Tobacco monopoly lean market supervision preventive case intelligent generation system
Automatically extract and generate tobacco monopoly market supervision cases through natural language processing and deep learning technology, solving the problems of human errors and case template updates in the existing technology, achieving efficient case sorting and early warning push, and improving the accuracy of supervision and forward-looking.
Patent Information
- Application Number
- CN202411094481.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-10
- Publication Date
- 2024-12-03
AI Technical Summary
The existing tobacco monopoly market supervision technology is prone to human errors during the process of certificate collection and document writing, which leads to the inability to effectively supervise the latest market behavior, and existing case templates cannot be updated in a timely manner.
Use natural language processing, video and audio recognition technology to automatically extract and format case elements, combine deep learning algorithms to generate new case scenarios, use data analysis and prediction models to push early warning information, and regularly analyze case template libraries to identify long-term problems and generate systems Review the recommendations.
Automatic case sorting and key factor extraction have been achieved, structured case templates have been formed, and the forward-looking and accurate supervision has been improved, and law enforcement personnel have a basis to take the initiative to prevent, and a regular system review mechanism has been formed.
Smart Images

Figure CN119067707A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of tobacco management, and in particular relates to an intelligent generation system for preventive cases of lean market supervision of tobacco monopoly. Background Art
[0002] Under the current lean market supervision mechanism of tobacco monopoly, supervisors collect evidence and write judicial documents in accordance with the law. After the case is over, the collected supervision case data will be used as the analysis object, that is, the existing case data will be counted and analyzed to find the key supervision objects and conduct precise supervision on retailers.
[0003] In the Chinese patent with publication number CN114493442A, a comprehensive query method based on tobacco-related intelligence data is mentioned. It includes the following steps: data management step: the server obtains tobacco industry data and stores it in the database; relationship analysis step: the tobacco industry data in the database is analyzed for relevance, and the tobacco industry data with relevance is marked as related data; data query step: the server receives the query request of the front-end device, and according to the query request of the front-end device, calls the requested data in the database, and the related data of the requested data, and feeds it back to the front-end device. Compared with the prior art, it can conduct a comprehensive analysis and exploration of tobacco data. Although the above scheme has many beneficial effects, it is inevitable that human errors will occur in the process of evidence collection, document writing, and information extraction in this scheme. The case object analyzed, as a past case, often leads to existing case templates, similar or related cases and punishment measures, and case-related information such as citing legal and regulatory content, which cannot effectively regulate the latest market behavior.
[0004] In response to this, the inventor proposed an intelligent generation system for preventive cases in lean market supervision of tobacco monopoly to solve the above problems. Summary of the invention
[0005] The purpose of the present invention is to provide an intelligent generation system for preventive cases of lean market supervision of tobacco monopoly to solve the problems raised in the above-mentioned background technology.
[0006] To achieve the above object, the present invention provides the following technical solutions:
[0007] The intelligent generation system of preventive cases for lean market supervision of tobacco monopoly includes:
[0008] The data processing module uses natural language processing (NLP) technology and video and audio recognition technology to automatically extract and format case-related elements from the transcripts collected at the scene;
[0009] The text verification module uses text verification based on case-related elements collected on-site, uses text processing and proofreading software, and combines legal-specific corpora to automatically verify, amend, and correct grammatical, spelling, and factual errors in judicial documents;
[0010] Establish a case template module, apply information extraction methods, extract key elements from a large amount of case data, and generate structured case templates;
[0011] The case coverage analysis and early warning push module uses data analysis and prediction models to calculate the coverage of case elements, and based on this, generates early warning information and pushes it to relevant supervisors;
[0012] The new case generation module, combined with the CNN deep learning algorithm, analyzes existing data and generates new case scenarios to simulate the occurrence of illegal situations;
[0013] The case template library analysis and system review module uses data mining and statistical analysis methods to regularly analyze the case template library, identify long-term problems and generate corresponding system review recommendations.
[0014] Preferably, the relevant elements include the time, location, and amount of the illegal sales.
[0015] Preferably, the case template includes time, place, retailer, goods, violation content, and penalty measures.
[0016] Preferably, the natural language processing is used to extract case elements from the text, using word frequency-inverse document frequency to measure the importance of words in the document, expressed as:
[0017] TF-IDF(t,d,D)=TF(t,d)×IDF(t,D)
[0018] Among them, TF(t,d) is the frequency of word t in document d, and IDF(t,D) is the inverse document frequency of word t in corpus D.
