Bidding document coincidence degree determination method, computer equipment and storage medium

Abstract

The invention relates to a bidding document coincidence degree determination method, computer equipment and a storage medium. The method comprises the following steps: carrying out feature extraction on obtained character information of a to-be-checked duplicate bidding document by adopting a character feature extraction model to obtain character feature information, and carrying out feature extraction on original image information of the to-be-checked duplicate bidding document by adopting an image feature extraction model to obtain image feature information, performing feature extraction on the enterprise information of the bidding document to be subjected to duplicate checking according to each preset enterprise attribute to obtain enterprise feature information; fusing the character feature information, the image feature information and the enterprise feature information to obtain bidding document feature information; and carrying out coincidence degree prediction on the bidding document feature information by adopting a bidding document coincidence degree prediction model to obtain target coincidence degree information of the bidding document to be subjected to duplicate checking. By adopting the method, the accuracy of the determined bidding document coincidence degree can be improved.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular to a method for determining the overlap of bids, a computer device and a storage medium. Background Art

[0002] In the bidding process, it is extremely important to ensure the fairness, impartiality and scientific nature of the bidding process. The bid document is an important carrier for bidders to demonstrate their strength, technical solutions, service capabilities, quotations and other key information. Therefore, the bid documents submitted by bidders are an important basis for evaluating their capabilities and solutions.

[0003] Bid rigging and bid collusion refer to the process of bidding, where bidders or bidders and the tenderer collude through improper means to exclude competitors or damage the interests of other bidders. Through the bid duplication check, it can be determined whether the bid submitted by the bidder is based on the bidder's own business and technical conditions, or whether it is a bid formed by plagiarism or imitation without the corresponding technology. Therefore, the bid duplication check can effectively discover whether there is bid rigging and bid collusion.

[0004] In the related art, the degree of overlap of bids is determined by checking for duplicate documents. However, the solution of determining the degree of overlap of bids by checking for duplicate documents has the problem of inaccurate degree of overlap of bids. Summary of the invention

[0005] Based on this, the present application provides a method for determining the degree of overlap of bids, a computer device and a storage medium, which can improve the accuracy of the determined degree of overlap of bids.

[0006] In a first aspect, the present application provides a method for determining the degree of overlap of bid documents, the method comprising: using a text feature extraction model to extract features from text information of the bid document to be checked for duplicates, to obtain text feature information; using an image feature extraction model to extract features from original image information of the bid document to be checked for duplicates, to obtain image feature information; and according to preset attributes of each enterprise, extracting features from enterprise information of the bid document to be checked for duplicates, to obtain enterprise feature information;

[0007] The text feature information, the image feature information and the enterprise feature information are integrated to obtain the bid feature information;

[0008] The bid overlap prediction model is used to predict the overlap of the bid feature information, and the target overlap information of the duplicate bids to be checked is obtained.

[0009] In some embodiments, a text feature extraction model is used to extract features from the text information of the duplicated documents to be checked, including:

[0010] Perform at least one of the following processes on the text information of the bid to be checked for duplicates: filtering stop words, segmenting words, and extracting stems, to obtain key information of the bid to be checked for duplicates;

[0011] The text feature extraction model is used to extract the key information of the bid to be checked for duplicates to obtain text feature information.

[0012] In some embodiments, a text feature extraction model is used to extract features of key information of the bid to be checked for duplicates, and text feature information is obtained, including:

[0013] Obtain a set of key words in the files in the file library; wherein the file library includes the files of the historical bidding and the files of the current bidding other than the bid documents to be checked for duplicates;

[0014] According to each target word in the key information that belongs to the key word set, a text feature extraction model is used to determine feature information of each target word, and a target value is determined as feature information of each designated word other than each target word in the key word set;

[0015] According to the order of words in the keyword set, the feature information of each target word and the feature information of each designated word are combined to obtain text feature information.

