Method and device for reviewing import and export goods declaration information
Through the feature extraction and prediction model of deep learning technology, the cargo tax number is automatically queried and compared, which solves the problems of low manual review efficiency and high error rate, and achieves efficient and accurate cargo customs declaration information review.
Patent Information
- Application Number
- CN202311872092.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-29
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2043-12-29
AI Technical Summary
In the prior art, manual review of goods customs declaration information is inefficient and can easily lead to incorrect query results, affecting customs declaration efficiency.
Using deep learning technology, the goods tax number is automatically queried through the feature extraction model and the tax number prediction model, the semantic analysis model and the CNN/LSTM module extract key information, and the model compares the declaration tax number and predicted tax number to determine whether the customs declaration information is pass.
It improves the efficiency of cargo customs declaration information query, reduces errors caused by manual negligence, realizes an efficient and accurate audit process, and reduces costs.
Smart Images

Figure CN117829158B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of trade management, and in particular to a method and device for reviewing customs declaration information of imported and exported goods. Background Art
[0002] In current global trade activities, customs declaration is an important link in the import and export process.
[0003] The existing customs declaration information review mainly relies on manual operation. Specifically, it manually reviews whether the tax number of the goods filled in the customs declaration information form is consistent with the actual tax number of the goods. Every time the tax number of the goods is manually reviewed, it is necessary to query the paper version of the goods tax number table, find the tax number corresponding to the current goods from the goods tax number table according to the correspondence between the goods type and the tax number, and then compare the tax number of the goods found with the goods tax number in the customs declaration information, and determine whether the customs declaration information is passed based on the comparison results. The method of manually querying the goods tax number table to obtain the tax number corresponding to the current goods and manually comparing the tax numbers is not only inefficient, but also prone to errors in the query results due to the negligence of the query personnel, thereby affecting the efficiency of customs declaration. Summary of the invention
[0004] The main purpose of the present invention is to provide a method and device for reviewing import and export goods declaration information, aiming to solve the technical problems in the prior art that manual review of goods declaration information is low in efficiency and easily leads to erroneous query results.
[0005] In one aspect, the present invention provides a method for reviewing import and export goods declaration information, comprising:
[0006] Obtain customs declaration information for goods;
[0007] Inputting the customs declaration information of the goods into the trained feature extraction model to obtain key information included in the customs declaration information of the goods; the key information is used to characterize the identity information of the goods;
[0008] Inputting the key information of the goods into the trained tax number prediction model to obtain the predicted tax number corresponding to the goods;
[0009] Obtain the declared tax number included in the customs declaration information of the goods, compare the declared tax number corresponding to the goods with the predicted tax number, and determine whether the customs declaration information of the goods is passed based on the comparison result.
[0010] Optionally, the feature extraction model is a semantic analysis model;
[0011] The semantic analysis model includes a word segmentation algorithm module, a part-of-speech tagging module, a dependency syntax analysis module and an entity recognition module;
[0012] The step of inputting the customs declaration information of the goods into the trained semantic analysis model to obtain key information included in the customs declaration information of the goods includes:
[0013] Inputting the customs declaration information of the goods into the word segmentation algorithm module, the word segmentation algorithm module is used to perform text decomposition on the customs declaration information of the goods to obtain multiple vocabulary units, and outputting the multiple vocabulary units to the part-of-speech tagging module;
[0014] The part-of-speech tagging module is used to identify the part of speech of each vocabulary unit, tag each vocabulary unit according to the identified part of speech, and output the tagged vocabulary unit to the dependency syntax analysis module;
[0015] The dependency syntax analysis module is used to analyze the dependency relationship between the multiple annotated vocabulary units, annotate the dependency relationship between the multiple vocabulary units according to the analyzed dependency relationship, and output the annotated multiple vocabulary units to the entity recognition module;
[0016] The entity recognition module is used to determine at least one of the multiple vocabulary units as the key information of the goods according to the parts of speech of the multiple vocabulary units and the dependency relationship between the multiple vocabulary units.
[0017] Optionally, the feature extraction model includes a CNN module, an LSTM module and an association rule learning module;
[0018] The step of inputting the customs declaration information of the goods into the trained feature extraction model to obtain key information included in the customs declaration information of the goods includes:
[0019] Inputting the customs declaration information of the goods into the CNN module, the CNN module is used to extract vocabulary units from the customs declaration information of the goods to obtain multiple vocabulary units, and inputting the multiple vocabulary units into the LSTM module;
[0020] The LSTM module is used to obtain the association relationship between the multiple vocabulary units, mark the association relationship between the multiple vocabulary units, and input the marked multiple vocabulary units into the association rule learning module;
[0021] The association rule learning module is used to determine at least one of the multiple vocabulary units as the key information of the goods according to the association relationship between the multiple vocabulary units.