[0019] Preferably, the text verification uses the edit distance to calculate the minimum number of edit operations between two strings, expressed as:
[0020]
[0021] Where d(i,j): the edit distance between the first i characters and the first j characters of strings s and t;
[0022] cost(si,tj): the replacement cost between characters si and tj;
[0023] If si=tj, then cost(si,tj)=0, otherwise 1;
[0024] Grammatical correction, based on the statistical language model n-gram model, is expressed as:
[0025]
[0026] Where P(w i ∣w i-n+1 ,…,w i-1 ): Given the first n-1 words, word w i Probability of occurrence;
[0027] Used to automatically check and correct grammatical and spelling errors in text to improve document quality.
[0028] Preferably, the information extraction method is based on TF-IDF, extracts key information from a large amount of case data, and generates a standardized case template.
[0029] Preferably, the calculation of the coverage is expressed as:
[0030]
[0031] Coverage: coverage rate, R: number of relevant cases; T: total number of cases;
[0032] The early warning model is based on time series analysis and ARI MA forecasting model, which is expressed as:
[0033] yt=α+βt+e
[0034] yt: observed value at time t, α: constant term, β: time trend coefficient, e: error term;
[0035] It is used to calculate the coverage rate of case elements and generate early warning information based on this, which will be pushed to relevant supervisory personnel.
[0036] Preferably, the CNN deep learning algorithm is used for processing image and video data, expressed as:
[0037] f(x) = ReLU(W*x+b)
[0038] Where W: convolution kernel, x: input feature map, b: bias, ReLU: activation function;
[0039]
[0040] Where D: discriminator, G: generator, x: real data, z: noise vector;
[0041] Deep learning technology is used to generate new case scenarios, simulate violations, and improve the foresight of supervision.
[0042] Preferably, the data mining adopts April association rule mining, and its support is expressed as:
[0043]
[0044] Where Count(X∪Y): the number of transactions that contain X and Y;
[0045] N: total number of transactions;
[0046] The confidence level is expressed as:
[0047]
[0048] Count(X): the number of transactions containing X;
[0049] Used to regularly analyze case data, identify chronic issues, and generate appropriate system review recommendations.
[0050] Compared with the prior art, the present invention has the following beneficial effects:
[0051] The present invention uses large model technology to complete a series of judicial procedures such as automatic formatting and classification starting from the evidence collection stage. After the case is sorted out, its key factors are automatically extracted to form a case template. Existing case data is regularly collected and sorted out. Using NLP technology, keywords of existing cases are extracted to form new cases that may occur. As active prevention objects, cases with a large coverage rate of case elements are actively pushed, so that law enforcement personnel have precedents to rely on, and a regular system review mechanism can be formed. BRIEF DESCRIPTION OF THE DRAWINGS
[0052] Figure 1 This is a block diagram of the intelligent generation system for preventive cases of lean market supervision of tobacco monopoly of the present invention. DETAILED DESCRIPTION
[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0054] Embodiment 1:
[0055] See also Figure 1 As shown in the figure, the intelligent generation system of preventive cases for lean market supervision of tobacco monopoly includes:
[0056] The data processing module uses natural language processing (NLP) technology and video and audio recognition technology to automatically extract and format case-related elements from the transcripts collected on site, including the time, location, and amount of illegal sales;
[0057] The text verification module uses text verification based on case-related elements collected on-site, uses text processing and proofreading software, and combines legal-specific corpora to automatically verify, amend, and correct grammatical, spelling, and factual errors in judicial documents;
[0058] Establish a case template module, apply information extraction methods, extract key elements from a large amount of case data, and generate a structured case template, including time, location, retailer, goods, violation content, and penalty measures;
[0059] The case coverage analysis and early warning push module uses data analysis and prediction models to calculate the coverage of case elements, and based on this, generates early warning information and pushes it to relevant supervisors;
[0060] The new case generation module, combined with the CNN deep learning algorithm, analyzes existing data and generates new case scenarios to simulate the occurrence of illegal situations;
[0061] The case template library analysis and system review module uses data mining and statistical analysis methods to regularly analyze the case template library, identify long-term problems and generate corresponding system review recommendations.
