Intelligent purchasing method based on AI

Abstract

The invention discloses an intelligent purchasing method based on AI, and belongs to the technical field of purchasing, the intelligent purchasing method can help a user to screen suppliers more quickly and select suppliers more suitable for the user by introducing a supplier recommendation mechanism, and the user experience is improved. And meanwhile, an AI analysis mechanism is introduced to intelligently analyze a purchase contract, a purchase invoice and a payment application, so that intelligent proofreading of information can be realized, and the information checking efficiency in the purchase process can be effectively improved, thereby avoiding potential legal risks.

Description

Technical Field

[0001] The present invention belongs to the field of procurement technology, and specifically relates to an AI-based intelligent procurement method. Background Art

[0002] In the procurement business, there are usually some problems and challenges in the process of supplier selection, procurement contract signing, invoice entry and payment. The procurement department, project team and finance department are responsible for the selection of suppliers, procurement contract signing, invoice collection, invoice and contract information entry system, payment application and payment approval. Some companies may still rely on manual selection of suppliers, manual processing of procurement contracts, invoices and payments, including paper contracts, manual entry of invoice and contract information, and manual verification of payment vouchers, which is inefficient and prone to errors. There may be compliance risks in the process of supplier selection, procurement contract execution, invoice management and payment. Failure to strictly comply with relevant laws and regulations or internal control requirements may lead to potential legal risks. Summary of the invention

[0003] The present invention provides an AI-based intelligent procurement method to solve the problems existing in the prior art.

[0004] An AI-based smart procurement method, including:

[0005] Construct supplier scoring rules, purchase invoice identification and verification rules, purchase contract verification rules, and payment application verification rules;

[0006] Obtaining the demand keywords input by the user, and recommending suppliers according to the supplier scoring rules and the demand keywords, so that the user can purchase from the recommended suppliers and generate purchase invoices, purchase contracts and payment applications;

[0007] Based on the purchase invoice identification and verification rules, AI analysis is performed on the purchase invoice to determine a first analysis result corresponding to the purchase invoice; wherein the first analysis result includes the presence of an abnormality or the absence of an abnormality;

[0008] Based on the procurement contract verification rule, AI analysis is performed on the procurement contract to determine a second analysis result corresponding to the procurement contract; wherein the second analysis result includes the presence of an abnormality or the absence of an abnormality;

[0009] Based on the payment application verification rule, AI analysis is performed on the payment application to determine a third analysis result corresponding to the payment application; wherein the third analysis result includes whether an abnormality exists or does not exist;

[0010] When the first analysis result, the second analysis result and / or the third analysis result are abnormal, it is determined that the verification has failed, and a manual review request is initiated to the server, and a warning is given;

[0011] When the first analysis result, the second analysis result and the third analysis result show that there is no abnormality, the verification is determined to be successful, the procurement data is archived, and the intelligent procurement process is completed.

[0012] Furthermore, supplier scoring rules, purchase invoice identification and verification rules, purchase contract verification rules, and payment application verification rules are constructed, including:

[0013] Determine supplier evaluation indicators, de-dimensionalize the supplier evaluation indicators, and then perform weighted summation on all supplier evaluation indicators to obtain supplier scoring rules;

[0014] Constructing a purchase invoice correction sub-rule, a sub-rule for cutting the first text area of ​​the purchase invoice, a sub-rule for identifying the first text in the first text area, and a sub-rule for sorting the identified first text, to obtain a purchase invoice identification and verification rule;

[0015] Constructing a procurement contract correction sub-rule, a sub-rule for cutting the second text area of ​​the procurement contract, a sub-rule for identifying the second text in the second text area, and a sub-rule for sorting the identified second text, to obtain a procurement contract identification and verification rule;

[0016] Construct verification rules for the business license, tax registration certificate, organization code certificate, legal representative ID card, and corporate bank account information in the payment application to obtain the payment application verification rules.

[0017] Furthermore, the supplier evaluation index is determined, and after the supplier evaluation index is de-dimensionalized, all supplier evaluation indexes are weighted and summed to obtain the supplier scoring rules, including:

[0018] Determine supplier evaluation indicators; wherein the supplier evaluation indicators include material unit price, material demand quantity, material receipt quantity, material qualification rate, supply cycle, supplier purchase amount, supplier purchase times, supplier registered capital and supplier grade; wherein the supplier grade is the average of the historical scores given by the purchaser to the supplier;

[0019] With respect to the supplier evaluation indicators, the supplier evaluation indicators are made homogenized to obtain the homogenized supplier evaluation indicators;

[0020] After de-dimensionalizing the supplier evaluation indicators after homogenization, all supplier evaluation indicators are weighted and summed to obtain the supplier scoring rules.

