Government affair information temporary storage type response system based on big data

Abstract

The invention discloses a government affair information temporary storage type response system based on big data. The government affair information temporary storage type response system comprises a data receiving module, a preprocessing module, a relay module, a storage module and a response module. The data receiving module receives and checks the validity and authenticity of the government affair information, and only forwards the checked government affair information to ensure the system security; the preprocessing module rejects repeated government affair information through a text similarity detection algorithm, automatically corrects spelling errors of the government affair information based on a text error correction network, unifies the format of the government affair information, autonomously gives level labels, and improves the processing efficiency of the government affair information; the relay module transfers the government affair information based on the level labels, the storage module and the response module accelerate the processing flow through the synergistic effect of the anti-delay strategy and the real-time coping strategy, and the government affair information processing efficiency is further improved.

Description

Technical Field

[0001] The present invention relates to the fields of information technology and big data processing, and in particular to a government information temporary storage response system based on big data. Background Art

[0002] At present, with the development of government digital transformation, the informatization and digitalization construction within the government has been greatly improved. The channel terminals provided to the public have developed from traditional service halls to now receiving government event reports from the public and grassroots staff through mobile apps, official websites, hotlines, public accounts, mini programs, etc. The increase in channels has also increased the amount of government information processing.

[0003] The invention patent with the existing publication number CN117350670A proposes a government affairs processing system and method, including an information release module, an information processing module, an information supervision module and an information decision module. Data is collected and published through the information collection module, and the data is stored in a data warehouse. A decision support system is established through the information decision module to analyze and make decisions on the data in the data warehouse.

[0004] Most existing government information processing systems use real-time processing methods. When faced with massive amounts of government information, they are unable to effectively filter out duplicate government information, resulting in low processing efficiency. In addition, government information uploaded through multiple channels often has problems with confusing formats and text errors, which further affects the system processing efficiency and cannot meet the needs of efficient and convenient government processing. Therefore, the present invention proposes a government information temporary storage response system based on big data to solve the above problems. Summary of the invention

[0005] The purpose of the present invention is to provide a government information temporary storage response system based on big data to pre-process and temporarily store government information, improve government processing efficiency, and meet the needs of efficient and convenient government processing.

[0006] The technical solution to achieve the purpose of the present invention is: A government information temporary storage response system based on big data, comprising a data receiving module, a preprocessing module, a relay module, a storage module and a response module; The data receiving module receives and verifies the government information, and forwards the verified government information to the pre-processing module; The pre-processing module cleans government affairs information, generates pending government affairs information and forwards it to the relay module; The relay module selects a sending storage module or a processing party based on the level label of the pending government affairs information; The storage module temporarily stores pending government affairs information based on anti-delay strategies and stores necessary information for a long time; The response module responds to the processing notification initiated by the processor based on the real-time response strategy and transmits the pending government affairs information to the processor.

[0007] Further, Government information of government departments Include department number , Reporting time , Government documents ,sign and Certificate ; Department Number Used to identify departments; Reporting time To send government information time; Governmental documents It is a text file, including the government name and government affairs ; sign By Government departments report time generate; Certificate In the It is issued by the system when registering with a government department.

[0008] Further, the data receiving module includes a verification unit and a transceiver unit; The verification unit obtains public information, and verifies the validity and authenticity of the government information based on the first equation and the second equation respectively, and the government information that satisfies the equation is deemed to have passed the verification; The transceiver unit receives government affairs information and forwards the government affairs information that has passed the inspection to the pre-processing module.

[0009] Furthermore, the verification unit verifies the government information including the following steps: Obtain government information and public information; Verify government information through the first equation effectiveness; Verify government information through the second equation authenticity; When both the first and second equations hold, government information The inspection is passed, and the rest are deemed as failed.

[0010] Further, the pre-processing module includes a de-duplication unit, an optimization unit, a rating unit, and a communication unit; The deduplication unit uses a text similarity detection algorithm to identify and remove duplicate government information; The optimization unit corrects spelling errors and standardizes government texts based on the text error correction network; The rating unit assigns level labels to government information based on the keyword detection algorithm and generates pending government information; The communication unit receives the government affairs information that has passed the inspection, sends the pending government affairs information to the relay module, and sends the supervision tag to the storage module at the same time.