[0062] Specifically, the natural language processing is used to extract case elements from the text, and the importance of words in the document is measured using word frequency-inverse document frequency, which is expressed as:
[0063] TF-IDF(t,d,D)=TF(t,d)×IDF(t,D)
[0064] Among them, TF(t,d) is the frequency of word t in document d, and IDF(t,D) is the inverse document frequency of word t in corpus D.
[0065] Specifically, the text verification uses the edit distance to calculate the minimum number of edit operations between two strings, which is expressed as:
[0066]
[0067] Where d(i,j): the edit distance between the first i characters and the first j characters of strings s and t;
[0068] cost(si,tj): the replacement cost between characters si and tj;
[0069] If si=tj, then cost(si,tj)=0, otherwise 1;
[0070] Grammatical correction, based on the statistical language model n-gram model, is expressed as:
[0071]
[0072] Where P(w i ∣w i-n+1 ,…,w i-1 ): Given the first n-1 words, word w i Probability of occurrence;
[0073] Used to automatically check and correct grammatical and spelling errors in text to improve document quality.
[0074] Specifically, the information extraction method is based on TF-IDF, extracts key information from a large amount of case data, and generates a standardized case template.
[0075] Specifically, the calculation of the coverage rate is expressed as:
[0076]
[0077] Coverage: coverage rate, R: number of relevant cases; T: total number of cases;
[0078] The early warning model is based on time series analysis and ARI MA forecasting model, which is expressed as:
[0079] yt=α+βt+e
[0080] yt: observed value at time t, α: constant term, β: time trend coefficient, e: error term;
[0081] It is used to calculate the coverage rate of case elements and generate early warning information based on this, which will be pushed to relevant supervisory personnel.
[0082] Specifically, the CNN deep learning algorithm is used for processing image and video data, which is expressed as:
[0083] f(x) = ReLU(W*x+b)
[0084] Where W: convolution kernel, x: input feature map, b: bias, ReLU: activation function;
[0085]
[0086] Where D: discriminator, G: generator, x: real data, z: noise vector;
[0087] Deep learning technology is used to generate new case scenarios, simulate violations, and improve the foresight of supervision.
[0088] Specifically, the data mining adopts April association rule mining, and its support is expressed as:
[0089]
[0090] Where Count(X∪Y): the number of transactions that contain X and Y;
[0091] N: total number of transactions;
[0092] The confidence level is expressed as:
[0093]
[0094] Count(X): the number of transactions containing X;
[0095] Used to regularly analyze case data, identify chronic issues, and generate appropriate system review recommendations.
[0096] For a video frame containing a scene of a tobacco shop, tobacco products are detected and recognized as follows:
[0097] Input image: Read the video frame and extract feature maps through the convolutional layer.
[0098] The feature map size is H×W×C, where H and W are the height and width respectively, and C is the number of channels;
[0099] Generate candidate regions: Use RPN to generate a series of candidate regions, each represented by a bounding box;
[0100] Proposal Region Classification and Regression: Classify (tobacco / non-tobacco) and regress bounding boxes for each candidate region.
[0101] Object detection results: Outputs the location and category of the detected tobacco products.
[0102] Consider a video clip that records transaction activities in a tobacco shop and wants to identify whether there are any illegal transactions.
[0103] Input video clip: Read the video clip and extract features through 3D convolutional layers.
[0104] The size of the video clip is T×H×W×C, specifically 16×112×112×3.
[0105] 3D convolution operation: Use 3D convolution kernels to perform convolution in time and space dimensions to extract spatiotemporal features.
[0106] Classification: Pass the extracted features through the fully connected layer and output the probability of illegal transaction / normal transaction.
[0107] From the above, it can be seen that the present invention uses large model technology to complete a series of judicial procedures such as automatic formatting and classification starting from the evidence collection stage. After the case is sorted out, its key factors are automatically extracted to form a case template. Existing case data is regularly collected and sorted out. Using NLP technology, keywords of existing cases are extracted to form new cases that may occur. As active prevention objects, cases with a large coverage rate of case elements are actively pushed, so that law enforcement personnel have precedents to rely on, and a regular system review mechanism can be formed.
[0108] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.