[0021] Furthermore, constructing a purchase invoice correction sub-rule, a sub-rule for cutting the first text area of ​​the purchase invoice, a sub-rule for identifying the first text in the first text area, and a sub-rule for sorting the identified first text, to obtain a purchase invoice recognition and verification rule, including:

[0022] Construct a rule for correcting the tilt of purchase invoices and obtain a purchase invoice correction sub-rule;

[0023] The YOLO algorithm is used to construct the recognition rule of the first text area, and the sub-rule of cutting the first text area of ​​the purchase invoice is obtained;

[0024] The CDNets algorithm is used to construct a sub-rule for identifying the first character in the first character region;

[0025] Constructing a rule for formatting text processing of the first character, and obtaining a sub-rule for collating the recognized first character;

[0026] The purchase invoice recognition and verification rule is obtained according to the purchase invoice correction sub-rule, the sub-rule for cutting the first text area of ​​the purchase invoice, the sub-rule for identifying the first text in the first text area and the sub-rule for sorting the identified first text.

[0027] Furthermore, constructing a procurement contract correction sub-rule, a sub-rule for cutting the second text area of ​​the procurement contract, a sub-rule for identifying the second text in the second text area, and a sub-rule for sorting the identified second text, to obtain a procurement contract identification and verification rule, including:

[0028] Construct a purchase contract tilt correction rule to obtain a purchase contract correction sub-rule;

[0029] The YOLO algorithm is used to construct the recognition rule of the second text area, and the sub-rule of cutting the second text area of ​​the procurement contract is obtained;

[0030] The CDNets algorithm is used to construct a sub-rule for identifying the second character in the second character area;

[0031] Constructing a rule for formatting text processing of the second character, and obtaining a sub-rule for collating the recognized second character;

[0032] According to the procurement contract correction sub-rule, the sub-rule for cutting the second text area of ​​the procurement contract, the sub-rule for identifying the second text in the second text area and the sub-rule for sorting the identified second text, the procurement contract recognition and verification rule is obtained.

[0033] Furthermore, a verification rule for the business license, tax registration certificate, organization code certificate, legal representative ID card and corporate bank account information in the payment application is constructed to obtain the payment application verification rule, including:

[0034] Construct text extraction and certificate number online query rules for business license, tax registration certificate, organization code certificate, legal representative ID card and corporate bank account information in payment applications to obtain payment application verification rules.

[0035] Further, the demand keywords input by the user are obtained, and suppliers are recommended according to the supplier scoring rules and the demand keywords, so that the user can purchase according to the recommended suppliers, and generate purchase invoices, purchase contracts and payment applications, including:

[0036] Obtain the demand keywords input by the user and determine the target suppliers with the demand keywords;

[0037] According to the supplier scoring rules, each target supplier is evaluated to determine the comprehensive score corresponding to the supplier evaluation index;

[0038] The top N target suppliers with the largest comprehensive scores corresponding to the supplier evaluation indicators are recommended to the user, so that the user can purchase according to the recommended suppliers and generate purchase invoices, purchase contracts and payment applications.

[0039] Furthermore, based on the purchase invoice identification and verification rules, AI analysis is performed on the purchase invoice to determine a first analysis result corresponding to the purchase invoice, including:

[0040] According to the purchase invoice correction sub-rule, the purchase invoice is corrected to obtain the corrected purchase invoice;

[0041] Using the corrected purchase invoice as an input of the recognition rule for the first text region to obtain the first text region in the purchase invoice;

[0042] Recognize the first character region using a sub-rule for recognizing the first character in the first character region to obtain the first character;

[0043] Arrange the first character using the sub-rule for arranging the recognized first character to obtain the arranged first character;

[0044] The sorted first text is checked against the bill content recorded in the financial management system to determine a first analysis result corresponding to the purchase invoice.

[0045] Furthermore, based on the procurement contract verification rules, AI analysis is performed on the procurement contract to determine a second analysis result corresponding to the procurement contract, including:

[0046] Correct the purchase contract according to the purchase contract correction sub-rule to obtain the corrected purchase contract;

[0047] Using the corrected purchase contract as an input of the recognition rule of the second text region to obtain the second text region in the purchase contract;

[0048] Recognize the second character region using a sub-rule for recognizing the second character in the second character region to obtain the second character;

[0049] Arrange the second character using the sub-rule for arranging the recognized second character to obtain an arranged second character;

[0050] The collated second text is checked against the contract content recorded in the financial management system to determine the second analysis result corresponding to the procurement contract.

[0051] Furthermore, based on the payment application verification rule, AI analysis is performed on the payment application to determine a third analysis result corresponding to the payment application, including:

[0052] Parse the text information of the business license, tax registration certificate, organization code certificate, legal representative ID card and corporate bank account information in the payment application to obtain key text information;

[0053] Based on the payment application verification rule, the key text information is queried online to determine the third analysis result corresponding to the payment application.

