A heterogeneous multi-modal data format synchronization method, system and terminal

Abstract

The application relates to a heterogeneous multi-modal data format synchronization method, system and terminal, and belongs to the technical field of data processing. The heterogeneous multi-modal data format synchronization method comprises the following steps: receiving heterogeneous multi-modal service data; constructing a data fingerprint for each piece of service data; analyzing the association between service data according to the data fingerprint, and establishing an association set; generating all feasible synchronization paths according to the association set, the source and format of the service data; calculating the synchronization cost corresponding to the feasible synchronization paths; selecting the feasible synchronization path with the minimum synchronization cost as a target synchronization path; calling a synchronization template from a template library according to the target synchronization path; and filling the heterogeneous multi-modal service data in the synchronization template to realize format synchronization. The application has the beneficial effects of improving the synchronization efficiency of heterogeneous multi-modal data and reducing resource consumption.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of data processing, in particular to a heterogeneous multi-modal data format synchronization method, system and terminal. BACKGROUND

[0002] With the rapid development of information technology, heterogeneous multi-modal data is increasingly widely used in various fields. Heterogeneous multi-modal data refers to data from different data sources, with different formats and characteristics, covering text, images, audio, video and other forms. These data contain rich information and have important value for enterprise decision-making, scientific research, intelligent services and the like. For example, in the medical field, electronic medical records, medical images, test reports and other heterogeneous multi-modal data can provide comprehensive patient information for doctors to assist in diagnosis and treatment; in the financial field, transaction records, market quotations, customer feedback and other data are helpful for risk assessment and investment decision-making. However, due to the heterogeneity and multi-modality of data, how to effectively integrate and synchronize these data has become a problem to be solved.

[0003] In the past technology, in order to solve the problem of synchronization of heterogeneous multi-modal data, the following means are usually adopted. One common method is to directly compare the original data. By writing specific program codes, the differences between different data are compared line by line and column by column to determine the association between the data. This method is suitable for cases where the data scale is small and the data structure is relatively simple. Another method is to use manual annotation. Professional personnel analyze and annotate business data to clearly define the relationship and rules between data. Although this method is highly accurate, it is inefficient and easily affected by human factors. Another method is to synchronize data based on preset rules. According to the rules and conditions set in advance, the data is filtered, converted and merged to achieve data synchronization. This method is more suitable for scenarios where rules are clear and data changes are small.

[0004] The existing technology has obvious defects. The traditional association analysis method needs to compare a large amount of original data, and the analysis efficiency is extremely low. Moreover, when facing multiple synchronization schemes, there is no effective method to filter out the optimal solution, resulting in large resource consumption in the data synchronization process, such as a large amount of occupation of computing resources, storage resources and network bandwidth, and low synchronization efficiency and long synchronization time. SUMMARY

[0005] In order to improve the synchronization efficiency of heterogeneous multi-modal data and reduce resource consumption, the application provides a heterogeneous multi-modal data format synchronization method, system and terminal.

[0006] In a first aspect, the application provides a heterogeneous multi-modal data format synchronization method, which adopts the following technical solution:

[0007] A heterogeneous multi-modal data format synchronization method, comprising:

[0008] receiving heterogeneous multi-modal service data;

[0009] constructing a data fingerprint for each piece of service data;

[0010] analyzing the association between service data according to the data fingerprint, and establishing an association set;

[0011] generating all feasible synchronization paths according to the association set, the source and format of the service data;

[0012] calculating the synchronization cost corresponding to the feasible synchronization paths;

[0013] selecting the feasible synchronization path with the minimum synchronization cost as the target synchronization path;

[0014] According to the target synchronization path, a synchronization template is retrieved from a template library;

[0015] filling the heterogeneous multi-modal service data in the synchronization template to realize format synchronization.

[0016] By adopting the above technical solution, the internal relationship between heterogeneous multi-modal service data can be quickly and accurately identified by constructing data fingerprints to analyze the association between service data and establish an association set. Compared with traditional association analysis methods, the use of data fingerprints greatly improves the analysis efficiency and avoids complex comparison of a large amount of original data, providing a solid foundation for subsequent data synchronization. Generating all feasible synchronization paths and calculating the synchronization cost, and finally selecting the path with the minimum cost as the target synchronization path, can filter out the optimal solution among multiple synchronization schemes, not only reducing resource consumption such as computing resources, storage resources and network bandwidth during data synchronization, but also improving synchronization efficiency and reducing synchronization time. According to the target synchronization path, a synchronization template is retrieved from a template library, and the heterogeneous multi-modal service data is filled into the template to realize format synchronization. The templating operation makes the data synchronization process more standardized and standardized, improving the accuracy and consistency of data synchronization.