[0011] Furthermore, the deduplication unit uses a text similarity detection algorithm to identify and remove duplicate government information, including the following steps: Obtain government information , based on department number Check whether the same department number already exists in the storage module Information on pending government affairs; If it does not exist, directly determine the government information No repetition; If it exists, get all the same department numbers To-do government affairs information to construct a to-do government affairs information set; Get the latest pending government affairs information from the pending government affairs information set , perform duplicate determination, and count government documents and recent government affairs to be done Length of text and ; Compare text differences Difference threshold The size relationship, difference threshold and text length and There is a fixed ratio between the larger of ; When the text is poor Greater than difference threshold , determine government information Recent government affairs information No duplication, remove the most recent pending government affairs information from the pending government affairs information set ; When the text is poor Less than or equal to the difference threshold , using the bidirectional LSTM+CRF combination model as the word segmenter, and recent government affairs to be done Perform word segmentation, obtain the word segmentation vector and the recent to-do word segmentation vector respectively, and merge them to generate a merged word segmentation set; For the first word segmentation, and calculate the Governmental texts and recent government affairs to be done Frequency of words and recent to-do word frequency , and calculate the The inverse text frequency of the word , to obtain the Governmental texts and recent government affairs to be done Weight and the weight of the most recent tasks ; Based on public information processing and merging word segmentation word segmentation, generate binary code, and based on weight and the weight of the most recent tasks Get weight encoding and the weight code of the recent to-do list ; Superimpose and merge the weight encodings of all the words in the word set and the weight code of the recent to-do list Generate superposition code Overlay coding with recent to-dos , determine the superposition coding Overlay coding with recent to-dos Is the Hamming distance of greater than the distance threshold? If the superposition coding Overlay coding with recent to-dos The Hamming distance of is greater than the distance threshold, and the government information is determined Recent government affairs information No duplication, remove the most recent pending government affairs information from the pending government affairs information set , otherwise, it is determined that government information Recent government affairs information Duplicate and remove government information And generate supervision labels to stop repeated judgments.

[0012] Execute the empty set judgment to determine whether the pending government affairs information set is an empty set. If it is an empty set, determine whether the pending government affairs information set is an empty set. If there is no duplication, then the latest to-do government affairs information is obtained from the to-do government affairs information set after the removal. Continue to perform duplicate determination and empty set determination.

[0013] Furthermore, the optimization unit corrects spelling errors of government texts based on the text error correction network, including the following steps: Constructing the Confusion Set , similar pronunciation character relationship diagram Relationship diagram of similar characters , and define the approximate character relationship graph , ; Re-representing the confusion set based on graph convolution operation and attention map combination operation The characters in the aggregate representation set are generated ; Extracting government texts through BERT , get the corresponding representation set And input the detector to calculate the government text Middle Characters The probability of error ; Get the aggregate representation set Character representation The most similar Aggregate representation, in the confusion set Index to the corresponding character and construct the candidate representation set respectively and candidate set ; Use the corrector to calculate the candidate set The candidate characters are positive examples The predicted probability ; Introduce contrast loss, maximize the prediction probability of positive examples through contrast learning, and minimize the prediction probability of negative examples, and define the final loss ,in, To detect the loss, To correct the loss, is the comparison loss.

[0014] Furthermore, the confusion set is a predefined character set, including characters, , considered Character similarities above; Characters with the same or similar pronunciation are used as graph nodes, and edges are defined and generated to build a graph of similar-pronunciation characters. , taking characters with similar shapes as graph nodes, defining and generating edges to construct a relationship graph of similar-shaped characters , and define the approximate character relationship graph ; side The definition of confusion set is Middle characters and The similarity of characters is based on characters and Character representation of characters and The distance is obtained.

[0015] Furthermore, the aggregated representation set is generated based on the graph convolution operation and the attention graph combination operation The following steps are involved: Extracting confusion sets through BERT To obtain the representation set ; Processing approximate character relationship graphs through lightweight convolutional layers Get the convolution map , Time Convolutional Graph Convolutional graph for pronunciation , Time Convolutional Graph Convolutional graph of shape ; The attention mechanism is used to combine the similarity of character pronunciation and character shape to generate an attention combination graph ; Superimposed attention combination diagram and the representation set To obtain the aggregate representation set .

[0016] Furthermore, optimizing the unit standardized government documents includes the following steps: Unified data format, dates are represented in ISO 8601 format, numbers are represented in floating point numbers, and strings are represented in UTF-8 encoding; Unified data structure: lists, sequences, and vectors are all represented by arrays, and tree structures are all represented by binary trees; Set data constraints: date cannot be empty, string cannot contain special symbols, binary tree cannot contain duplicate elements; The unified naming of government affairs texts is “department number-reporting time-government affairs name”.