[0054] The present invention provides an AI-based intelligent procurement method, which can help users screen suppliers more quickly and select suppliers that are more suitable for them by introducing a supplier recommendation mechanism. At the same time, it introduces an AI analysis mechanism to perform intelligent analysis on procurement contracts, procurement invoices and payment applications, which can realize intelligent proofreading of information and effectively improve the efficiency of information verification in the procurement process, thereby avoiding potential legal risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0056] Figure 1 A flowchart of an AI-based intelligent procurement method provided in an embodiment of the present invention.

[0057] The above drawings have shown clear embodiments of the present invention, which will be described in more detail below. These drawings and text descriptions are not intended to limit the scope of the present invention in any way, but to illustrate the concept of the present invention for those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0058] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0059] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0060] like Figure 1 As shown, an embodiment of the present invention provides an AI-based intelligent procurement method, including:

[0061] S101. Construct supplier scoring rules, purchase invoice identification and verification rules, purchase contract verification rules, and payment application verification rules;

[0062] The supplier scoring rules can be rules for statistically analyzing the indicators of suppliers. The comprehensive evaluation of suppliers can be determined through this rule, and suppliers with high comprehensive evaluations can be recommended to users, thereby realizing procurement recommendations and making it more convenient for users to choose suppliers that are suitable for them.

[0063] The purchase invoice identification and verification rules can be some rules for verifying the invoice content, for example, the text information in the paper purchase invoice can be identified, and the text information can be compared with the archived electronic files to determine whether there are any problems with the information. At the same time, the information in the purchase invoice can also be queried online to determine whether the information in the invoice is correct, for example, the invoice header and tax number can be queried online to determine whether the information is wrong.

[0064] The procurement contract verification rules can be similar to the procurement invoice identification and verification rules. For example, the text information in the paper procurement contract can be compared with the archived electronic file to determine whether the information in the procurement contract is accurate and avoid potential legal risks.

[0065] The payment application verification rules can be to conduct online query of the business license, tax registration certificate, organization code certificate, legal representative ID card and corporate bank account information to determine whether there are any abnormalities in the payment information.

[0066] S102, obtaining the demand keywords input by the user, and recommending suppliers according to the supplier scoring rules and the demand keywords, so that the user can purchase from the recommended suppliers and generate a purchase invoice, a purchase contract and a payment application;

[0067] Each supplier is pre-set with its own corresponding supply keywords. The similarity between the demand keywords and the supply keywords can be calculated to determine the suppliers that meet the demand keywords. Then, the determined suppliers are scored according to the supplier scoring rules to achieve TOPN recommendation. If the TOPN recommendation is not met, the found suppliers will be arranged in descending order according to the score and recommended.

[0068] S103. Based on the purchase invoice recognition and verification rules, perform AI (Artificial Intelligence) analysis on the purchase invoice to determine a first analysis result corresponding to the purchase invoice; wherein the first analysis result includes whether an abnormality exists or does not exist;

[0069] After users make purchases with suppliers, purchase invoices, purchase contracts and payment applications will be generated. By collecting the paper and electronic files of these documents, real-time verification can be performed to ensure that the information is correct.

[0070] These checks are not limited to online query of information in paper documents, online query of information in electronic files, and consistency check of information in paper documents and electronic files. The embodiment of the present invention mainly performs consistency check of information in paper documents and electronic files to determine the correctness of the information.

[0071] S104. Based on the procurement contract verification rule, perform AI analysis on the procurement contract to determine a second analysis result corresponding to the procurement contract; wherein the second analysis result includes whether an abnormality exists or does not exist;

[0072] As mentioned above, the procurement contract verification rules can be used to conduct online inquiries on the procurement contract or to check the consistency of the information in the paper document and the electronic file to determine whether the procurement contract information is accurate.

[0073] S105. Based on the payment application verification rule, perform AI analysis on the payment application to determine a third analysis result corresponding to the payment application; wherein the third analysis result includes whether an abnormality exists or does not exist;

[0074] Payment applications need to strictly ensure the accuracy of data, so the information in the payment application can be parsed and queried online to determine whether the payment application is abnormal.

[0075] S106: When the first analysis result, the second analysis result and / or the third analysis result are abnormal, it is determined that the verification has failed, a manual review request is sent to the server, and a warning is given;

[0076] For example: when any analysis result is abnormal, an abnormal warning message about the analysis can be generated and a manual review can be requested, so as to effectively troubleshoot the abnormality and improve the security of the procurement process and the accuracy of information.

[0077] S107. When the first analysis result, the second analysis result and the third analysis result show that there is no abnormality, it is determined that the verification is successful, the procurement data is archived, and the intelligent procurement process is completed.