[0022] Optionally, before inputting the key information of the goods into the trained tax number prediction model, the method further includes:
[0023] Obtain key sample information corresponding to the cargo samples, as well as the sample tax number corresponding to each cargo sample;
[0024] Input the sample key information into the initial tax number prediction model to obtain the predicted sample tax number corresponding to each cargo sample;
[0025] According to the sample tax number and the predicted sample tax number, the model parameters of the initial tax number prediction model are updated to obtain a trained tax number prediction model.
[0026] Optionally, after obtaining the trained tax number prediction model, the following is further included:
[0027] Obtain the sample key information corresponding to the cargo test sample and the sample tax number corresponding to each cargo test sample;
[0028] Inputting the sample key information corresponding to the cargo test sample into the trained tax number prediction model to obtain the test sample tax number;
[0029] Determine whether the trained tax number prediction model meets the preset requirements based on the sample tax number corresponding to the cargo test sample and the test sample tax number. If so, terminate the model training and output the trained tax number prediction model. Otherwise, increase the number of iterations and continue to train the trained tax number prediction model.
[0030] Optionally, comparing the declared tax number corresponding to the goods with the predicted tax number, and determining whether the customs declaration information of the goods is passed according to the comparison result, includes:
[0031] If the declared tax number and the predicted tax number corresponding to the goods are consistent, it is determined that the customs declaration information of the goods has passed; otherwise, it is determined that the customs declaration information of the goods is wrong, and an early warning signal is issued, which is used to indicate that the declared tax number and the predicted tax number corresponding to the goods are inconsistent.
[0032] Optionally, after the warning signal is issued, the method further includes:
[0033] Sending the warning signal to a designated client or database; and / or
[0034] The warning signal is pushed to a designated client, and a warning signal pop-up window is generated on the display interface of the client.
[0035] On the other hand, the present invention also provides an import and export goods declaration information review device, comprising:
[0036] An acquisition unit, used to acquire the customs declaration information of goods;
[0037] A feature extraction unit, used for inputting the customs declaration information of the goods into the trained feature extraction model to obtain key information included in the customs declaration information of the goods;
[0038] A tax number prediction unit, wherein the key information is used to characterize the identity information of the goods;
[0039] Inputting the key information of the goods into the trained tax number prediction model to obtain the predicted tax number corresponding to the goods;
[0040] The audit unit is used to obtain the declared tax number included in the customs declaration information of the goods, compare the declared tax number corresponding to the goods with the predicted tax number, and determine whether the customs declaration information of the goods is passed according to the comparison result.
[0041] On the other hand, the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for reviewing import and export goods declaration information as described in any one of the above items is implemented.
[0042] On the other hand, the present invention also provides a storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method for reviewing import and export goods declaration information as described in any one of the above items are implemented.
[0043] According to the method for reviewing the customs declaration information of imported and exported goods provided by the present invention, firstly, the customs declaration information of the goods is obtained; then the customs declaration information of the goods is input into the trained feature extraction model to obtain the key information included in the customs declaration information of the goods; the key information is used to characterize the identity information of the goods; the key information of the goods is input into the trained tax number prediction model to obtain the predicted tax number corresponding to the goods, and finally the declared tax number corresponding to the goods and the predicted tax number are compared, and whether the customs declaration information of the goods is passed is determined according to the comparison result. It can be seen that the method for reviewing the customs declaration information of the goods of the present invention can automatically query the tax number corresponding to the goods by using the principle of deep learning, and the query efficiency is high, and the technical defect of incorrect query results caused by manual negligence is overcome. Finally, the declared tax number corresponding to the goods and the predicted tax number are intelligently compared to determine whether the customs declaration information of the goods is passed, and the whole process does not require manual participation, and the audit efficiency is high and the cost is low. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Figure 1 It is a structural diagram of an import and export goods declaration information review device in a hardware operating environment involved in an embodiment of the present invention;
[0045] Figure 2 A schematic diagram of a method for reviewing import and export goods declaration information provided by an embodiment of the present invention;
[0046] Figure 3 A schematic diagram of the structure of a semantic analysis model provided by an embodiment of the present invention;
[0047] Figure 4 A schematic diagram of the structure of a feature extraction model provided by an embodiment of the present invention;
[0048] Figure 5 A schematic diagram of a training method for a tax number prediction model provided in an embodiment of the present invention;
[0049] Figure 6 A schematic diagram of the structure of an import and export goods declaration information review device provided by an embodiment of the present invention.