[0017] Furthermore, the rating unit assigns level labels based on the keyword detection algorithm. , generating pending government affairs information includes the following steps: Obtain government information Government documents in ; Get the department number , query the department level table and obtain the initial level label ; The bidirectional LSTM+CRF combined model is used to analyze government documents Perform word segmentation and obtain word segmentation vector; Count the total number of word segmentations in the word segmentation vector and the total number of key words , calculate the increase in word segmentation vector ; Compute level tags And provide government information , generate pending government affairs information ,in, Is the highest level.

[0018] Further, the relay module includes a determination control unit and a forwarding unit; Determine the control unit to obtain pending government affairs information Level labels in , determine whether it is the highest level If yes, the next hop of the control forwarding unit is the response module, otherwise, the next hop of the control forwarding unit is the storage module; The forwarding unit forwards pending government affairs information To the next hop.

[0019] Further, the storage module includes a temporary storage unit, a main storage unit and a transmission unit; The temporary storage unit temporarily stores pending government affairs information based on the anti-delay strategy and adjusts the temporary storage order; The main memory unit stores necessary data for a long time, including corpus, key dictionary and department level table, among which the corpus is constructed by historical government information; The transmission unit receives pending government affairs information, supervision labels and archiving instructions, transmits the pending government affairs information to the response module, and transfers the historical government affairs information to the main storage unit based on the archiving instructions.

[0020] Furthermore, the temporary storage unit temporarily stores pending government affairs information based on the delay prevention strategy and adjusts the temporary storage order, including the following steps: Get pending government affairs information Reporting time in and level labels , set the maximum storage duration , real-time recording of storage duration , determine the temporary storage order ; How to obtain pending government affairs information The corresponding urging tag updates the number of urging times and determines whether it is greater than 3 times. If so, it will be ignored. Otherwise, the temporary storage order will be updated; If storage time Maximum storage time The ratio is greater than or equal to the alarm ratio , the pending government affairs information is directly transmitted through the transmission unit to the response module.

[0021] Furthermore, the transmission unit transfers the historical government information to the main storage unit based on the archiving instruction, including the following steps: Get archive instructions ,in, for completion time; Generate historical government information and transferred to the main memory unit for long-term storage; Clear pending government affairs information in the temporary storage unit .

[0022] Further, the response module responds to the processing notification initiated by the processor based on the real-time response strategy and transmits the pending government affairs information to the processor, including the following steps; Get the processing notification and determine the processing notification type based on the processing tag; If the processing tag indicates that it is a regular call request, the pending government affairs information with the first temporary storage order is selected and transmitted to the processing party; If the processing tag indicates a specific call request, obtain the department number in the specific call request , select the pending government affairs information with the same department number and the earliest temporary storage order and transmit it to the processing party; If the processing tag indicates processing feedback, get the department number in the processing feedback , Reporting time , Level Label and completion time , generate archive instructions and sent to the storage module; Obtain pending government affairs information and forward it to the processing party.

[0023] Compared with the prior art, the present invention has the following significant advantages: 1. Set up a pre-processing unit to remove duplicate government information through text similarity detection algorithm, automatically correct spelling errors in government information based on text error correction network, unify the data format, data structure and naming of government information, and independently assign level labels to improve the efficiency of government information processing; 2. Set up a temporary storage unit in the storage module to temporarily store pending government information. Through the synergy of the anti-delay strategy and the real-time response strategy of the response module, ensure that government information can be processed in a timely manner, speed up the transfer process of government information, and further improve the efficiency of government information processing; 3. Set up inspection steps in the data receiving module to ensure the reliability and authenticity of the source of government information and ensure the security of government information. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a schematic diagram of a temporary storage response system for government information based on big data; Figure 2This is the flow chart of the text similarity detection algorithm; Figure 3 This is a diagram of the text error correction network model; Figure 4 Flowchart of delay prevention strategy. DETAILED DESCRIPTION

[0025] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0026] The embodiment provided by the present invention is a government information temporary storage response system based on big data, comprising a data receiving module, a preprocessing module, a relay module, a storage module and a response module; The data receiving module receives and verifies the government information, and forwards the verified government information to the pre-processing module; The pre-processing module cleans government information, generates pending government information and forwards it to the relay module. Cleaning government information includes deduplication, optimization and rating of government information. Relay module based on the level label of pending government affairs information Select the sending storage module or processor; The storage module temporarily stores pending government affairs information based on the anti-delay strategy and stores necessary data for a long time, including corpus, confusion set, pronunciation-similar character relationship graph, shape-similar character relationship graph, key dictionary and department level table; The response module responds to the processing notification initiated by the processor based on the real-time response strategy and transmits the pending government affairs information to the processor; Further, from the Government information of government departments Include department number , Reporting time , Government documents ,sign and Certificate ; Department Number Used to identify the department where the government information comes from; Reporting time For the Government departments send government information time; Governmental documents For the Government departments report time The text file sent includes the name of the government and government affairs ; sign By Government departments report time Using department private key and a public hash function Generate, that is ; Certificate In the Issued by the system when registering a government department, it is based on the system private key , Public department key and a public hash function Generate, that is .