[0078] The present invention provides an AI-based intelligent procurement method, which can help users screen suppliers more quickly and select suppliers that are more suitable for them by introducing a supplier recommendation mechanism. At the same time, it introduces an AI analysis mechanism to perform intelligent analysis on procurement contracts, procurement invoices and payment applications, which can realize intelligent proofreading of information and effectively improve the efficiency of information verification in the procurement process, thereby avoiding potential legal risks.

[0079] In an embodiment of the present invention, supplier scoring rules, purchase invoice identification and verification rules, purchase contract verification rules, and payment application verification rules are constructed, including:

[0080] Determine supplier evaluation indicators, de-dimensionalize the supplier evaluation indicators, and then perform weighted summation on all supplier evaluation indicators to obtain supplier scoring rules;

[0081] Constructing a purchase invoice correction sub-rule, a sub-rule for cutting the first text area of ​​the purchase invoice, a sub-rule for identifying the first text in the first text area, and a sub-rule for sorting the identified first text, to obtain a purchase invoice identification and verification rule;

[0082] Constructing a procurement contract correction sub-rule, a sub-rule for cutting the second text area of ​​the procurement contract, a sub-rule for identifying the second text in the second text area, and a sub-rule for sorting the identified second text, to obtain a procurement contract identification and verification rule;

[0083] Construct verification rules for the business license, tax registration certificate, organization code certificate, legal representative ID card, and corporate bank account information in the payment application to obtain the payment application verification rules.

[0084] The embodiment of the present invention can intelligently verify the information in the procurement process by constructing supplier scoring rules, procurement invoice identification and verification rules, procurement contract verification rules and payment application verification rules, which can effectively improve the verification efficiency and verification accuracy.

[0085] In the embodiment of the present invention, after determining the supplier evaluation index and de-dimensionalizing the supplier evaluation index, weighted summing is performed on all supplier evaluation indexes to obtain supplier scoring rules, including:

[0086] Determine supplier evaluation indicators; wherein the supplier evaluation indicators include material unit price, material demand quantity, material receipt quantity, material qualification rate, supply cycle, supplier purchase amount, supplier purchase times, supplier registered capital and supplier grade; wherein the supplier grade is the average of the historical scores given by the purchaser to the supplier;

[0087] With respect to the supplier evaluation indicators, the supplier evaluation indicators are made homogenized to obtain the homogenized supplier evaluation indicators;

[0088] For example, the lower the material unit price, the better, while the material demand quantity and material receipt quantity can be combined into one item, that is, delivery rate = material receipt quantity / material demand quantity. The higher the delivery rate, the better. Therefore, the reciprocal of the material unit price can be taken to make the delivery rate and the material unit price move in the same direction. Similarly, the lower the supply cycle, the better. This can be processed in the same direction. The larger the better, the better. No further processing is required.

[0089] After de-dimensionalizing the supplier evaluation indicators after homogenization, all supplier evaluation indicators are weighted and summed to obtain the supplier scoring rules.

[0090] De-dimensionalizing the supplier evaluation indicators after homogenization may include: normalizing any supplier evaluation indicator after homogenization to obtain the normalized supplier evaluation indicator. For example, normalizing the unit price of the material after homogenization can map this data to between [0,1], which can not only reduce the amount of data processing, but also make each supplier evaluation indicator in the same interval, which is convenient for summing. It is worth noting that in the process of weighted summation, the weight parameter corresponding to each supplier evaluation indicator is pre-set.

[0091] Optionally, users can also specify supplier evaluation indicators, and screen suppliers based on the specified supplier evaluation indicators and supplier scoring rules, thereby achieving customized recommendations and helping users quickly find suppliers that suit them.

[0092] In the embodiment of the present invention, a purchase invoice correction sub-rule, a sub-rule for cutting a first text area of ​​a purchase invoice, a sub-rule for identifying a first text area in the first text area, and a sub-rule for sorting the identified first text are constructed to obtain a purchase invoice recognition and verification rule, including:

[0093] Construct a rule for correcting the tilt of purchase invoices and obtain a purchase invoice correction sub-rule;

[0094] For example, OpenCV can be used to filter out the text in the purchase invoice, obtain the longest straight line in the invoice, and correct the longest straight line horizontally to achieve invoice correction.

[0095] The YOLO algorithm is used to construct the recognition rule of the first text area, and the sub-rule of cutting the first text area of ​​the purchase invoice is obtained;

[0096] For example, the historical invoices may be marked first to determine the text area, and then the historical invoices may be used as input data for the YOLO algorithm to obtain a calibration box for the predicted text area. Thus, the loss function corresponding to the YOLO algorithm may be determined based on the marked text area and the calibration box for the predicted text area. With the goal of minimizing the loss function, the particle swarm algorithm or other intelligent optimization algorithms may be used to adjust the algorithm parameters of the YOLO algorithm so that the YOLO algorithm can identify the first text area to obtain sub-rules for cutting the first text area of ​​the purchase invoice.