[0109] In the drawings of the embodiments disclosed in the present invention, only the structures related to the embodiments disclosed in the present invention are involved, and other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of the present invention can be combined with each other.
[0110] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. The intelligent generation system of preventive cases for lean market supervision of tobacco monopoly is characterized by: include: The data processing module uses natural language processing (NLP) technology and video and audio recognition technology to automatically extract and format case-related elements from the transcripts collected at the scene; The text verification module uses text verification based on case-related elements collected on-site, uses text processing and proofreading software, and combines legal-specific corpora to automatically verify, amend, and correct grammatical, spelling, and factual errors in judicial documents; Establish a case template module, apply information extraction methods, extract key elements from a large amount of case data, and generate structured case templates; The case coverage analysis and early warning push module uses data analysis and prediction models to calculate the coverage of case elements, and based on this, generates early warning information and pushes it to relevant supervisors; The new case generation module, combined with the CNN deep learning algorithm, analyzes existing data and generates new case scenarios to simulate the occurrence of illegal situations; The case template library analysis and system review module uses data mining and statistical analysis methods to regularly analyze the case template library, identify long-term problems and generate corresponding system review recommendations.
2. The intelligent generation system for preventive cases of lean market supervision of tobacco monopoly according to claim 1 is characterized by: The relevant elements include the time, location and amount of the illegal sales.
3. The intelligent generation system for preventive cases of lean market supervision of tobacco monopoly according to claim 1 is characterized by: The case template includes time, place, retailer, goods, violation content, and penalty measures.
4. The intelligent generation system for preventive cases of lean market supervision of tobacco monopoly according to claim 1 is characterized by: The natural language processing is used to extract case elements from the text, and the importance of words in the document is measured using word frequency-inverse document frequency, which is expressed as: TF-IDF(t,d,D)=TF(t,d)×IDF(t,D) Among them, TF(t,d) is the frequency of word t in document d, and IDF(t,D) is the inverse document frequency of word t in corpus D.
5. The intelligent generation system for preventive cases of lean market supervision of tobacco monopoly according to claim 1 is characterized by: The text check uses the edit distance to calculate the minimum number of edit operations between two strings, expressed as: Where d(i,j): the edit distance between the first i characters and the first j characters of strings s and t; cost(si,tj): the replacement cost between characters si and tj; If si=tj, then cost(si,tj)=0, otherwise 1; Grammatical correction, based on the statistical language model n-gram model, is expressed as: Where P(w i ∣w i-n+1 ,…,w i-1 ): Given the first n-1 words, word w i Probability of occurrence; Used to automatically check and correct grammatical and spelling errors in text to improve document quality.
6. The intelligent generation system for preventive cases of lean market supervision of tobacco monopoly according to claim 1 is characterized by: The information extraction method is based on TF-IDF, extracts key information from a large amount of case data, and generates a standardized case template.
7. The intelligent generation system for preventive cases of lean market supervision of tobacco monopoly according to claim 1 is characterized by: The calculation of the coverage is expressed as: Coverage: coverage rate, R: number of relevant cases; T: total number of cases; The early warning model is based on time series analysis and ARIMA forecasting model, which can be expressed as: yt=α+βt+e yt: observed value at time t, α: constant term, β: time trend coefficient, e: error term; It is used to calculate the coverage rate of case elements and generate early warning information based on this, which will be pushed to relevant supervisory personnel.
8. The intelligent generation system for preventive cases of lean market supervision of tobacco monopoly according to claim 1 is characterized by: The CNN deep learning algorithm is used for processing image and video data, which is expressed as: f(x)=ReLU(W*x+b) Where W: convolution kernel, x: input feature map, b: bias, ReLU: activation function; Where D: discriminator, G: generator, x: real data, z: noise vector; Deep learning technology is used to generate new case scenarios, simulate violations, and improve the foresight of supervision.
9. The intelligent generation system for preventive cases of lean market supervision of tobacco monopoly according to claim 1 is characterized by: The data mining adopts Apriori association rule mining, and its support is expressed as: Where Count(X∪Y): the number of transactions that contain X and Y; N: total number of transactions; The confidence level is expressed as: Count(X): the number of transactions containing X; Used to regularly analyze case data, identify chronic issues, and generate appropriate system review recommendations.
Citation Information
Patent Citations
Comprehensive query method based on smoke-related intelligence data
CN114493442A