[0017] Optionally, the step of constructing a data fingerprint for each piece of service data comprises:

[0018] identifying the data type of each piece of service data;

[0019] According to the data type, the corresponding feature algorithm is called to extract the key features of the service data;

[0020] The key features are converted into fixed-length data fingerprints through a hash algorithm.

[0021] Optionally, the step of analyzing the association between the service data according to the data fingerprints and establishing an association set comprises:

[0022] calculating the data similarity between the data fingerprints;

[0023] when the data similarity exceeds a similarity threshold, marking the corresponding data fingerprint as an association fingerprint, the association fingerprint indicating that there is an association between the corresponding service data;

[0024] constructing an initial association set according to the association fingerprints;

[0025] calculating the confidence of the association fingerprints in the initial association set;

[0026] eliminating the association fingerprints in the initial association set with a confidence less than a confidence threshold to generate a final association set.

[0027] By using the above technical solution, through the double screening of "similarity + confidence", the relevance (similarity) and stability score (confidence) of the association are considered, thereby reducing the false association.

[0028] Optionally, the step of calculating the confidence of the association fingerprint comprises:

[0029] finding a historical association fingerprint matching the association fingerprint in a set time window from a historical database;

[0030] obtaining the matching number of the historical association fingerprint;

[0031] obtaining the set number of the historical association set corresponding to each historical association fingerprint;

[0032] obtaining the authority value of the association fingerprint;

[0033] calculating the confidence of the association fingerprint according to the matching number, the set number and the authority value.

[0034] By employing the aforementioned technical solution, historical fingerprints that match the associated fingerprint within a set time window are retrieved from the historical database, and the number of matches is obtained. This step utilizes information from historical data. In practical applications, historical data can reflect the past performance and patterns of associated fingerprints. By analyzing the number of matched associated fingerprints in historical data, the frequency and stability score of the associated fingerprint can be understood. Obtaining the number of historical association sets corresponding to each historical associated fingerprint further enriches the evaluation dimensions. Different associated fingerprints may be associated with different numbers of association sets; the number of association sets reflects the complexity and influence of the associated fingerprint. By considering the number of historical association sets, the value and reliability of the associated fingerprint can be evaluated more comprehensively, avoiding evaluation bias caused by ignoring association set information. Obtaining the authority value of the associated fingerprint and combining it with the number of matches and the number of sets to calculate the confidence level, by introducing an objective measurement standard, reduces the interference of subjective factors and makes the evaluation results more objective and fair. Combining the number of matches, the number of sets, and the authority value to calculate the confidence level of the associated fingerprint avoids the limitations of single-factor evaluation. Different factors may have different levels of importance in the evaluation process. By taking these factors into account, the true confidence level of associated fingerprints can be reflected more accurately.

[0035] Optionally, the step of calculating the synchronization cost corresponding to the feasible synchronization path includes:

[0036] Obtain the synchronization duration, data loss rate, and resource consumption rate for each feasible synchronization path;

[0037] Based on the synchronization duration, the data loss rate, and the resource consumption rate, calculate the evaluation value corresponding to each feasible synchronization path;

[0038] Obtain the reliability of each of the aforementioned feasible synchronization paths;

[0039] Based on the evaluation value and the reliability, calculate the synchronization cost corresponding to the feasible synchronization path.

[0040] By adopting the technical solution, when calculating the synchronization cost, the synchronization duration, data loss rate and resource consumption rate of each feasible synchronization path are obtained, and evaluation values are calculated according to these indexes, and multiple key factors in the synchronization process are comprehensively considered. For example, the synchronization duration reflects the efficiency of the synchronization operation, the data loss rate relates to the integrity of the data, and the resource consumption rate reflects the occupation of the system resources. Different synchronization paths may have different performances on these indexes, and by comprehensively calculating the evaluation values of these indexes, each feasible synchronization path can be more comprehensively and objectively evaluated, thereby providing a scientific basis for selecting the optimal path. The reliability of each feasible synchronization path is obtained, and the synchronization cost is calculated in combination with the evaluation values, thereby further improving the evaluation of the synchronization path, avoiding selecting a synchronization path that appears to be good but actually has low reliability, and improving the accuracy and reliability of the selection.