[0097] In order to improve the accuracy of parameter adjustment, an embodiment of the present invention provides a loss function to accurately determine whether the YOLO algorithm is adjusted. The loss function can be:

[0098]

[0099] Among them, L represents the loss function, λ cood Represents the weight coefficient between the predicted bounding box error and the category error, S 2 Indicates that the input image is divided into S×S cells, B represents the total number of borders within the cell, Indicates that the value is 1 only if the i-th cell contains a border, otherwise the value is 0; x i Represents the actual horizontal coordinate of the center of the jth border in the ith cell, represents the predicted horizontal coordinate of the center of the jth bounding box in the i-th cell, y ij It represents the actual vertical coordinate of the center of the jth border in the i-th cell. represents the predicted ordinate of the center of the jth bounding box in the i-th cell, w ij represents the actual width of the jth border in the ith cell, represents the predicted width of the jth border in the ith cell, h ij represents the actual height of the jth border in the ith cell, represents the predicted height of the jth border in the ith cell, It means that the value is 1 only when the intersection-over-intersection ratio between the j-th predicted box and the true box of the i-th cell is the largest, otherwise the value is 0; C ij represents the actual confidence of the jth bounding box in the i-th cell, represents the prediction confidence of the jth bounding box in the i-th cell, λ noobj Represents the weight coefficient to alleviate the problem of sample imbalance, p i (c) represents the probability of the actual category, Represents the probability of the predicted category, and classes represents the category set.

[0100] Optionally, an embodiment of the present invention provides a training method, which may include:

[0101] Training step 1: Initialize the algorithm parameters (such as connection weights) of the YOLO algorithm, construct a vector with all the initialized parameters, and repeatedly obtain multiple vectors;

[0102] Training step 2: According to the loss function, obtain the loss function value, and determine the vector with the smallest loss function value as the optimal vector;

[0103] Training step three: Perform a neighborhood search on the vector:

[0104]

[0105]

[0106] in, represents the i-th vector in the t-th iteration, i=1,2,…,I, I represents the total number of vectors, Represents the vector after local search Indicates the vector in the t+1th iteration process The corresponding increment, represents a random vector, and Not equal to R1 represents the first random number between (0,1), π represents the ratio of pi, and R2 represents the second random number between (0,1);

[0107] Training step 4: Perform guided search on the vector after neighborhood search:

[0108]

[0109]

[0110] in, Represents the vector after the mth neighborhood search during the t+1th iteration The corresponding increment, Represents the vector after the mth neighborhood search during the tth iteration The corresponding increments, α1 represents the first guiding coefficient, α2 represents the second guiding coefficient, and β represents the search range control coefficient. Representation vector The corresponding historical optimal value, d P Representation vector Its historical best value The Euclidean distance between t represents the optimal vector, that is, the vector with the smallest error function value; d G Representation vector The Euclidean distance from the optimal vector, e represents a natural constant.

[0111] Training step 5: Perform a global search on the vector after the guided search:

[0112]

[0113] in, represents the vector after the nth guided search, R3 represents the third random number between (0,1), R4 represents the fourth random number between (0,1), Represents the vector after global search γ represents the search control factor, levy represents the Levy flight step length, represents a random vector, and Not equal to

[0114] Training step six: When the number of training times does not reach the preset value, return to training step two; when the number of training times reaches the preset value, the optimal vector is output as the final parameter of the YOLO algorithm to complete the training process.

[0115] The training method provided by the embodiment of the present invention can not only realize the exploration of more solution spaces, but also has the characteristics of fast convergence speed and global optimal convergence, and can effectively improve the recognition ability of the algorithm.

[0116] The CDNets algorithm is used to construct a sub-rule for identifying the first character in the first character region;

[0117] The CDNets model uses the CTPN text detection algorithm, which consists of three parts: VGG-16 network architecture, BLSTM and fully connected layer. This text detection algorithm integrates spatial features and text time series features, which can well solve the above problems and achieve accurate text detection.

[0118] The CDNets algorithm can be used to detect four areas: the buyer area, the goods or taxable services area, the price and tax total area, and the seller area, thereby realizing the recognition of the first character.

[0119] Constructing a rule for formatting text processing of the first character, and obtaining a sub-rule for collating the recognized first character;

[0120] After the first character is recognized, a space may appear, but the character recognition automatically wraps to the next line. This is also a defect of OCR character recognition. Therefore, the embodiment of the present invention constructs a rule for formatting the first character to solve this problem.

[0121] Constructing a rule for formatting text processing of the first word may include:

[0122] Step 1: Create an empty first list. Store each line of character information recognized by CTPN and DenseNets in the first list one by one. In the next step, the subscript parameter information of the list can be used to operate on the recognized characters.