[0027] Further, the data receiving module includes a verification unit and a transceiver unit; The verification unit obtains public information, and verifies the validity and authenticity of the government information based on the first equation and the second equation respectively, and the government information that satisfies the equation is deemed to have passed the verification; The transceiver unit receives government affairs information and forwards the government affairs information that has passed the inspection to the pre-processing module.

[0028] Furthermore, the verification unit verifies the government information including the following steps: Get from Government information of government departments ; Obtain public information, including Public key of each government department , System public key , bilinear mapping , hash function and generating factors , where the generation factor Used to generate a public key that matches the private key, using the system public key For example, ; Verify government information through the first equation The validity of the first equation is as follows , Due to the certificate , system public key , when the first equation holds, the certificate Issued by the system, government information The source is valid; Verify government information through the second equation The authenticity of the second equation is as follows ; Due to the signature , department public key , so when the second equation holds, the signature Untampered, government documents The content is true; When both the first and second equations hold, government information The inspection is passed, and the rest are deemed as failed.

[0029] Further, the pre-processing module includes a de-duplication unit, an optimization unit, a rating unit, and a communication unit; The deduplication unit uses a text similarity detection algorithm to identify and remove duplicate government information; The optimization unit corrects spelling errors in government texts based on the text error correction network and standardizes government texts; The rating unit assesses the urgency of government information based on the keyword detection algorithm and assigns a level label to the government information. Generate pending government affairs information; The communication unit receives the government affairs information verified by the data receiving module, sends the pending government affairs information to the relay module, and simultaneously sends the supervision tag to the storage module.

[0030] Furthermore, the deduplication unit uses a text similarity detection algorithm to identify and remove duplicate government information, including the following steps: Get the Government information of government departments ; Query whether there is pending government affairs information of the i-th government affairs department in the storage module; If there is no pending government affairs information of the i-th government department, then the government affairs information No repetition; If there is pending government affairs information of the i-th government affairs department, all pending government affairs information of the i-th government affairs department is extracted from the storage module, and all pending government affairs information of the i-th government affairs department is sorted in reverse chronological order and placed into the empty set in sequence to construct a pending government affairs information set; Get the reporting time of all pending government affairs information from the pending government affairs information set, and convert the reporting time The most recent reporting time is defined as the most recent reporting time , get the latest pending government affairs information Perform duplicate determination. It is worth noting that the most recent reporting time It is not a fixed value, but always represents the time from the time when pending government affairs information is concentrated to the time when it is reported. The most recent reporting time. Assuming that the most recent pending government affairs information corresponding to the most recent reporting time in the current pending government affairs information set is removed, the most recent reporting time will be updated based on the pending government affairs information set after removal, and updated to the reporting time in the pending government affairs information set before removal. The second most recent reporting time; Statistics government documents and recent government affairs to be done Length of text and ,The text length refers to the total number of characters of the government affairs name and government affairs content; Determine the text difference Is it within the difference threshold? In which, The text difference ratio is preset; When the text is poor Greater than difference threshold , determine government information Recent government affairs information No duplication, remove the most recent pending government affairs information from the pending government affairs information set ; When the text is poor Less than or equal to the difference threshold , determine government information and recent pending government affairs information Suspected duplication; The bidirectional LSTM+CRF combined model is used as the word segmenter to analyze government texts. and recent government affairs to be done Perform word segmentation, obtain word segmentation vectors and recent to-do word segmentation vectors respectively, extract the words in the word segmentation vectors and recent to-do word segmentation vectors, and if the same word is encountered, only keep one, and obtain the merged word segmentation set; For the first word segmentation, and calculate the Governmental texts and recent government affairs to be done Frequency of words and recent to-do word frequency The specific calculation formula is as follows , in, Refers to the reporting time and the most recent reporting time ; Calculate the The inverse text frequency of the word The specific calculation formula is as follows , Among them, the corpus is constructed by the historical government information stored in the storage module for a long time; Calculate the Governmental texts and recent government affairs to be done Weight and the weight of the most recent tasks , the specific calculation formula is as follows: , in, Refers to the reporting time and the most recent reporting time ; Hash functions exposed by the system Process the merged word set word segmentation, generate a binary code with a fixed number of bits, and assign a weight to the position of 1 in the binary code and the weight of the most recent tasks , the position of 0 is given weight and the weight of the most recent tasks The opposite number of , generating weight encoding and the weight code of the recent to-do list , assuming that The binary encoding of a word is , and the weight and the weight of the most recent tasks are 1 and 2 respectively, then the weight encoding , recent to-do weight coding ; Superimpose and merge the weight encodings of all the words in the word set and the weight code of the recent to-do list Generate superposition code Overlay coding with recent to-dos The specific formula is as follows , in, Refers to the reporting time and the most recent reporting time , is the total number of segmentations in the merged segmentation set; Determine the superposition coding Overlay coding with recent to-dos Whether the Hamming distance of is greater than the distance threshold; If the superposition coding Overlay coding with recent to-dos If the Hamming distance is less than or equal to the distance threshold, then the government information is determined to be Recent government affairs information Duplicate and remove government information And generate supervision labels to stop repeated judgments; If the superposition coding Overlay coding with recent to-dos If the Hamming distance of is greater than the distance threshold, then the government information is determined to be Recent government affairs information No duplication, remove the most recent pending government affairs information from the pending government affairs information set ; Execute the empty set judgment to determine whether the pending government affairs information set is an empty set. If it is empty, determine whether the pending government affairs information set is an empty set. No duplication, if not empty, further update the latest reporting time The time from the day when the pending government affairs information after elimination is reported The closest reporting time, to obtain the latest pending government affairs information , continue to perform duplicate judgment and empty set judgment.