[0050] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings in conjunction with the embodiments. DETAILED DESCRIPTION
[0051] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.
[0052] Reference Figure 1 , Figure 1 The present invention is a schematic diagram of the structure of the import and export goods declaration information review equipment in the hardware operating environment involved in the embodiment of the present invention.
[0053] like Figure 1 As shown, the import and export goods declaration information review device (i.e., electronic device) may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a wireless fidelity (WIreless-FIdelity, WI-FI) interface). The memory 1005 may be a high-speed random access memory (Random Access Memory, RAM) memory, or a stable non-volatile memory (Non-Volatile Memory, NVM), such as a disk memory. The memory 1005 may also be a storage device independent of the aforementioned processor 1001.
[0054] Those skilled in the art will understand that Figure 1The structure shown in the figure does not constitute a limitation on the import and export goods declaration information review equipment, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.
[0055] like Figure 1 As shown, the memory 1005 as a storage medium may include an operating system, a data storage module, a network communication module, a user interface module, and an import and export goods declaration information review program.
[0056] exist Figure 1 In the import and export goods declaration information review device shown, the network interface 1004 is mainly used for data communication with other devices; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the customs declaration information review device of the present invention can be set in the customs declaration information review device, and the import and export goods declaration information review device calls the import and export goods declaration information review program stored in the memory 1005 through the processor 1001, and executes the import and export goods declaration information review method provided by the embodiment of the present invention.
[0057] The existing customs declaration information review mainly relies on manual review, which is not only inefficient, but also prone to errors in the review results due to human negligence. In addition, with the increase in trade volume, the burden of manual review is getting heavier and it is difficult to cope with the complex and changing international trade rules.
[0058] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method for reviewing the customs declaration information of imported and exported goods. After obtaining the customs declaration information of the goods, based on the concept of deep learning, a feature extraction model is used to extract the key information of the goods, and the key information of the goods is input into the trained tax number prediction model to obtain the predicted tax number corresponding to the goods; finally, the declared tax number corresponding to the goods and the predicted tax number are compared, and whether the customs declaration information of the goods is passed is determined according to the comparison result. It can be seen that the method for reviewing the customs declaration information of the goods of the present invention can automatically query the tax number corresponding to the goods by using the principle of deep learning, and the query efficiency is high, and the technical defect of incorrect query results caused by manual negligence is overcome. Finally, the declared tax number corresponding to the goods and the predicted tax number are intelligently compared to determine whether the customs declaration information of the goods is passed. The whole process does not require manual participation, and the review efficiency is high and the cost is low.
[0059] The embodiment of the present invention provides a method for reviewing import and export goods declaration information, referring to Figure 2 , Figure 2 A schematic diagram of the flow chart of a method for reviewing import and export goods declaration information provided by an embodiment of the present invention.
[0060] In this embodiment, the method for reviewing import and export goods declaration information includes:
[0061] Step S201: Obtain customs declaration information of goods.
[0062] In one embodiment, if the customs declaration information of the goods is a paper version of the customs declaration information form, the customs declaration information of the goods can be obtained by scanning and extracting text; or the paper version of the customs declaration information form can be photographed and then the customs declaration information of the goods can be obtained by extracting text information.
[0063] In another embodiment, if the customs declaration information of the goods is an electronic customs declaration information form, the customs declaration information of the goods can be directly obtained by a text information extraction method.
[0064] Step S202: input the customs declaration information of the goods into the trained feature extraction model to obtain key information included in the customs declaration information of the goods; the key information is used to characterize the identity information of the goods.
[0065] In one embodiment, the feature extraction model may be a semantic analysis model. For example, the present embodiment may use natural language processing (NLP) technology to construct a semantic analysis model. The core function of the model is to understand and parse the commodity name and specification model fields on the customs declaration form, thereby extracting key information from the customs declaration information on the customs declaration form and obtaining the key information included in the goods customs declaration information.
[0066] Figure 3 For a schematic diagram of the semantic analysis model structure provided by an embodiment of the present invention, please refer to Figure 3 As shown, the semantic analysis model includes a word segmentation algorithm module 301 , a part-of-speech tagging module 302 , a dependency syntax analysis module 303 and an entity recognition module 304 .
[0067] When extracting key information included in the cargo declaration information, first, the word segmentation algorithm module 301 decomposes the text in the commodity name and specification model fields into separate vocabulary units through the word segmentation algorithm. This step is the basis of semantic analysis, because the general prediction model can only recognize and process a single word, but cannot process the entire text block.