[0041] Optionally, the step of obtaining the reliability of each feasible synchronization path comprises:

[0042] obtaining current environment data;

[0043] According to the current environment data, obtaining an environment adaptability score corresponding to each feasible synchronization path;

[0044] According to historical synchronization data, obtaining a success rate corresponding to each feasible synchronization path;

[0045] obtaining a stability score corresponding to each feasible synchronization path;

[0046] According to the environment adaptability score, the success rate and the stability score, calculating the reliability of each feasible synchronization path.

[0047] By adopting the technical solution, by obtaining the current environment data and obtaining the environment adaptability score corresponding to each feasible synchronization path, the influence of the actual situation on the synchronization path can be fully considered. The success rate corresponding to each feasible synchronization path is obtained, which is based on the past performance to evaluate the reliability. The historical data can reflect the stability score and effectiveness of the synchronization path under different conditions. The stability score corresponding to each feasible synchronization path is obtained, which can evaluate the fluctuation in the synchronization process. The synchronization path with high stability score means that the possibility of abnormality in the synchronization process is small, and the synchronization task can be more reliably completed. According to the environment adaptability score, the success rate and the stability score, the reliability of each feasible synchronization path is calculated, which comprehensively considers multiple important factors, and can more comprehensively evaluate the quality of the synchronization path. The evaluation of a single factor may have limitations, and the comprehensive evaluation can make up for this deficiency.

[0048] Optionally, the step of retrieving a synchronization template from the template library according to the target synchronization path comprises:

[0049] finding a historical target synchronization path matching the target synchronization path;

[0050] calling a historical synchronization template associated with the historical target synchronization path;

[0051] obtaining a score value of the historical synchronization template;

[0052] determining whether there is the historical synchronization template with a score value exceeding a set threshold value;

[0053] if yes, directly calling the historical synchronization template with the maximum score value as the synchronization template for this time;

[0054] if no, finding a latest synchronization template matching the target synchronization path from the template library.

[0055] By using the above technical solution, the historical target synchronization path matching the target synchronization path is found, and the historical synchronization template associated therewith is called, so that the previous synchronization experience is fully utilized. In actual application, similar synchronization paths may have similar synchronization requirements, so that it is possible to avoid searching the matching template from the template library every time, and the efficiency of template selection is greatly improved. The score value of the historical synchronization template is obtained, and it is determined whether there is the historical synchronization template with a score value exceeding a set threshold value. The score value reflects the performance and effect of the template in the previous use process, and the template with a high score value often performs better in terms of accuracy, stability score, etc. The historical synchronization template with the maximum score value is directly called as the synchronization template for this time, so that it is possible to ensure that the template with high quality is selected, and the accuracy and reliability of synchronization are improved. By preferentially selecting the historical synchronization template with a high score value or the latest synchronization template, it is possible to reduce the trial and error cost caused by selecting an unsuitable template.

[0056] In a second aspect, the application provides a heterogeneous multi-modal data format synchronization system, which adopts the following technical solution:

[0057] A heterogeneous multi-modal data format synchronization system comprises:

[0058] a data receiving module configured to receive heterogeneous multi-modal service data;

[0059] a data processing module configured to construct a data fingerprint for each piece of service data, analyze the association between service data according to the data fingerprint, and establish an association set;

[0060] a synchronization path processing module configured to generate all feasible synchronization paths according to the association set, the source and format of service data, calculate the synchronization cost corresponding to the feasible synchronization paths, and select the feasible synchronization path with the minimum synchronization cost as a target synchronization path;

[0061] a data format synchronization module, configured to call a synchronization template from a template library according to the target synchronization path, and fill the heterogeneous multi-modal service data into the synchronization template to realize format synchronization.

[0062] In a third aspect, the present application provides a terminal, which adopts the following technical solution:

[0063] A terminal comprises:

[0064] a memory, which stores a heterogeneous multi-modal data format synchronization program;

[0065] a processor, configured to execute the program stored on the memory to realize the steps of the above-mentioned heterogeneous multi-modal data format synchronization method.

[0066] In summary, the present application has at least the following beneficial effects:

[0067] By constructing data fingerprints to analyze the association between service data and establish an association set, the internal relationship between heterogeneous multi-modal service data can be quickly and accurately identified. Compared with traditional association analysis methods, the use of data fingerprints greatly improves the analysis efficiency and avoids complex comparison of a large number of original data, thereby providing a solid foundation for subsequent data synchronization. All feasible synchronization paths are generated and the synchronization cost is calculated, and finally the path with the minimum cost is selected as the target synchronization path, which can filter out the optimal solution from multiple synchronization schemes, not only reducing the resource consumption in the data synchronization process, such as computing resources, storage resources and network bandwidth, but also improving the synchronization efficiency and reducing the synchronization time. According to the target synchronization path, a synchronization template is called from a template library, and the heterogeneous multi-modal service data is filled into the template to realize format synchronization. The templating operation makes the data synchronization process more standardized and normalized, thereby improving the accuracy and consistency of data synchronization. BRIEF DESCRIPTION OF DRAWINGS