[0123] Step 2: Create an empty first string variable. Loop through the elements in the first list in step 1 and concatenate them.

[0124] Step 3: Since the characters in step 2 are stored in the UTF-8 encoding format, they need to be decoded and assigned to the new second string variable and converted into text information in the format of Chinese characters, letters, numbers, etc.

[0125] Step 4: Use Python's built-in function sub() and regular expressions to delete the name, taxpayer identification number, address, phone number, bank and account number key fields from the second string variable and assign them to the new third string variable.

[0126] Step 5: Use regular expressions to match Arabic numerals, uppercase English letters and '-' in the third string variable, extract the information corresponding to the taxpayer identification number, telephone number and account number key fields, store them in a new second list, and remove the empty values ​​in the list.

[0127] Step 6: While obtaining the values ​​in the second list through the enumerate() function, you can also get an additional serial number parameter. Since the element values ​​in the second list are different, and the values ​​in the second list are between the name, address, and bank, you can use the taxpayer identification number, phone number, and account number as segmentation elements, and segment them cyclically. The first time you use the taxpayer identification number to segment, the first item obtained is the information corresponding to the name. The second time you use the phone number to segment, the second item obtained is the information corresponding to the address. The third time you use the account number to segment, the third item obtained is the information corresponding to the bank, and the three items of information obtained are stored in a new third list.

[0128] Step 7: Concatenate the elements in the second list and the third list, sequentially concatenate the first item in the third list, the first item in the second list, and the line break operation; the second item in the third list, the second item in the second list, and the line break operation; the third item in the third list, the third item in the second list. At this point, all fields are integrated and the purpose of standard formatted text output is achieved, which can be saved to a file or database.

[0129] The purchase invoice recognition and verification rule is obtained according to the purchase invoice correction sub-rule, the sub-rule for cutting the first text area of ​​the purchase invoice, the sub-rule for identifying the first text in the first text area and the sub-rule for sorting the identified first text.

[0130] In the embodiment of the present invention, a procurement contract correction sub-rule, a sub-rule for cutting the second text area of ​​the procurement contract, a sub-rule for identifying the second text in the second text area, and a sub-rule for sorting the identified second text are constructed to obtain a procurement contract recognition and verification rule, including:

[0131] Construct a purchase contract tilt correction rule to obtain a purchase contract correction sub-rule;

[0132] The YOLO algorithm is used to construct the recognition rule of the second text area, and the sub-rule of cutting the second text area of ​​the procurement contract is obtained;

[0133] The CDNets algorithm is used to construct a sub-rule for identifying the second character in the second character area;

[0134] Constructing a rule for formatting text processing of the second character, and obtaining a sub-rule for collating the recognized second character;

[0135] According to the procurement contract correction sub-rule, the sub-rule for cutting the second text area of ​​the procurement contract, the sub-rule for identifying the second text in the second text area and the sub-rule for sorting the identified second text, the procurement contract recognition and verification rule is obtained.

[0136] In an embodiment of the present invention, a verification rule for the business license, tax registration certificate, organization code certificate, legal representative ID card and corporate bank account information in a payment application is constructed to obtain a payment application verification rule, including:

[0137] Construct text extraction and certificate number online query rules for business license, tax registration certificate, organization code certificate, legal representative ID card and corporate bank account information in payment applications to obtain payment application verification rules.

[0138] In an embodiment of the present invention, the demand keywords input by the user are obtained, and suppliers are recommended according to the supplier scoring rules and the demand keywords, so that the user can purchase according to the recommended suppliers, and generate a purchase invoice, a purchase contract and a payment application, including:

[0139] Obtain the demand keywords input by the user and determine the target suppliers with the demand keywords;

[0140] According to the supplier scoring rules, each target supplier is evaluated to determine the comprehensive score corresponding to the supplier evaluation index;

[0141] The top N target suppliers with the largest comprehensive scores corresponding to the supplier evaluation indicators are recommended to the user, so that the user can purchase according to the recommended suppliers and generate purchase invoices, purchase contracts and payment applications.

[0142] In an embodiment of the present invention, based on the purchase invoice identification and verification rules, AI analysis is performed on the purchase invoice to determine a first analysis result corresponding to the purchase invoice, including:

[0143] According to the purchase invoice correction sub-rule, the purchase invoice is corrected to obtain the corrected purchase invoice;

[0144] Using the corrected purchase invoice as an input of the recognition rule for the first text region to obtain the first text region in the purchase invoice;

[0145] Recognize the first character region using a sub-rule for recognizing the first character in the first character region to obtain the first character;

[0146] Arrange the first character using the sub-rule for arranging the recognized first character to obtain the arranged first character;

[0147] The sorted first text is checked against the bill content recorded in the financial management system to determine a first analysis result corresponding to the purchase invoice.