[0068] Next, the part-of-speech tagging module 302 uses part-of-speech tagging technology to identify the grammatical function of each vocabulary unit, such as noun, verb or adjective, etc. This helps the model understand the role and meaning of each vocabulary in the text sentence.
[0069] Then, the dependency syntax analysis module 303 uses dependency syntax analysis to determine the relationship between words, such as which words are modifiers or objects of other words. This step is particularly important for understanding complex product names and specifications, because they usually contain multiple attributes and descriptive phrases. Through this dependency analysis, the model can construct a structured representation between words, revealing the semantic relationship between vocabulary units in the text.
[0070] Finally, the entity recognition module 304 uses entity recognition technology to identify and classify key information in the text, such as product type, brand, model and other specifications. This information is key to reviewing customs declaration information because it is directly related to the tax number and classification of the product.
[0071] In one embodiment, the inputting of the customs declaration information of the goods into the trained semantic analysis model to obtain key information included in the customs declaration information of the goods includes:
[0072] Input the customs declaration information of the goods into the word segmentation algorithm module 301, the word segmentation algorithm module 301 is used to perform text decomposition on the customs declaration information of the goods to obtain multiple vocabulary units, and output the multiple vocabulary units to the part-of-speech tagging module 302;
[0073] The part-of-speech tagging module 302 is used to identify the part of speech of each vocabulary unit, tag each vocabulary unit according to the identified part of speech, and output the tagged vocabulary unit to the dependency syntax analysis module 303;
[0074] The dependency syntax analysis module 303 is used to analyze the dependency relationship between the multiple annotated vocabulary units, annotate the dependency relationship between the multiple vocabulary units according to the analyzed dependency relationship, and output the annotated multiple vocabulary units to the entity recognition module 304;
[0075] The entity recognition module 304 is used to determine at least one of the multiple vocabulary units as the key information of the goods according to the parts of speech of the multiple vocabulary units and the dependency relationship between the multiple vocabulary units.
[0076] It should be noted that the above-mentioned word segmentation algorithm module 301, part-of-speech tagging module 302, dependency syntax analysis module 303 and entity recognition module 304 are all functional modules that have been trained and verified by neural networks. The training data can be the field of goods trade in this solution, or the field of text processing technology. The training method can refer to the training of existing neural networks. For example, the word segmentation algorithm module 301 can use a statistical-based method to determine the most likely vocabulary boundaries, while the entity recognition module 304 can use a machine learning model to identify and classify entities. Through the semantic analysis model composed of the above-mentioned trained functional modules, this embodiment can automatically extract key information for commodity tax number prediction from customs declaration information, replacing the customs declaration information review work that originally required a lot of manual labor, thereby improving the efficiency and accuracy of customs declaration information review.
[0077] For example, using the semantic analysis model provided in this embodiment, assume that the name of the commodity on the customs declaration is "Brand A, Model B, 5G mobile phone". The semantic analysis model first segments it into "Brand A", "Model B", "5G" and "mobile phone". Then, the semantic analysis model recognizes "Brand A, Model B" as the brand and model, and "5G" and "mobile phone" as the key attributes of the commodity. Finally, the key information of the current goods can be output as "mobile phone".
[0078] In one embodiment, the embodiment also provides a feature extraction model structure, Figure 4 For a schematic diagram of the feature extraction model structure provided by an embodiment of the present invention, please refer to Figure 4 The feature extraction model includes a CNN module 401, a LSTM module 402 and an association rule learning module 403.
[0079] CNN module 401 uses a convolutional neural network (CNN) to extract features from text data. CNN module 401 can capture local correlations when processing text data, and is used to identify key information in product names. For example, if the input product name is "Brand A Model B 5G Mobile Phone", CNN module 401 can identify "Brand A Model B" and distinguish it from "5G" and "Mobile Phone".
[0080] LSTM module 402 uses a long short-term memory network (LSTM) to perform sequence modeling on the extracted features. LSTM can handle long-distance dependency problems in text data and help the model understand the contextual relationship in the product name. For example, LSTM can recognize that "5G" is a supplementary description of the product feature "mobile phone".
[0081] The association rule learning module 403 uses an association rule learning algorithm, such as the Apriori algorithm, to learn the association rules between commodity features and tax numbers from historical customs declaration data. This step enables the model to understand which commodity features are usually associated with a specific tax number, thereby helping to extract key information for predicting the tax number.