[0068] Figure 1 is a first flowchart of the method embodiment of the present application;

[0069] Figure 2 is a second flowchart of the method embodiment of the present application;

[0070] Figure 3 is a third flowchart of the method embodiment of the present application;

[0071] Figure 4 is a fourth flowchart of the method embodiment of the present application;

[0072] Figure 5 is a fifth flowchart of the method embodiment of the present application;

[0073] Figure 6is a sixth flow chart of the method embodiment of the present application. DETAILED DESCRIPTION

[0074] To make the purposes, technical solutions, and advantages of the embodiments of the present application clearer, the following will combine the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application. Figure 1 - the drawings Figure 6 To make the purposes, technical solutions, and advantages of the embodiments of the present application clearer, the following will combine the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0075] The first embodiment of the present application discloses a heterogeneous multi-modal data format synchronization method. Referring to Figure 1 The heterogeneous multi-modal data format synchronization method can include S110-S180:

[0076] S110, receiving heterogeneous multi-modal service data;

[0077] S120, constructing a data fingerprint for each piece of service data;

[0078] S130, analyzing the association between the service data according to the data fingerprint, and establishing an association set;

[0079] S140, generating all feasible synchronization paths according to the association set, the source, and the format of the service data;

[0080] S150, calculating the synchronization cost corresponding to the feasible synchronization path;

[0081] S160, selecting the feasible synchronization path with the minimum synchronization cost as the target synchronization path;

[0082] S170, according to the target synchronization path, calling a synchronization template from a template library;

[0083] S180, filling the heterogeneous multi-modal service data in the synchronization template to realize format synchronization.

[0084] Specifically, a receiving service based on a network interface can be built, for example, a web service is created using the Flask framework of Python, and the client can send text, image, and audio data to the receive_data interface through a POST request.

[0085] The specific steps of S120, constructing a data fingerprint for each piece of service data, can include:

[0086] identifying the data type of each piece of service data;

[0087] According to the data type, the corresponding feature algorithm is called to extract the key features of the business data;

[0088] The key features are converted into fixed-length data fingerprints through a hash algorithm.

[0089] Specifically, data type identification can be performed through file extension combined with file header information verification. For example, for text data, common extensions are.txt,.csv,.json, etc., and the file header usually has no special identifier; for image data, the.jpg file header is FF D8 FF, and the.png file header is 89 50 4E 47; for audio data, the.mp3 file header contains an ID3 identifier, and the.wav file header is RIFF format header 12.

[0090] For text data, the TF-IDF value of each word in the text is first calculated, and the keywords with high TF-IDF values are selected. These keywords and their TF-IDF values are combined into a vector, and then the vector is converted into a fixed-length fingerprint through dimension reduction techniques such as principal component analysis. For image data, the relative brightness relationship between each pixel point and its neighborhood pixels in the image can be calculated to generate LBP codes. The LBP code histogram of the image is taken as the feature vector of the image, and then the feature vector is converted into a fingerprint through a hash algorithm. For audio data, the audio signal can be converted into frequency domain features (such as Mel Frequency Cepstral Coefficients, MFCC), and the MFCC features can be processed and quantized to generate a fixed-length hash value as an audio fingerprint.

[0091] Referring to Figure 2 For S130, according to the data fingerprints, the step of analyzing the association between the business data and establishing the association set includes S210-S250:

[0092] S210, calculating the data similarity between the data fingerprints;

[0093] S220, when the data similarity exceeds the similarity threshold, marking the corresponding data fingerprint as an association fingerprint, which indicates that there is an association between the corresponding business data;

[0094] S230, constructing an initial association set according to the association fingerprints;

[0095] S240, calculating the confidence of the association fingerprints;

[0096] S250, removing the association fingerprints in the initial association set with a confidence less than a confidence threshold to generate a final association set.

[0097] Specifically, the cosine similarity can be used to calculate the data similarity between the data fingerprints corresponding to different types of business data. For two data fingerprint vectors and , the cosine similarity calculation formula is: wherein, are the values of the i-th dimension of the vectors and , and n is the dimension of the vector. The closer the value of the cosine similarity is to 1, the more similar the two data fingerprints are.