[0031] Furthermore, the optimization unit corrects spelling errors of government texts based on the text error correction network, including the following steps: Constructing the Confusion Set , similar pronunciation character relationship diagram Relationship diagram of similar characters , and define the approximate character relationship graph ,in, ; Re-representing the confusion set based on graph convolution operation and attention map combination operation The characters in the aggregate representation set are generated ; Extracting government texts through BERT , get government text The corresponding representation set ,in, is the length of the text; Will represent the set Input detector, calculate government text Middle Characters The probability of error , the detector consists of a linear layer and a sigmoid activation function. The specific formula is as follows , in, To be learnable dimensional weight vector, is the bias value, For the Characters The corresponding characters represent, ; Get the aggregate representation set Character representation The most similar Aggregate representations to build a candidate representation set , the candidate representation set includes Aggregate representation and character representation , the dimension is ; In the confusion set Based on Aggregate representation indexes to the corresponding characters and constructs a candidate set , candidate set Including index Characters and government text The characters, among which there is only 1 correct character, defined as a positive example ,the remaining characters are all wrong characters, defined as negative examples ; Use the corrector to calculate the candidate set The candidate characters are positive examples The predicted probability , the corrector is implemented by a linear layer, a batch normalization layer, a feedforward layer, and a relu activation function. The specific formula is as follows , in, The candidate representation set Middle The candidate representation corresponding to the candidate character, To be learnable dimensional weight matrix, for dimensional bias vector, ; The contrast loss is introduced to maximize the prediction probability of positive examples and minimize the prediction probability of negative examples through contrast learning. Therefore, the final loss ,in, To detect the loss, To correct the loss, is the contrast loss, the specific formula is as follows , , .

[0032] Furthermore, the confusion set is a predefined character set, including characters, , considered The above character similarity includes similar shape, same pronunciation and same tone, same pronunciation and different tone, similar pronunciation and same tone, and similar pronunciation and different tone; Characters with the same or similar pronunciation are used as graph nodes, and edges between any two graph nodes are defined and generated to logically connect the two graph nodes, thereby constructing a graph of similar-pronunciation characters. , taking characters with similar shapes as graph nodes, defining and generating edges between any two graph nodes to logically connect the two graph nodes, thus constructing a relationship graph of similar-shaped characters , the dimensions are , define the approximate character relationship graph ,exist and The relationship diagram of characters with similar pronunciation Relationship diagram of similar characters ; side The definition of confusion set is Middle characters and The similarity of characters, using BERT to extract the confusion set , get the representation set , calculate the characters and The character corresponding to the character represents and The distance includes Euclidean distance, cosine distance and Hamming distance. The normalized distance can be obtained characters and The similarity of characters.