[0082] In one embodiment, the inputting of the customs declaration information of the goods into the trained feature extraction model to obtain key information included in the customs declaration information of the goods includes:
[0083] The customs declaration information of the goods is input into the CNN module 401, and the CNN module 401 is used to extract vocabulary units from the customs declaration information of the goods to obtain multiple vocabulary units, and input the multiple vocabulary units into the LSTM module 402;
[0084] The LSTM module 402 is used to obtain the association relationship between the multiple vocabulary units, mark the association relationship between the multiple vocabulary units, and input the marked multiple vocabulary units into the association rule learning module 403;
[0085] The association rule learning module 403 is used to determine at least one of the multiple vocabulary units as the key information of the goods according to the association relationship between the multiple vocabulary units.
[0086] In one embodiment, the feature extraction model further includes a data preprocessing module, which is used to preprocess the customs declaration information before inputting it into the CNN module 401. The preprocessing method includes but is not limited to data cleaning, error correction, missing information completion, data unification and data standardization. For example, the data preprocessing module first cleans the commodity name and specification model fields on the customs declaration form, including removing illegal characters, unifying synonyms, etc., to ensure the quality of the data input into the CNN module 401.
[0087] For example, for the input customs declaration information of "Brand A, Model B 5G mobile phone", the model first uses the data preprocessing module to preprocess the data, and then uses the CNN module 401 to extract vocabulary units (equivalent to keywords), such as extracting the vocabulary units "Brand A", "Model B" and "mobile phone". Next, the LSTM module 402 processes these vocabulary units to capture the sequence relationship between "5G" and "mobile phone". The Apriori algorithm model learns that it is usually associated with a specific tax number. Finally, the association rule learning module 403 determines the vocabulary unit "mobile phone" as the key information of the goods based on the association relationship between multiple vocabulary units.
[0088] Step S203: input the key information of the goods into the trained tax number prediction model to obtain the predicted tax number corresponding to the goods.
[0089] In this embodiment, tax number prediction model training is a crucial step. In order to improve the prediction accuracy and generalization ability of the tax number prediction model, this embodiment uses a large amount of historical customs declaration data (i.e., customs declaration information) as a training set. These historical customs declaration data contain the name, specification model and corresponding tax number of the goods, which can be selected from past customs declarations and obtained through preprocessing. In the model training process, the cross-validation method can be used to avoid overfitting, and the performance of the model can be optimized by adjusting the model parameters.
[0090] This embodiment provides a method for training a tax number prediction model. Figure 5 For a flow chart of the training method of the tax number prediction model provided in the embodiment of the present invention, please refer to Figure 5 As shown, the training method includes:
[0091] S501, obtaining sample key information corresponding to the goods sample and the sample tax number corresponding to each goods sample;
[0092] S502, inputting the sample key information into the initial tax number prediction model to obtain the predicted sample tax number corresponding to each cargo sample;
[0093] S503. Update the model parameters of the initial tax number prediction model according to the sample tax number and the predicted sample tax number to obtain a trained tax number prediction model.
[0094] Specifically, when training the initial tax number prediction model, the historical customs declaration data is first divided into a training set and a validation set, where the training set is used to learn the model, and the validation set is used to evaluate the prediction effect of the model. During the training phase, the model will try to learn the key features in the product name and specification model, and associate these features with the correct tax number. A loss function can be used to measure the difference between the predicted tax number output by the model and the actual tax number, and the model parameters can be continuously adjusted through the gradient descent algorithm to minimize the value of the loss function, thereby obtaining a trained tax number prediction model.
[0095] For example, the training process of the tax number prediction model can be expressed by the following formula:
[0096]
[0097] In the above formula, L(θ) is the loss function, θ is the model parameter, N is the number of training samples, and y i is the actual tax number of the i-th sample, f(x i ; θ) is the tax number predicted by the model, xi is the input feature of the i-th sample. The training goal is to find a set of optimal model parameters that minimizes the value of the loss function.
[0098] For example, a batch of historical customs declaration data is obtained, which contains 10,000 records. First, these data are divided into a training set and a validation set in a ratio of 8:2, that is, 8,000 records are used for training the tax number prediction model, and 2,000 records are used for verifying the tax number prediction model. The tax number prediction model of this embodiment can adopt a support vector machine (SVM) model. Through multiple rounds of iterative training, it is observed that the value of the model loss function gradually decreases and the prediction accuracy of the model gradually increases. In the verification stage, the verification set can be used to evaluate the performance of the model, and it is determined that the accuracy of the model has reached more than 90%, which indicates that the model has good prediction ability. In this way, a tax number prediction model with high accuracy can be successfully trained, which provides strong technical support for the review of customs declaration information of goods in customs declaration forms.