[0098] All the calculated data similarity values are traversed, and a conditional judgment statement (such as an if statement in Python) is used. When a certain data similarity value is greater than a set threshold value, for example, 0.8, the corresponding two data fingerprints are marked as associated fingerprints. A set (Set) data structure is used to construct an initial association set. For each pair of associated fingerprints, they are added to the set as an element.

[0099] Referring to Figure 3 , for S230, the specific steps of calculating the confidence of the associated fingerprint can include S310-S350:

[0100] S310, find the historical associated fingerprints matching the associated fingerprint in a set time window from the historical database;

[0101] S320, obtain the matching number of the historical associated fingerprints;

[0102] S330, obtain the set number of the historical association set corresponding to each historical associated fingerprint;

[0103] S340, obtain the authority value of the associated fingerprint;

[0104] S350, calculate the confidence of the associated fingerprint according to the matching number, the set number, and the authority value.

[0105] Specifically, the associated fingerprint is stored in the form of a "fingerprint pair" (such as (F1, F2)), and the historical database table contains fields for storing the associated fingerprint pair, recording the generation time, etc. The historical data can be filtered by Structured Query Language (SQL) combined with a time window condition. The time window can be dynamically adjusted according to business requirements (such as the last 30 days, the last 1 year, etc.). Based on the query result of S310, the total number of records that meet the conditions is counted. In SQL, the COUNT() function can be directly used. If the return result is match_count = 150, it means that the associated fingerprint matches 150 historical records in the time window.

[0106] Each historical associated fingerprint belongs to an association set, and the association set ID is stored in the association_set_id field. The set number can be calculated using the DISTINCT keyword. If the result is set_count = 50, it means that the associated fingerprint is distributed in 50 different historical association sets.

[0107] The calculation of the authority value can be based on the business data source reliability score and the business data compliance level score; the authority value=(source reliability score*0.6)+(compliance score*0.4).

[0108] The rules of the business data source reliability score are as follows:

[0109] If the business data is from an official database (such as an enterprise core system), the corresponding reliability score is 0.9-1; if the business data is from a partner API interface, the corresponding reliability score is 0.7-0.8; if the business data is from public crawler data, the corresponding reliability score is 0.4-0.6; if the business data is from unknown source data, the corresponding reliability score is 0.1-0.3. If the data corresponding to the association fingerprint is from multiple channels, the weighted average value of the scores of each channel is taken.

[0110] The rules of the business data compliance level score are as follows:

[0111] If the business data is completely compliant, the corresponding score is 0.9-1; if the business data is partially compliant, the corresponding score is 0.5-0.8; if the business data is completely non-compliant, the corresponding score is 0-0.4.

[0112] For S350, the confidence level is the product of (the number of matches / (set number+1)) and the authority value, and the calculation result is normalized to [0,1].

[0113] For S140, according to the association set, the source and format of the business data, the specific way to generate all feasible synchronization paths is:

[0114] For each piece of business data in the association set, key features are extracted from its source and format. Source features include the system, device, department, etc. where the data is generated; format features cover file types (such as CSV, JSON, XML), data structures (such as table type, hierarchical type), encoding methods, etc. According to the extracted features, the business data is classified. For example, data from the same system and with the same format are classified into one category, which can simplify the subsequent generation process of the synchronization path. For different data formats, corresponding conversion rules are defined in advance. Combined with the data source, synchronization rules compatible with different sources are developed. Each piece of business data in the association set is taken as a potential starting point, and the target format and target source to which synchronization is needed are taken as target points. Based on the defined synchronization rules, from the starting point, all possible synchronization steps are searched through recursion or iteration until the target point is reached. Each complete step sequence is a feasible synchronization path. The generated synchronization paths are checked, and the paths that do not conform to the synchronization rules or cannot be implemented are removed.

[0115] Reference Figure 4For S150, the specific steps of calculating the synchronization cost corresponding to the feasible synchronization path can include S410-S440:

[0116] S410, obtaining the synchronization duration, data loss rate and resource consumption rate of each feasible synchronization path;

[0117] S420, calculating the evaluation value corresponding to each feasible synchronization path according to the synchronization duration, data loss rate and resource consumption rate;

[0118] S430, obtaining the reliability of each feasible synchronization path;

[0119] S440, calculating the synchronization cost corresponding to the feasible synchronization path according to the evaluation value and the reliability.