[0016] In some embodiments, the feature information of each target word is determined using a text feature extraction model, including:

[0017] According to the number of occurrences of each target word in the key information and the total number of words in the key information, the word frequency determination algorithm in the text feature extraction model is used to determine the word frequency of each target word in the key information;

[0018] According to the total number of files in the file library and the number of sub-files containing each target word in the file library, the inverse document frequency determination algorithm in the text feature extraction model is used to determine the inverse document frequency of each target word;

[0019] The feature information of each target word is determined according to the word frequency of each target word and the inverse document frequency of each target word.

[0020] In some embodiments, an image feature extraction model is used to extract features from the original image information of the document to be checked for duplicate content, and image feature information is obtained, including:

[0021] Performing histogram equalization and / or denoising processing on each original image information of the duplicated documents to be checked to obtain each converted image information;

[0022] Perform pixel size conversion processing on each converted image information to obtain each target image information of a preset pixel size;

[0023] The image feature extraction model is used to extract features of each target image information to obtain feature information of each image.

[0024] In some embodiments, feature extraction is performed on the enterprise information of the duplicate bid to be checked according to the preset attributes of each enterprise to obtain enterprise feature information, including:

[0025] Determine the enterprise similarity information between the attribute information of the enterprise information of the to-be-checked duplicate bid under each enterprise attribute and the attribute information of the enterprise information of each bid under each enterprise attribute in all bids collected in this bidding;

[0026] According to the enterprise similarity information, determine the attribute overlap of the enterprise information of the duplicate bid to be checked under each enterprise attribute;

[0027] The attribute overlaps under each enterprise attribute are combined to obtain enterprise characteristic information.

[0028] In some embodiments, a bid overlap prediction model is used to predict the overlap of bid feature information to obtain target overlap information of the duplicate bid to be checked, including:

[0029] The bid overlap prediction model is used to predict the overlap of the bid feature information, and the predetermined overlap information of the duplicate bids to be checked is obtained;

[0030] Determine the text overlap according to the text feature information, determine the image overlap according to the image feature information, and determine the enterprise overlap according to the enterprise feature information;

[0031] The text overlap, image overlap and enterprise overlap are integrated to obtain the set overlap information;

[0032] The predetermined overlap information and the set overlap information are merged to obtain the target overlap information.

[0033] In some embodiments, determining the text overlap according to the text feature information, determining the image overlap according to the image feature information, and determining the enterprise overlap according to the enterprise feature information includes:

[0034] Determine the text similarity information between the text feature information of the bid to be checked for duplicates and the text feature information of each file in the file library, and determine the text overlap degree based on the text similarity information; wherein the file library includes the files of the historical bidding and the files of the current bidding other than the bid to be checked for duplicates;

[0035] Determine the image similarity information between the image feature information corresponding to each original image in the to-be-checked duplicated book and the image feature information corresponding to each image in each file in the file library, and determine the image coincidence degree based on the image similarity information;

[0036] The attribute overlap of each enterprise attribute included in the enterprise characteristic information is weighted and fused to obtain the enterprise overlap.

[0037] In a second aspect, the present application provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any of the above methods when executing the computer program.

[0038] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of any of the above methods when the computer program is executed by a processor.

[0039] In the technical solution of the embodiment of the present application, by respectively extracting features of the text information, original image information and enterprise information of the bid to be checked for duplicates, feature extraction of multimodal data can be performed according to the characteristics of the bid, and the feature information of the multimodal data can be fused to obtain the bid feature information. The bid feature information can comprehensively reflect the characteristics of the bid, thereby improving the comprehensiveness and pertinence of the bid feature extraction. Therefore, using the feature information of multimodal data to predict the overlap information can improve the accuracy of the target overlap information of the determined bid to be checked for duplicates. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0041] Figure 1 A schematic flow chart of a method for determining the degree of overlap of bids provided in the first embodiment;

[0042] Figure 2 A schematic flow chart of a method for determining the degree of overlap of bids provided in the second embodiment;

[0043] Figure 3 A schematic flow chart of a method for determining the degree of overlap of bids provided in the third embodiment;

[0044] Figure 4 A schematic flow chart of a method for determining the degree of overlap of bids provided in the fourth embodiment;

[0045] Figure 5 A schematic flow chart of a method for determining the degree of overlap of bids provided in the fifth embodiment;

[0046] Figure 6 A schematic diagram of the structure of a computer device is provided for some embodiments. DETAILED DESCRIPTION

[0047] The following embodiments of the technical solution of the present application are described in detail in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present application, and are therefore only used as examples, and cannot be used to limit the scope of protection of the present application.