[0099] Step S204: obtaining the declared tax number included in the customs declaration information of the goods, comparing the declared tax number corresponding to the goods with the predicted tax number, and determining whether the customs declaration information of the goods is passed according to the comparison result.
[0100] In one embodiment, if the declared tax number and the predicted tax number corresponding to the goods are consistent, it is determined that the customs declaration information of the goods has passed; otherwise, it is determined that the customs declaration information of the goods is wrong, and an early warning signal is issued, wherein the early warning signal is used to indicate that the declared tax number and the predicted tax number corresponding to the goods are inconsistent.
[0101] It can be understood that according to the audit method of this embodiment, when it is determined that the declared tax number and the predicted tax number corresponding to the goods are inconsistent, the early warning mechanism can be triggered immediately to remind the staff to deal with it in time, which can greatly improve the efficiency and accuracy of the customs declaration form review.
[0102] In one embodiment, after the warning signal is issued, the warning signal is also sent to a designated client or database. For example, the warning signal is sent to the mobile phone of a designated staff member to prompt that there is inconsistency in the customs declaration information of the current goods and it needs to be processed in time. Or the warning signal is sent to a designated database for storage to facilitate later viewing.
[0103] In one embodiment, after the warning signal is issued, the warning signal is also pushed to a designated client, and a warning signal pop-up window is generated on the display interface of the client, so as to remind the staff in real time to deal with the customs declaration information errors in time. The warning signal can be an email notification, SMS reminder or a pop-up window warning within the system to ensure that relevant personnel can be notified in time and take corresponding measures.
[0104] For example, in real-time monitoring, the audit system processes a customs declaration information. The predicted tax number predicted by the tax number prediction model is "1234", while the declared tax number in the customs declaration information form is "5678". After the audit system detects this inconsistency in tax numbers, it immediately marks the record in red on the monitoring interface and sends an email notification to the auditor. After receiving the notification, the auditor can quickly log in to the system to view the detailed information and conduct further manual review of the customs declaration information form. This real-time early warning mechanism reduces the workload of manual review, and also improves the accuracy and timeliness of the customs declaration information review. The audit method provided in this embodiment can ensure the compliance of the commodity declaration information on the customs declaration form, and effectively prevent the occurrence of inconsistencies, false reports, concealment, etc.
[0105] The import and export goods declaration information review device provided by the present invention is described below. The import and export goods declaration information review device described below and the import and export goods declaration information review method described above can be referenced to each other.
[0106] Figure 6 For a schematic diagram of the structure of the import and export goods declaration information review device provided by the embodiment of the present invention, please refer to Figure 6 As shown, the customs declaration information review device includes:
[0107] The acquisition unit 601 is used to acquire the customs declaration information of the goods;
[0108] A feature extraction unit 602 is used to input the customs declaration information of the goods into the trained feature extraction model to obtain key information included in the customs declaration information of the goods;
[0109] The tax number prediction unit 603, wherein the key information is used to characterize the identity information of the goods;
[0110] Inputting the key information of the goods into the trained tax number prediction model to obtain the predicted tax number corresponding to the goods;
[0111] The review unit 604 is used to obtain the declared tax number included in the customs declaration information of the goods, compare the declared tax number corresponding to the goods with the predicted tax number, and determine whether the customs declaration information of the goods is passed according to the comparison result.
[0112] In one embodiment, the feature extraction model is a semantic analysis model; the semantic analysis model includes a word segmentation algorithm module, a part-of-speech tagging module, a dependency syntax analysis module and an entity recognition module.
[0113] In one embodiment, the feature extraction unit 602 is specifically used for:
[0114] The step of inputting the customs declaration information of the goods into the trained semantic analysis model to obtain key information included in the customs declaration information of the goods includes:
[0115] Inputting the customs declaration information of the goods into the word segmentation algorithm module, the word segmentation algorithm module is used to perform text decomposition on the customs declaration information of the goods to obtain multiple vocabulary units, and outputting the multiple vocabulary units to the part-of-speech tagging module;
[0116] The part-of-speech tagging module is used to identify the part of speech of each vocabulary unit, tag each vocabulary unit according to the identified part of speech, and output the tagged vocabulary unit to the dependency syntax analysis module;
[0117] The dependency syntax analysis module is used to analyze the dependency relationship between the multiple annotated vocabulary units, annotate the dependency relationship between the multiple vocabulary units according to the analyzed dependency relationship, and output the annotated multiple vocabulary units to the entity recognition module;
[0118] The entity recognition module is used to determine at least one of the multiple vocabulary units as the key information of the goods according to the parts of speech of the multiple vocabulary units and the dependency relationship between the multiple vocabulary units.