[0033] Furthermore, the aggregated representation set is generated based on the graph convolution operation and the attention graph combination operation The following steps are involved: Extracting confusion sets through BERT , get the representation set , the dimension is ,in, To confuse the concentration characters, For the Characters The corresponding characters represent, ; Processing approximate character relationship graphs through lightweight convolutional layers Get the convolution map , convolutional graph Incorporates character relationship graph The information of adjacent characters in is The specific formula is as follows , in, , Time Convolutional Graph Representing the pronunciation convolution map , Time Convolutional Graph Representing shape convolution graph , is a learnable weight matrix with dimension , For the approximate character relationship diagram The standardized representation of ; The attention mechanism is used to combine the similarity of character pronunciation and character shape to generate an attention combination graph , the dimension is , where the attention combination graph is generated No. OK The specific formula is as follows , in, , Convolutional graph No. row, corresponding to Characters , For the Characters For the approximate character relationship graph The attention weight, attention weight The specific formula is as follows , in, , is a learnable vector with dimension , is a hyperparameter used to control the smoothness of the attention weight; Superimposed attention combination diagram and the representation set To obtain the aggregate representation set , the dimension is .

[0034] Furthermore, optimizing the unit standardized government documents includes the following steps: Unified data format, dates are represented in ISO 8601 format, numbers are represented in floating point numbers, and strings are represented in UTF-8 encoding; Unified data structure: lists, sequences, and vectors are all represented by arrays, and tree structures are all represented by binary trees; Set data constraints: date cannot be empty, string cannot contain special symbols, binary tree cannot contain duplicate elements; The government affairs documents are uniformly named as “department number-reporting time-government affairs name”, starting with Government information of government departments Take the following example to get the government text: Name of government affairs , renamed .

[0035] Furthermore, the rating unit assesses the urgency of government information based on the keyword detection algorithm and assigns a level label to the government information. Generating pending government affairs information includes the following steps: Get the Government information of government departments Government documents in ; Get the department number , query the department level table and use the department level corresponding to the department number as the initial level label ,and , is the highest level; The bidirectional LSTM+CRF combined model is used as the word segmenter to analyze government texts. Perform word segmentation and obtain word segmentation vector; Count the total number of word segmentations in the word segmentation vector and the total number of key words ,Key word refers to the word vector that exists in the key dictionary; Based on the total number of key words Calculate the increase in word segmentation vector The specific formula is as follows , in, Indicates rounding operation. A pre-set grading scale, with the highest level Decide, , ; Compute level tags And provide government information , generate pending government affairs information .

[0036] Further, the relay module includes a determination control unit and a forwarding unit; Determine the control unit to obtain pending government affairs information Level labels in , determine the level label Is the value of the highest level If yes, the next hop of the control forwarding unit is the processing party, if no, the next hop of the control forwarding unit is the storage module; The forwarding unit forwards pending government affairs information To the next hop.

[0037] Further, the storage module includes a temporary storage unit, a main storage unit and a transmission unit; The temporary storage unit stores pending government affairs information in sequence based on the anti-delay strategy and adjusts the temporary storage order in real time; The main memory unit stores necessary data for a long time, including corpus, key dictionary and department level table, among which the corpus is constructed by historical government information; The transmission unit receives pending government affairs information, supervision labels and archiving instructions, transmits the pending government affairs information to the response module based on the anti-delay strategy or response call request, and transfers the historical government affairs information to the main storage unit based on the archiving instruction.

[0038] Furthermore, the temporary storage unit stores pending government affairs information in sequence based on the anti-delay strategy and adjusts the temporary storage order in real time. The following steps are involved: Get pending government affairs information Reporting time in and level labels , according to the level label Set the maximum storage duration , and record the storage time in real time ; Determine the order of temporary storage , The smaller the value, the earlier the temporary storage order, and the larger the level label value, the pending government affairs information The smaller the value, the earlier the pending government affairs information of the same level label is reported. The smaller the value; How to obtain pending government affairs information The corresponding supervision label, let the number of supervisions increase by 1, and judge whether the number of supervisions is greater than 3. If it is greater than 3, it will be ignored. If it is less than or equal to 3, further based on the level label and reporting time Determine the order of temporary storage , if the temporary storage order , then no adjustment will be made. If the temporary storage order , then the update temporary storage order is ; If storage time Maximum storage time The ratio is greater than or equal to the alarm ratio , it is considered as pending government affairs information Timeout is about to occur, and pending government affairs information is directly transmitted through the transmission unit to the response module.