[0048] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" in the specification and claims of this application and the above-mentioned figure descriptions and any variations thereof are intended to cover non-exclusive inclusions.

[0049] In the description of the embodiments of the present application, the technical terms "first", "second", etc. are only used to distinguish different objects, and cannot be understood as indicating or implying relative importance or implicitly indicating the number, specific order or primary and secondary relationship of the indicated technical features. In the description of the embodiments of the present application, the meaning of "multiple" is more than two, unless otherwise clearly and specifically defined. In the description of the embodiments of the present application, "each" means each or each of the multiple, unless otherwise clearly and specifically defined.

[0050] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0051] In the description of the embodiments of the present application, the term "and / or" is only a description of the association relationship of the associated objects, indicating that there may be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0052] During the bid evaluation process, a quantitative indicator is needed to check for duplicate bids for collusion. The degree of overlap between bids is calculated by borrowing the calculation method of document overlap to achieve this.

[0053] However, since the application scenarios of bid plagiarism checking are different from those of document plagiarism checking (such as paper plagiarism checking), its focus is on text, pictures, legal person information, company qualifications, etc., so it is also necessary to design a specific formula for calculating the degree of overlap in bid plagiarism checking.

[0054] Document duplication checking (such as paper duplication checking) in related technologies uses the Jaccard coefficient method, cosine similarity method or other methods. Among them, the Jaccard coefficient method is one of the most commonly used methods for calculating document overlap. The Jaccard coefficient measures the similarity between two document sets by calculating the ratio of the intersection and union of the two document sets. Cosine similarity is another commonly used document similarity calculation method, which measures the similarity between documents by representing documents as vectors and calculating the cosine values ​​between these vectors. Other methods include similarity calculation based on term frequency-inverse document frequency (TF-IDF), similarity calculation based on latent semantic analysis (LSA), etc.

[0055] However, document duplication checking (such as paper duplication checking) in related technologies is based on semantic analysis and natural language processing. As long as the meaning of the text is the same, it will be found. However, bid duplication checking needs to check for overlaps and plagiarism of completely identical texts. For situations with the same meaning and similar grammar, they are not enough to constitute a contradiction in bid duplication checking. According to the traditional practice of document duplication checking (such as paper duplication checking), it will bring serious interference to the bid review. Since the application scenarios of bid duplication checking are completely different from those of document duplication checking (such as paper duplication checking), its focus is on text, pictures, legal person information, company qualifications, etc., and the focus of document duplication checking is not here at all. In addition, bid query requires the design of a specific bid duplication checking overlap calculation method. Usually, text duplication checking does not construct a text overlap calculation method.

[0056] The computer device in the embodiments of the present application may include one or a combination of at least two of the following: a server, a mobile phone, a tablet computer, a computer with a transceiver function, a PDA, a desktop computer, a personal digital assistant, a portable media player, a smart speaker, a navigation device, a smart watch, a smart glasses, a smart necklace and other wearable devices, a pedometer, a digital TV, a virtual reality (VR) device, an augmented reality (AR) device, a terminal in industrial control, a terminal in self-driving, a terminal in remote medical surgery, a terminal in smart grid, a terminal in transportation safety, a terminal in smart city, a terminal in smart home, a vehicle, a vehicle-mounted device, a vehicle-mounted module, and the like.