[0119] In one embodiment, the feature extraction model includes a CNN module, a LSTM module and an association rule learning module.
[0120] In one embodiment, the feature extraction unit 602 is specifically used for:
[0121] Inputting the customs declaration information of the goods into the CNN module, the CNN module is used to extract vocabulary units from the customs declaration information of the goods to obtain multiple vocabulary units, and inputting the multiple vocabulary units into the LSTM module;
[0122] The LSTM module is used to obtain the association relationship between the multiple vocabulary units, mark the association relationship between the multiple vocabulary units, and input the marked multiple vocabulary units into the association rule learning module;
[0123] The association rule learning module is used to determine at least one of the multiple vocabulary units as the key information of the goods according to the association relationship between the multiple vocabulary units.
[0124] In one embodiment, the customs declaration information review device also includes a model training unit, which is used to train the feature extraction model and the tax number prediction model using historical data.
[0125] In one embodiment, the model training unit is specifically used for:
[0126] Obtain key sample information corresponding to the cargo samples, as well as the sample tax number corresponding to each cargo sample;
[0127] Input the sample key information into the initial tax number prediction model to obtain the predicted sample tax number corresponding to each cargo sample;
[0128] According to the sample tax number and the predicted sample tax number, the model parameters of the initial tax number prediction model are updated to obtain a trained tax number prediction model.
[0129] In one embodiment, the model training unit is further used to:
[0130] After obtaining the trained tax number prediction model, obtain the sample key information corresponding to the goods test sample and the sample tax number corresponding to each goods test sample;
[0131] Inputting the sample key information corresponding to the cargo test sample into the trained tax number prediction model to obtain the test sample tax number;
[0132] Determine whether the trained tax number prediction model meets the preset requirements based on the sample tax number corresponding to the cargo test sample and the test sample tax number. If so, terminate the model training and output the trained tax number prediction model. Otherwise, increase the number of iterations and continue to train the trained tax number prediction model.
[0133] In one embodiment, the review unit 604 is further configured to:
[0134] If the declared tax number and the predicted tax number corresponding to the goods are consistent, it is determined that the customs declaration information of the goods has passed; otherwise, it is determined that the customs declaration information of the goods is wrong, and an early warning signal is issued, which is used to indicate that the declared tax number and the predicted tax number corresponding to the goods are inconsistent.
[0135] In one embodiment, the review unit 604 is further configured to:
[0136] The warning signal is sent to a designated client or database; and / or the warning signal is pushed to a designated client, and a warning signal pop-up window is generated on a display interface of the client.
[0137] In one embodiment, the customs declaration information review device also includes a display unit, which includes a display interface for displaying the customs declaration information of the goods, customs declaration review results or early warning information, etc., to support customs declaration personnel to conduct detailed review and processing of this information.
[0138] In one embodiment, the customs declaration information review device further includes an input unit for the customs declaration personnel to input manual review result information when processing the customs declaration information, customs declaration review results or early warning information.
[0139] The review unit 604 sends the review result of the customs declaration information to the display unit, and the sending method includes but is not limited to email, text message, system notification, etc. After receiving the warning information, the customs declaration personnel can directly access the relevant customs declaration record through the user interface of the display unit and start the review process. During the review process, the customs declaration personnel can view the detailed description of the goods, relevant trade regulations, and historical declaration data, etc., to help determine whether there are indeed problems with the declaration information.
[0140] The present embodiment also provides a storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method for reviewing the customs declaration information of imported and exported goods are implemented as described above; the method for reviewing the customs declaration information includes: obtaining the customs declaration information of the goods; inputting the customs declaration information of the goods into a trained feature extraction model to obtain key information included in the customs declaration information of the goods; the key information is used to characterize the identity information of the goods; inputting the key information of the goods into a trained tax number prediction model to obtain the predicted tax number corresponding to the goods; obtaining the declared tax number included in the customs declaration information of the goods, comparing the declared tax number corresponding to the goods with the predicted tax number, and determining whether the customs declaration information of the goods is passed based on the comparison result.
[0141] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.
[0142] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.
[0143] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.
[0144] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.