[0120] Specifically, the historical synchronization records are called to count the average time consumption of the same or similar synchronization path as a reference. If a path contains the step of "text to speech -> speech compression", the execution time of multiple times of the combination step in the past can be found, and the average value is taken as the estimated duration of the current path. The checksum (such as MD5) of the data is calculated before and after synchronization respectively, and the data integrity is determined by comparing the checksum difference. If the checksum of the original data is A and the checksum of the data after synchronization is B, the data loss rate = (1- same checksum data amount / total data amount) x 100%. For example, a piece of audio has 100 data packets before synchronization, and 98 packets have consistent checksum after synchronization, then the loss rate is 2%. The resource optimization and configuration module of the project management system is used to monitor the resource indicators such as CPU utilization, memory occupation, network bandwidth consumption and the like in the synchronization process in real time. The resource consumption rate can be defined as (actual resource usage / maximum available resource of the system) x 100%. The weight factors of the synchronization duration (T), the data loss rate (L) and the resource consumption rate (R) are set, and the sum of the weights is 1. The indexes are standardized to eliminate the influence of dimension. For example, the duration standardized value = synchronization duration / maximum possible duration, and the loss rate and consumption rate are already in percentage form. The evaluation value = w1 x duration standardized value + w2 x data loss rate + w3 x resource consumption rate.

[0121] Referring to Figure 5 For S430, the steps of obtaining the reliability of each feasible synchronization path include S510-S550:

[0122] S510, obtaining the current environment data;

[0123] S520, obtaining the environment adaptability score corresponding to each feasible synchronization path according to the current environment data;

[0124] S530, obtaining the success rate corresponding to each feasible synchronization path according to the historical synchronization data;

[0125] S540, obtaining a stability score corresponding to each feasible synchronization path;

[0126] S550, calculating a reliability of each feasible synchronization path according to the environmental adaptability score, the success rate and the stability score.

[0127] Specifically, the server hardware indicators are collected by a system monitoring tool (such as Prometheus, Zabbix), including CPU usage (such as current load 20%), memory occupation (8GB / 16GB), network bandwidth (uplink / downlink rate 50Mbps / 100Mbps) and delay (average 10ms). The format attributes of the data to be synchronized (such as image resolution 1920x1080, text encoding UTF-8, audio sampling rate 44.1kHz), data size (single file 200MB) and real-time requirements are scanned. The status of third-party services is queried, such as API interface response time (such as OCR recognition interface average 300ms), database connection stability score (continuous 3 times ping test all successful) and permission authentication validity (Token remaining validity period 2 hours). The preset environmental threshold of each synchronization link (such as format conversion, cross-system transmission) in the path is compared with the current environmental data to generate an adaptability score of 0-1. According to the influence weight of each link on the synchronization result (such as transmission link weight 0.4, conversion link 0.6), the weighted sum is obtained to get the total score of the path environmental adaptability score.

[0128] From the synchronization log database, the historical records of the same path structure (such as "text OCR recognition -> JSON format conversion -> cloud storage transmission") within the last 3 months are extracted. The success rate = (number of successful times ÷ total number of attempts) x 100%. In addition, if the historical data is insufficient (such as <100), the industry average success rate of similar paths can be introduced as a supplement.

[0129] Based on the historical synchronization data, the mean and standard deviation of the core indicators (such as synchronization time) in the feasible synchronization path are calculated, and the fluctuation coefficient is calculated, the fluctuation coefficient = standard deviation / synchronization time mean, the fluctuation coefficient is mapped to a stability score of 0-1, such as the fluctuation coefficient ≤0.1, the score is 1.0, 0.1-0.3 linearly reduced, >0.3, the score is 0.

[0130] Reliability = environmental adaptability score total score x 0.3 + success rate x 0.5 + stability score x 0.2, wherein the weight can be adjusted according to the business scenario.

[0131] For S440, the way to calculate the synchronization cost corresponding to the feasible synchronization path according to the evaluation value and the reliability is: synchronization cost = evaluation value x (1-reliability).

[0132] Reference Figure 6For S170, the specific steps of retrieving the synchronization template from the template library according to the target synchronization path can include S610-S660:

[0133] S610, searching for a historical target synchronization path matching the target synchronization path;

[0134] S620, retrieving a historical synchronization template associated with the historical target synchronization path;

[0135] S630, obtaining a score value of the historical synchronization template;

[0136] S640, determining whether there is a historical synchronization template with a score value exceeding a set threshold;

[0137] S650, if yes, directly invoking the historical synchronization template with the largest score value as the synchronization template for this time;

[0138] S660, if no, searching for the latest synchronization template matching the target synchronization path from the template library.