[0057] Figure 1 A flowchart of a method for determining the degree of overlap of bids provided in the first embodiment is shown in FIG. Figure 1 As shown, the method is applied to a computer device, and the method comprises the following steps:

[0058] S101, using a text feature extraction model to extract features from the text information of the duplicate bid to be checked, to obtain text feature information.

[0059] For example, the model in this article can be understood in the same way as a network.

[0060] In some embodiments, the text feature extraction model may include one of the following: a bag-of-words model, a word frequency-inverse file frequency model, a word embedding module, or a deep learning model. Exemplarily, the word embedding model may include a word to vector (Word to Vector, Word2Vec) model or a global word vector representation (Global Vectors for Word Representation, GloVe) model, etc. Exemplarily, the deep learning model may include a bidirectional encoder representation from transformers (BERT) model or a generative pre-trained transformer (GPT).

[0061] In some embodiments, the text information of the duplicated bid to be checked can be input into a text feature extraction model to obtain text feature information from the text feature extraction model. In other embodiments, the text information of the duplicated bid to be checked can be preprocessed to obtain key information of the duplicated bid to be checked, and the key information of the duplicated bid to be checked can be input into a text feature extraction model to obtain text feature information from the text feature extraction model.

[0062] In some embodiments, for different bids, the dimensions of the text feature information obtained are the same.

[0063] S102, using an image feature extraction model to extract features from the original image information of the duplicated document to be checked, to obtain image feature information.

[0064] In some embodiments, the image feature extraction model may include a convolutional neural network (CNN) or other image feature extraction networks.

[0065] In some embodiments, the original image information of the document to be checked for duplicates may be input into an image feature extraction model, and the image feature information may be obtained from the image feature extraction model. In other embodiments, the original image information of the document to be checked for duplicates may be preprocessed to obtain target image information, and the target image information may be input into an image feature extraction model, and the image feature information may be obtained from the image feature extraction model.

[0066] In some embodiments, for different bids, the dimensions of the acquired image feature information are the same.

[0067] S103, extracting features from the enterprise information of the duplicate bids to be checked according to the preset attributes of each enterprise, and obtaining enterprise feature information.

[0068] In some embodiments, at least two enterprise attributes may be pre-set. For example, the at least two enterprise attributes may include at least one of the following: enterprise name, legal person information, company qualification information, enterprise corresponding email information, enterprise contact information, etc.

[0069] Exemplarily, the enterprise characteristic information may include characteristic information under each enterprise attribute. Exemplarily, the enterprise characteristic information may include attribute overlap under each enterprise attribute.

[0070] In some embodiments, for different bids, the dimensions of the enterprise characteristic information obtained are the same.

[0071] In some embodiments, the steps of extracting features from text information, extracting features from original image information, and extracting features from corporate information may be performed simultaneously.

[0072] S104: Fusing the text feature information, the image feature information and the enterprise feature information to obtain the bid feature information.

[0073] In some embodiments, the text feature information, image feature information and enterprise feature information are integrated to obtain the bid feature information, which can be achieved by splicing the text feature information, image feature information and enterprise feature information to obtain the bid feature information.

[0074] In other embodiments, the text feature information, image feature information and enterprise feature information are integrated to obtain the bid feature information, which can be achieved in the following ways: dimensionally adjust the text feature information, image feature information and enterprise feature information to obtain text feature adjustment information of a preset dimension, image feature adjustment information of a preset dimension and enterprise feature adjustment information of a preset dimension, and weightedly integrate the text feature adjustment information, image feature adjustment information and enterprise feature adjustment information to obtain the bid feature information.

[0075] S105, using a bid overlap prediction model to predict the overlap of the bid feature information, and obtaining target overlap information of the duplicate bids to be checked.

[0076] In some embodiments, the bid overlap prediction model may include a back-propagation (BP) neural network.

[0077] In some embodiments, the bid overlap prediction model may be a trained prediction model. In some embodiments, the bid overlap prediction model may be trained in the following manner: obtaining a sample (also referred to as a sample bid), obtaining the bid feature information of the sample using the same steps as S101 and S102, obtaining the overlap information of each sample determined by an expert, inputting the bid feature information of each sample and the overlap information of each sample into the bid overlap prediction model to be trained, thereby obtaining the trained bid overlap prediction model.