Claims
1. A method for reviewing import and export goods declaration information, characterized in that: include: Obtain customs declaration information for goods; Inputting the customs declaration information of the goods into the trained feature extraction model to obtain key information included in the customs declaration information of the goods; the key information is used to characterize the identity information of the goods; Inputting the key information of the goods into the trained tax number prediction model to obtain the predicted tax number corresponding to the goods; Obtaining the declared tax number included in the customs declaration information of the goods, comparing the declared tax number corresponding to the goods with the predicted tax number, and determining whether the customs declaration information of the goods is passed according to the comparison result; The feature extraction model includes a CNN module, an LSTM module and an association rule learning module; The step of inputting the customs declaration information of the goods into the trained feature extraction model to obtain key information included in the customs declaration information of the goods includes: Inputting the customs declaration information of the goods into the CNN module, the CNN module is used to extract vocabulary units from the customs declaration information of the goods to obtain multiple vocabulary units, and inputting the multiple vocabulary units into the LSTM module; The LSTM module is used to obtain the association relationship between the multiple vocabulary units, mark the association relationship between the multiple vocabulary units, and input the marked multiple vocabulary units into the association rule learning module; The association rule learning module is used to learn the association rules between commodity features and tax numbers according to the association rule learning algorithm, which helps to extract key information for predicting tax numbers and determine at least one of the multiple vocabulary units as the key information of the goods.
2. The method for reviewing import and export goods declaration information according to claim 1, characterized in that: Before inputting the key information of the goods into the trained tax number prediction model, it also includes: Obtain key sample information corresponding to the cargo samples, as well as the sample tax number corresponding to each cargo sample; Input the sample key information into the initial tax number prediction model to obtain the predicted sample tax number corresponding to each cargo sample; According to the sample tax number and the predicted sample tax number, the model parameters of the initial tax number prediction model are updated to obtain a trained tax number prediction model.
3. The method for reviewing import and export goods declaration information as claimed in claim 2, characterized in that: After obtaining the trained tax number prediction model, it also includes: Obtain the sample key information corresponding to the cargo test sample and the sample tax number corresponding to each cargo test sample; Inputting the sample key information corresponding to the cargo test sample into the trained tax number prediction model to obtain the test sample tax number; Determine whether the trained tax number prediction model meets the preset requirements based on the sample tax number corresponding to the cargo test sample and the test sample tax number. If so, terminate the model training and output the trained tax number prediction model. Otherwise, increase the number of iterations and continue to train the trained tax number prediction model.
4. The method for reviewing import and export goods declaration information according to claim 1, characterized in that: The process of comparing the declared tax number and the predicted tax number corresponding to the goods, and determining whether the customs declaration information of the goods is passed according to the comparison result, includes: If the declared tax number and the predicted tax number corresponding to the goods are consistent, it is determined that the customs declaration information of the goods has passed; otherwise, it is determined that the customs declaration information of the goods is wrong, and an early warning signal is issued, which is used to indicate that the declared tax number and the predicted tax number corresponding to the goods are inconsistent.
5. The method for reviewing import and export goods declaration information as claimed in claim 4, characterized in that: After the early warning signal is issued, it also includes: Sending the warning signal to a designated client or database; and / or The warning signal is pushed to a designated client, and a warning signal pop-up window is generated on the display interface of the client.
6. A device for reviewing import and export goods declaration information, characterized in that: include: An acquisition unit, used to acquire the customs declaration information of goods; A feature extraction unit, used for inputting the customs declaration information of the goods into the trained feature extraction model to obtain key information included in the customs declaration information of the goods; the feature extraction model includes a CNN module, an LSTM module and an association rule learning module; A tax number prediction unit, wherein the key information is used to characterize the identity information of the goods; Inputting the key information of the goods into the trained tax number prediction model to obtain the predicted tax number corresponding to the goods; An audit unit, used to obtain the declared tax number included in the customs declaration information of the goods, compare the declared tax number corresponding to the goods with the predicted tax number, and determine whether the customs declaration information of the goods is passed according to the comparison result; The feature extraction unit is specifically used to: input the customs declaration information of the goods into the CNN module, the CNN module is used to extract vocabulary units from the customs declaration information of the goods to obtain multiple vocabulary units, and input the multiple vocabulary units into the LSTM module; the LSTM module is used to obtain the association relationship between the multiple vocabulary units, annotate the association relationship between the multiple vocabulary units, and input the annotated multiple vocabulary units into the association rule learning module; The association rule learning module is used to learn the association rules between commodity features and tax numbers according to the association rule learning algorithm, which helps to extract key information for predicting tax numbers and determine at least one of the multiple vocabulary units as the key information of the goods.
7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method for reviewing import and export goods declaration information as described in any one of claims 1 to 5 is implemented.
8. A storage medium, characterized in that: The storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for reviewing import and export goods declaration information as described in any one of claims 1 to 5.