[0039] Furthermore, the transmission unit transfers the historical government information to the main storage unit based on the archiving instruction, including the following steps: Get archive instructions ,in, for completion time; Match and confirm pending government affairs information in the temporary storage unit , generate historical government information ; Clear pending government affairs information in the temporary storage unit , and transmit historical government information To main memory unit for long-term storage.

[0040] Further, the response module responds to the processing notification initiated by the processor based on the real-time response strategy and transmits the pending government affairs information, including the following steps; Get the processing notification and determine the processing notification type based on the processing tag; If the processing tag indicates that the processing notification is a regular call request, the pending government affairs information with the first temporary storage order in the call storage module is transmitted to the processing party; If the processing tag indicates that the processing notification is a specific call request, obtain the department number in the specific call request , call the pending government affairs information with the same department number and the earliest temporary storage order in the storage module and transmit it to the processing party; If the processing tag indicates that the processing notification is a processing feedback, then obtain the department number in the processing feedback , Reporting time , Level Label and completion time , generate archive instructions and sent to the storage module; Obtain pending government affairs information and forward it directly to the processing party.

[0041] The present invention discloses a temporary storage response system for government information based on big data, comprising a data receiving module, a preprocessing module, a relay module, a storage module and a response module; the data receiving module receives and verifies the validity and authenticity of government information, and only forwards the government information that has passed the verification to ensure system security; the preprocessing module removes duplicate government information through a text similarity detection algorithm, automatically corrects spelling errors in government information based on a text error correction network, unifies the format of government information and autonomously assigns level labels, thereby improving the efficiency of government information processing; the relay module transfers government information based on the level label, and the storage module and the response module accelerate the processing flow through the synergistic effect of anti-delay strategies and real-time response strategies, thereby further improving the efficiency of government information processing.

[0042] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A government information temporary storage response system based on big data, characterized in that: It includes a data receiving module, a preprocessing module, a relay module, a storage module and a response module; The data receiving module receives and verifies the government information, and forwards the verified government information to the pre-processing module; The pre-processing module cleans the government affairs information, generates pending government affairs information and forwards it to the relay module; The relay module selects a sending storage module or a processing party based on the level label of the pending government affairs information; The storage module temporarily stores pending government affairs information based on anti-delay strategies and stores necessary information for a long time; The response module responds to the processing notification based on the real-time response strategy and transmits the pending government affairs information to the processing party.

2. A government information temporary storage response system based on big data as claimed in claim 1, characterized in that: The pre-processing module includes a de-duplication unit, an optimization unit, a rating unit and a communication unit; The deduplication unit uses a text similarity detection algorithm to identify and remove duplicate government information; The optimization unit corrects spelling errors based on a text error correction network and standardizes government texts; The rating unit assigns a level label to the government affairs information based on a keyword detection algorithm and generates pending government affairs information; The communication unit receives the government affairs information that has passed the inspection, sends the pending government affairs information to the relay module, and simultaneously sends the supervision tag to the storage module.

3. A government information temporary storage response system based on big data as claimed in claim 2, characterized in that: The deduplication unit uses a text similarity detection algorithm to identify and remove duplicate government information, including the following steps: Check whether there is pending government affairs information with the same department number to preliminarily determine whether the government affairs information is repeated; If not, it is determined that the government affairs information is not repeated, otherwise all pending government affairs information with the same department number is extracted to construct a pending government affairs information set; Acquire the latest pending government affairs information from the pending government affairs information set, perform duplication determination, and determine whether the government affairs information and the latest pending government affairs information are duplicated based on text differences; If there is no repetition, the most recent pending government affairs information is removed from the pending government affairs information set, and on the contrary, a bidirectional LSTM+CRF combined model is used for word segmentation to obtain a merged word segmentation set; Calculate the word frequency and recent to-do word frequency of a single word in the combined word set, calculate the inverse text frequency of a single word to obtain the weight of the single word and the recent to-do weight, and further generate the weight code of the single word and the recent to-do weight code; Obtain the superposition codes of all word segments and the superposition codes of the most recent to-do items, and determine whether the government information is repeated with the most recent to-do government information based on the Hamming distance. If repeated, remove the government information and generate a urging label. Otherwise, remove the most recent to-do government information from the to-do government information set. Execute an empty set judgment to determine whether the pending government affairs information set is an empty set. If it is an empty set, determine that the government affairs information is not repeated. Otherwise, obtain new and recent pending government affairs information from the pending government affairs information set to perform duplication judgment and empty set judgment.