[0148] In the embodiment of the present invention, based on the procurement contract verification rule, AI analysis is performed on the procurement contract to determine the second analysis result corresponding to the procurement contract, including:

[0149] Correct the purchase contract according to the purchase contract correction sub-rule to obtain the corrected purchase contract;

[0150] Using the corrected purchase contract as an input of the recognition rule of the second text region to obtain the second text region in the purchase contract;

[0151] Recognize the second character region using a sub-rule for recognizing the second character in the second character region to obtain the second character;

[0152] Arrange the second character using the sub-rule for arranging the recognized second character to obtain an arranged second character;

[0153] The collated second text is checked against the contract content recorded in the financial management system to determine the second analysis result corresponding to the procurement contract.

[0154] In the embodiment of the present invention, based on the payment application verification rule, AI analysis is performed on the payment application to determine the third analysis result corresponding to the payment application, including:

[0155] Parse the text information of the business license, tax registration certificate, organization code certificate, legal representative ID card and corporate bank account information in the payment application to obtain key text information;

[0156] Based on the payment application verification rules, the key text information is queried online to determine the third analysis result corresponding to the payment application.

[0157] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0158] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0159] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0160] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0161] A person of ordinary skill in the art can understand that all or part of the steps in realizing the above-mentioned facts and methods can be completed by instructing the relevant hardware through a program, and the program involved or the program described can be stored in a computer-readable storage medium. When the program is executed, it includes the following steps: At this time, the corresponding method steps are derived, and the storage medium can be ROM / RAM, a disk, an optical disk, etc.

[0162] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An AI-based intelligent procurement method, characterized in that: include: Construct supplier scoring rules, purchase invoice identification and verification rules, purchase contract verification rules, and payment application verification rules; Obtaining the demand keywords input by the user, and recommending suppliers according to the supplier scoring rules and the demand keywords, so that the user can purchase from the recommended suppliers and generate purchase invoices, purchase contracts and payment applications; Based on the purchase invoice identification and verification rules, AI analysis is performed on the purchase invoice to determine a first analysis result corresponding to the purchase invoice; wherein the first analysis result includes the presence of an abnormality or the absence of an abnormality; Based on the procurement contract verification rule, AI analysis is performed on the procurement contract to determine a second analysis result corresponding to the procurement contract; wherein the second analysis result includes the presence of an abnormality or the absence of an abnormality; Based on the payment application verification rule, AI analysis is performed on the payment application to determine a third analysis result corresponding to the payment application; wherein the third analysis result includes whether an abnormality exists or does not exist; When the first analysis result, the second analysis result and / or the third analysis result are abnormal, it is determined that the verification has failed, and a manual review request is initiated to the server, and a warning is given; When the first analysis result, the second analysis result and the third analysis result show that there is no abnormality, the verification is determined to be successful, the procurement data is archived, and the intelligent procurement process is completed.

2. The AI-based intelligent procurement method according to claim 1, characterized in that: Build supplier scoring rules, purchase invoice identification and verification rules, purchase contract verification rules, and payment application verification rules, including: Determine supplier evaluation indicators, de-dimensionalize the supplier evaluation indicators, and then perform weighted summation on all supplier evaluation indicators to obtain supplier scoring rules; Constructing a purchase invoice correction sub-rule, a sub-rule for cutting the first text area of ​​the purchase invoice, a sub-rule for identifying the first text in the first text area, and a sub-rule for sorting the identified first text, to obtain a purchase invoice identification and verification rule; Constructing a procurement contract correction sub-rule, a sub-rule for cutting the second text area of ​​the procurement contract, a sub-rule for identifying the second text in the second text area, and a sub-rule for sorting the identified second text, to obtain a procurement contract identification and verification rule; Construct verification rules for the business license, tax registration certificate, organization code certificate, legal representative ID card, and corporate bank account information in the payment application to obtain the payment application verification rules.

3. The AI-based intelligent procurement method according to claim 2, characterized in that: After determining the supplier evaluation indicators and de-dimensionalizing the supplier evaluation indicators, all supplier evaluation indicators are weighted and summed to obtain supplier scoring rules, including: Determine supplier evaluation indicators; wherein the supplier evaluation indicators include material unit price, material demand quantity, material receipt quantity, material qualification rate, supply cycle, supplier purchase amount, supplier purchase times, supplier registered capital and supplier grade; wherein the supplier grade is the average of the historical scores given by the purchaser to the supplier; With respect to the supplier evaluation indicators, the supplier evaluation indicators are made homogenized to obtain the homogenized supplier evaluation indicators; After de-dimensionalizing the supplier evaluation indicators after homogenization, all supplier evaluation indicators are weighted and summed to obtain the supplier scoring rules.

