifier to the digital fingerprint. Specifically, this includes: In response to the competition launch command, the system reads the original file data of the entries and divides the original file data into multiple data blocks of fixed byte length. Perform hashing on each data block to obtain a block hash value. Concatenate the block hash values ​​according to the original order of the data blocks and perform hashing again to obtain the digital fingerprint of the entry. Obtain the current system time as a timestamp, read the participant's unique registration identifier as a participant identifier, and associate and store the digital fingerprint, timestamp, and participant identifier as a record.

7. The method for managing digital art competitions using big data according to claim 2, characterized in that, The process involves broadcasting the current block to the distributed network, where distributed nodes verify the current block sequentially based on its timestamp compared to the previous block. Once more than half of the distributed nodes confirm the verification, the current block is appended to the end of the chain structure, completing the registration and certification of the entry. This process specifically includes: The assembled current block is broadcast to all online nodes in the distributed network. After receiving the current block, each online node extracts the timestamp of the current block and the hash value of the previous block. Each online node compares the timestamp of the current block with the timestamp of the latest block in its own stored chain structure to see if the timestamp of the current block is later than the timestamp of the latest block, and whether the hash value of the previous block is consistent with the hash value of the latest block. If all comparison results are yes, each online node returns an acknowledgment signal to the broadcast source. When the number of acknowledgment signals received by the broadcast source exceeds half of the total number of online nodes, the current block is written to the tail of the chain structure, and an append completion notification is sent to all online nodes.

8. The method for managing digital art competitions using big data according to claim 3, characterized in that, The extraction of successfully matched art style tags specifically includes: The descriptive text of the entries is extracted from the registered entries and then segmented according to punctuation marks and stop words to obtain a set of candidate words. Read each tag in the preset art style tag library one by one, perform a character-complete match between each tag and the candidate words in the candidate word set, and mark the tag as a successfully matched art style tag when a character-complete match is successful. The successfully matched art style tags are sorted according to their first appearance in the work description text, forming an ordered sequence of art style tags.

9. The method for managing digital art competitions using big data according to claim 3, characterized in that, The process of arranging the extracted art style tags according to their position in the work description text, and then quantifying and encoding the arranged art style tags to generate work feature codes, specifically includes: Read the binary code corresponding to each art style tag in the preset encoding table according to the order of the tags in the art style tag sequence; The binary codes corresponding to each art style tag are concatenated end to end in the order of arrangement to generate a concatenated binary sequence as the initial feature code; The code length of the initial feature code is checked to see if it reaches the preset fixed code length. If the code length of the initial feature code is less than the fixed code length, padding is added to the end of the initial feature code until the fixed code length is reached. If the code length of the initial feature code is greater than the fixed code length, a segment of fixed code length is extracted from the first part of the initial feature code as the work's feature code.

10. A big data digital art competition management system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the big data digital art competition management method according to any one of claims 1 to 9.

Images