4. A government information temporary storage response system based on big data as claimed in claim 2, characterized in that: The optimization unit corrects spelling errors of government texts based on the text error correction network, including the following steps: Construct confusion sets, pronunciation-similar character relationship graphs, and shape-similar character relationship graphs; Re-represent the characters in the confusion set based on the graph convolution operation and the attention map combination operation to generate an aggregated representation set; The government text is extracted through BERT to obtain a representation set, and the error probability of a single character in the government text is calculated through the detector; Obtain multiple aggregate representations in the aggregate representation set that are most similar to the character representation, index them to the corresponding characters in the confusion set, and construct a candidate representation set and a candidate set respectively; The corrector is used to calculate the predicted probability of a candidate character in the candidate set being a positive example; Contrastive loss is introduced to maximize the prediction probability of positive examples and minimize the prediction probability of negative examples through contrastive learning. The final loss is defined as the sum of detection loss, correction loss, and contrast loss.

5. A government information temporary storage response system based on big data as claimed in claim 4, characterized in that: The generating of the aggregated representation set based on the graph convolution operation and the attention graph combination operation comprises the following steps: Extract the confusion set through BERT to obtain the representation set; Processing the pronunciation-approximate character relationship graph and the shape-approximate character relationship graph through a lightweight convolution layer to obtain a pronunciation convolution graph and a shape convolution graph; The attention mechanism is used to combine the similarity of character pronunciation and character shape to generate an attention combination graph; The attention combination map and the representation set are superimposed to obtain the aggregated representation set.

6. A government information temporary storage response system based on big data as claimed in claim 2, characterized in that: The rating unit assigns a level label based on a keyword detection algorithm, and generates pending government affairs information, including the following steps: Get the department number, query the department level table, and get the initial level label; Obtain government text from government information, use a bidirectional LSTM+CRF combination model to segment the government text, and obtain segmentation vectors; Count the total number of segmented words and key segmented words in the segmented word vector, and calculate the increment of the segmented word vector; Calculate level labels and assign government information, and generate pending government information.

7. A government information temporary storage response system based on big data as claimed in claim 1, characterized in that: The storage module includes a temporary storage unit, a main storage unit and a transmission unit; The temporary storage unit temporarily stores pending government affairs information based on the anti-delay strategy and adjusts the temporary storage order; The main storage unit stores necessary data for a long period of time, and the necessary data includes a corpus, which is constructed from historical government information; The transmission unit receives pending government affairs information, supervision labels and archiving instructions, transmits the pending government affairs information to the response module, generates historical government affairs information based on the archiving instructions and transfers it to the main storage unit, and clears the pending government affairs information in the temporary storage unit.

8. A government information temporary storage response system based on big data as claimed in claim 7, characterized in that: The temporary storage unit temporarily stores pending government affairs information based on the delay prevention strategy and adjusts the temporary storage order, including the following steps: Obtain the reporting time and level label in the pending government affairs information, set the maximum storage time, record the storage time in real time, and determine the temporary storage order; If the urging tag corresponding to the pending government affairs information is obtained, the number of urgings is updated, and based on the number of urgings, it is determined whether to update the temporary storage order; If the ratio of the storage time to the maximum storage time is greater than or equal to the alarm ratio, the pending government affairs information is transmitted to the response module through the transmission unit.

9. A government information temporary storage response system based on big data as claimed in claim 1, characterized in that: The response module responds to the notification based on the real-time response strategy and transmits the pending government affairs information, including the following steps: Get the processing notification and determine the processing notification type based on the processing tag; If the processing tag indicates a regular call request, the pending government affairs information with the first temporary storage order is called and transmitted to the processing party; If the processing tag indicates a specific call request, the department number in the specific call request is called, and the pending government affairs information with the same department number and the first temporary storage order is selected and transmitted to the processing party; If the processing label indicates processing feedback, obtain the department number, reporting time, level label and completion time in the processing feedback, generate an archive instruction and send it to the storage module; Obtain pending government affairs information and transmit it directly to the processor.

10. A government information temporary storage response system based on big data as claimed in claim 1, characterized in that: The data receiving module includes a verification unit and a transceiver unit; The verification unit obtains government information and public information, verifies the validity of the government information through the first equation, and verifies the authenticity of the government information through the second equation. The verification passes when both the first equation and the second equation are satisfied; The transceiver unit receives government affairs information and forwards the government affairs information that has passed the inspection to the pre-processing module.

Images