4. The AI-based intelligent procurement method according to claim 2, characterized in that: Constructing a purchase invoice correction sub-rule, a sub-rule for cutting the first text area of ​​the purchase invoice, a sub-rule for identifying the first text in the first text area, and a sub-rule for sorting the identified first text, to obtain a purchase invoice recognition and verification rule, including: Construct a rule for correcting the tilt of purchase invoices and obtain a purchase invoice correction sub-rule; The YOLO algorithm is used to construct the recognition rule of the first text area, and the sub-rule of cutting the first text area of ​​the purchase invoice is obtained; The CDNets algorithm is used to construct a sub-rule for identifying the first character in the first character region; Constructing a rule for formatting text processing of the first character, and obtaining a sub-rule for collating the recognized first character; The purchase invoice recognition and verification rule is obtained according to the purchase invoice correction sub-rule, the sub-rule for cutting the first text area of ​​the purchase invoice, the sub-rule for identifying the first text in the first text area and the sub-rule for sorting the identified first text.

5. The AI-based intelligent procurement method according to claim 2, characterized in that: Constructing a procurement contract correction sub-rule, a sub-rule for cutting the second text area of ​​the procurement contract, a sub-rule for identifying the second text in the second text area, and a sub-rule for sorting the identified second text, to obtain a procurement contract identification and verification rule, including: Construct a purchase contract tilt correction rule to obtain a purchase contract correction sub-rule; The YOLO algorithm is used to construct the recognition rule of the second text area, and the sub-rule of cutting the second text area of ​​the procurement contract is obtained; The CDNets algorithm is used to construct a sub-rule for identifying the second character in the second character area; Constructing a rule for formatting text processing of the second character, and obtaining a sub-rule for collating the recognized second character; According to the procurement contract correction sub-rule, the sub-rule for cutting the second text area of ​​the procurement contract, the sub-rule for identifying the second text in the second text area and the sub-rule for sorting the identified second text, the procurement contract recognition and verification rule is obtained.

6. The AI-based intelligent procurement method according to claim 2, characterized in that: Construct verification rules for the business license, tax registration certificate, organization code certificate, legal representative ID card, and corporate bank account information in the payment application to obtain the payment application verification rules, including: Construct text extraction and online query rules for certificate numbers of business licenses, tax registration certificates, organization code certificates, legal representative ID cards, and corporate bank account information in payment applications to obtain payment application verification rules.

7. The AI-based intelligent procurement method according to claim 3, characterized in that: Obtain the demand keywords input by the user, and recommend suppliers according to the supplier scoring rules and demand keywords, so that the user can purchase from the recommended suppliers, generate purchase invoices, purchase contracts and payment applications, including: Obtain the demand keywords input by the user and determine the target suppliers with the demand keywords; According to the supplier scoring rules, each target supplier is evaluated to determine the comprehensive score corresponding to the supplier evaluation index; The top N target suppliers with the largest comprehensive scores corresponding to the supplier evaluation indicators are recommended to the user, so that the user can purchase according to the recommended suppliers and generate purchase invoices, purchase contracts and payment applications.

8. The AI-based intelligent procurement method according to claim 4, characterized in that: Based on the purchase invoice recognition and verification rules, AI analysis is performed on the purchase invoice to determine a first analysis result corresponding to the purchase invoice, including: According to the purchase invoice correction sub-rule, the purchase invoice is corrected to obtain the corrected purchase invoice; Using the corrected purchase invoice as an input of the recognition rule for the first text region to obtain the first text region in the purchase invoice; Recognize the first character region using a sub-rule for recognizing the first character in the first character region to obtain the first character; Arrange the first character using the sub-rule for arranging the recognized first character to obtain the arranged first character; The sorted first text is checked against the bill content recorded in the financial management system to determine a first analysis result corresponding to the purchase invoice.

9. The AI-based intelligent procurement method according to claim 5, characterized in that: Based on the procurement contract verification rules, AI analysis is performed on the procurement contract to determine the second analysis result corresponding to the procurement contract, including: Correct the purchase contract according to the purchase contract correction sub-rule to obtain the corrected purchase contract; Using the corrected purchase contract as an input of the recognition rule of the second text region to obtain the second text region in the purchase contract; Recognize the second character region using a sub-rule for recognizing the second character in the second character region to obtain the second character; Arrange the second character using the sub-rule for arranging the recognized second character to obtain an arranged second character; The collated second text is checked against the contract content recorded in the financial management system to determine the second analysis result corresponding to the procurement contract.

10. The AI-based intelligent procurement method according to claim 6, characterized in that: Based on the payment application verification rule, AI analysis is performed on the payment application to determine a third analysis result corresponding to the payment application, including: Parse the text information of the business license, tax registration certificate, organization code certificate, legal representative ID card and corporate bank account information in the payment application to obtain key text information; Based on the payment application verification rules, the key text information is queried online to determine the third analysis result corresponding to the payment application.

Images