[0139] Specifically, a historical target synchronization path database is established in advance, which records all past synchronization path information. When receiving the target synchronization path, a string matching algorithm (such as regular expression matching) is used to match each historical target synchronization path in the database. An association table is established in advance, which records the mapping relationship between the historical target synchronization path and the historical synchronization template. When the matching historical target synchronization path is found, the associated historical synchronization template is obtained by querying the association table. A score database is established in advance, which records the score value of each historical synchronization template. When the historical synchronization template is obtained, the corresponding score value is obtained by querying the score database. The obtained score value list is traversed to determine whether there is a score value exceeding the set threshold. If there is a historical synchronization template with a score value exceeding the set threshold, the template with the largest score value among these templates is found.

[0140] In addition, the creation time and associated synchronization path of each synchronization template are recorded in the template library. When there is no historical synchronization template with a score value exceeding the set threshold, the template library is traversed to find the synchronization template matching the target synchronization path and having the latest creation time, and then the heterogeneous multi-modal business data is filled in the synchronization template to realize format synchronization; the matching refers to the conversion requirement of the target synchronization path.

[0141] An implementation scenario of the embodiment is as follows:

[0142] After receiving the heterogeneous multi-modal business data, the data type of each business data is identified, and according to the data type, the corresponding feature algorithm is called to extract the key features of the business data, and then the key features are converted into fixed-length data fingerprints through a hash algorithm.

[0143] The data similarity between different types of data fingerprints is calculated, and the corresponding data fingerprint is marked as an associated fingerprint when the data similarity exceeds a similarity threshold; then, an initial association set is constructed according to the associated fingerprints, and the confidence of the associated fingerprints is calculated, and the associated fingerprints with a confidence less than a confidence threshold in the initial association set are removed to generate a final association set.

[0144] Then, all feasible synchronization paths are generated according to the association set, the sources and formats of the business data, the synchronization duration, the data loss rate and the resource consumption rate of each feasible synchronization path are obtained, the reliability of each feasible synchronization path is obtained, and the synchronization cost of the corresponding feasible synchronization path is calculated according to the evaluation value and the reliability, and the feasible synchronization path with the minimum synchronization cost is selected as the target synchronization path.

[0145] A historical target synchronization path matching the target synchronization path is searched, and a historical synchronization template associated with the historical target synchronization path is called, then the score value of the historical synchronization template is obtained, it is judged whether there is a historical synchronization template with a score value exceeding a set threshold, if yes, the historical synchronization template with the maximum score value is directly called as the synchronization template this time, and if not, the latest synchronization template matching the target synchronization path is searched from the template library. The heterogeneous multi-modal business data is filled in the synchronization template to realize format synchronization.

[0146] Based on the above method embodiments, the second embodiment of the present application discloses a heterogeneous multi-modal data format synchronization system. The heterogeneous multi-modal data format synchronization system of the embodiment of the present application can implement any of the above-mentioned heterogeneous multi-modal data format synchronization methods, and the specific working process of each module in the heterogeneous multi-modal data format synchronization system can refer to the corresponding process in the above-mentioned method embodiments.

[0147] For ease of understanding, an example is as follows: a heterogeneous multi-modal data format synchronization system, comprising:

[0148] A data receiving module is configured to receive the heterogeneous multi-modal business data.

[0149] A data processing module is configured to construct a data fingerprint for each piece of business data, and analyze the association between the business data according to the data fingerprint, and establish an association set.

[0150] A synchronization path processing module is configured to generate all feasible synchronization paths according to the association set, the sources and formats of the business data, and calculate the synchronization cost of the feasible synchronization paths, and select the feasible synchronization path with the minimum synchronization cost as the target synchronization path.

[0151] The data format synchronization module is configured to call a synchronization template from the template library according to a target synchronization path, fill the heterogeneous multi-modal service data into the synchronization template, and realize format synchronization.

[0152] The third embodiment of the present application provides a terminal, which can include a memory and a processor, and as an embodiment of the terminal,

[0153] The memory is configured to store a heterogeneous multi-modal data format synchronization program.

[0154] The processor is configured to execute the program stored on the memory to implement the steps of the above-mentioned heterogeneous multi-modal data format synchronization method.

[0155] The memory can be in communication connection with the processor through a communication bus, and the communication bus can be an address bus, a data bus, a control bus, etc.

[0156] In addition, the memory can include a random access memory (RAM) and can also include a non-volatile memory (NVM), such as at least one disk memory.

[0157] The processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc., and can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component, etc.

[0158] The above are preferred embodiments of the present application, and are not used to limit the protection scope of the present application. Any feature disclosed in the present specification (including the abstract and the drawings) can be replaced by other equivalent or similar features unless specifically stated. That is, each feature is only an example of a series of equivalent or similar features unless specifically stated.

Claims

1. A method for synchronizing heterogeneous multimodal data formats, characterized in that, include: Receive heterogeneous multimodal service data; Build a data fingerprint for each piece of business data; Based on the data fingerprint, analyze the relationships between business data and establish a set of relationships; Based on the association set, the source and format of the business data, generate all feasible synchronization paths; Calculate the synchronization cost corresponding to the feasible synchronization path; Select the feasible synchronization path with the lowest synchronization cost as the target synchronization path; Based on the target synchronization path, retrieve the synchronization template from the template library; The heterogeneous multimodal service data is filled into the synchronization template to achieve format synchronization; The step of calculating the synchronization cost corresponding to the feasible synchronization path includes: Obtain the synchronization duration, data loss rate, and resource consumption rate for each feasible synchronization path; Based on the synchronization duration, the data loss rate, and the resource consumption rate, calculate the evaluation value corresponding to each feasible synchronization path; Obtain the reliability of each of the aforementioned feasible synchronization paths; Based on the evaluation value and the reliability, calculate the synchronization cost corresponding to the feasible synchronization path; The step of obtaining the reliability of each of the feasible synchronization paths includes: Obtain current environmental data; Based on the current environmental data, obtain the environmental adaptability score corresponding to each feasible synchronization path; Based on historical synchronization data, obtain the success rate corresponding to each feasible synchronization path; Obtain the stability score corresponding to each of the feasible synchronization paths; The reliability of each feasible synchronization path is calculated based on the environmental adaptability score, the success rate, and the stability score.

2. The heterogeneous multimodal data format synchronization method according to claim 1, characterized in that, The steps for constructing a data fingerprint for each piece of business data include: Identify the data type of each piece of business data; Based on the data type, the corresponding feature algorithm is invoked to extract the key features of the business data; The key features are converted into a fixed-length data fingerprint using a hash algorithm.

3. The heterogeneous multimodal data format synchronization method according to claim 1, characterized in that, The step of analyzing the correlations between business data based on the data fingerprint and establishing a correlation set includes: Calculate the data similarity between the data fingerprints; When the data similarity exceeds a similarity threshold, the corresponding data fingerprint is marked as an associated fingerprint, which indicates that there is an association between the corresponding business data. Based on the associated fingerprints, construct an initial association set; Calculate the confidence level of the associated fingerprints in the initial association set; Remove association fingerprints with confidence scores less than a confidence threshold from the initial association set to generate the final association set.

4. The heterogeneous multimodal data format synchronization method according to claim 3, characterized in that, The step of calculating the confidence level of the associated fingerprint includes: Search the historical database for historical fingerprints that match the associated fingerprint within a set time window; Obtain the number of matches for the historical associated fingerprints; Obtain the number of sets of historical associations corresponding to each of the historical association fingerprints; Obtain the authority value of the associated fingerprint; The confidence level of the associated fingerprint is calculated based on the number of matches, the number of sets, and the authority value.

5. The heterogeneous multimodal data format synchronization method according to claim 1, characterized in that, The step of retrieving the synchronization template from the template library according to the target synchronization path includes: Find a historical target synchronization path that matches the target synchronization path; Retrieve the historical synchronization template associated with the historical target synchronization path; Obtain the score value of the historical synchronization template; Determine whether there is a historical synchronization template whose score exceeds a set threshold; If so, the historical synchronization template with the highest score will be used directly as the synchronization template for this time; If not, then search the template library for the latest synchronization template that matches the target synchronization path.

6. A heterogeneous multimodal data format synchronization system, characterized in that, Performing the heterogeneous multimodal data format synchronization method as described in any one of claims 1-5 includes: The data receiving module is used to receive heterogeneous multimodal service data; The data processing module is used to construct a data fingerprint for each piece of business data, and to analyze the correlation between business data based on the data fingerprint and establish a correlation set. The synchronization path processing module is used to generate all feasible synchronization paths based on the association set, the source and format of the business data, calculate the synchronization cost corresponding to the feasible synchronization path, and select the feasible synchronization path with the lowest synchronization cost as the target synchronization path. The data format synchronization module is used to retrieve a synchronization template from the template library according to the target synchronization path, and fill the heterogeneous multimodal business data into the synchronization template to achieve format synchronization.

7. A terminal, characterized in that, include: The memory stores synchronization programs for heterogeneous multimodal data formats; A processor is configured to execute a program stored in the memory to implement the steps of the heterogeneous multimodal data format synchronization method as described in any one of claims 1-5.

Images