[0078] In the technical solution of the embodiment of the present application, by respectively extracting features of the text information, original image information and enterprise information of the bid to be checked for duplicates, feature extraction of multimodal data can be performed according to the characteristics of the bid, and the feature information of the multimodal data can be fused to obtain the bid feature information. The bid feature information can comprehensively reflect the characteristics of the bid, thereby improving the comprehensiveness and pertinence of the bid feature extraction. Therefore, using the feature information of multimodal data to predict the overlap information can improve the accuracy of the target overlap information of the determined bid to be checked for duplicates.

[0079] Figure 2 A flow chart of a method for determining the degree of overlap of bids provided in the second embodiment is shown in FIG. Figure 2 As shown, the method is applied to a computer device, Figure 2 Example compared to Figure 1 The difference between the embodiments is that S101 includes S1011 and S1012:

[0080] S1011, performing at least one of filtering stop words, word segmentation, and stem extraction on the text information of the bid to be checked for duplicates, and obtaining key information of the bid to be checked for duplicates.

[0081] Stop word filtering refers to the process of removing common words that have little practical meaning or have little impact on the analysis results in the process of natural language processing (NLP). Stop words are common words that do not need to be considered in text analysis. They appear frequently in the text but do not provide substantial help for text analysis, only occupying storage space and time. For example, in English, common stop words include "the", "a", "an", etc.; in Chinese, common stop words include "的", "是", "和", etc.

[0082] Word segmentation is a basic task in natural language processing (NLP), especially important in Chinese processing. Its main purpose is to split a continuous text sequence into meaningful words or other language units. Since Chinese writing does not have obvious space separators between words like English, word segmentation technology is needed to determine the lexical boundaries of sentences.

[0083] Stemming is a rule-based text processing method. Exemplarily, it extracts the stem of a word by removing the suffix of the word. Its purpose is to transform a word into its basic linguistic form, that is, the stem, without considering the grammar and semantics of the word. For example, it transforms "running", "runs", and "ran" into the stem "run".

[0084] S1012. Use a text feature extraction model to extract features from the key information of the bid to be checked for duplication, and obtain text feature information.

[0085] In some embodiments, using a text feature extraction model to extract features from the key information of the bid to be checked for duplication to obtain text feature information includes: obtaining a set of keyword phrases of the files in the file library; wherein, the file library includes files in historical bidding and tendering and files other than the bid to be checked for duplication in the current bidding and tendering; according to each target word in the key information that belongs to the set of keyword phrases, using the text feature extraction model to determine the feature information of each target word, and determining the target value as the feature information of each specified word other than each target word in the set of keyword phrases; combining the feature information of each target word and the feature information of each specified word in the order of the words in the set of keyword phrases to obtain text feature information.

[0086] Exemplarily, the files in historical bidding and tendering can include all files collected during historical bidding and tendering. For example, the files in historical bidding and tendering can include at least one of the following: bids, tender invitations, reference files, project reports, papers, etc. in historical bidding and tendering processes.

[0087] Exemplarily, the documents other than the bid documents to be checked for duplicates in the current bidding documents may include the documents other than the bid documents to be checked for duplicates in all the documents collected during the current bidding process. For example, the documents other than the bid documents to be checked for duplicates in the current bidding documents may include at least one of the following: bidding documents, reference documents, project reports, papers, bid documents other than the bid documents to be checked for duplicates in the current bidding process, etc.

[0088] In some embodiments, at least one of filtering stop words, word segmentation, and stem extraction may be performed on the files in the file library to obtain a set of key words for the files in the file library.

[0089] For example, the target value may be 0, 1 or other numerical values.

[0090] By combining the feature information of each target word and the feature information of each designated word according to the order of words in the keyword set, the text feature information is obtained, so that the feature information of the same dimension in the text feature information corresponding to different bids all represents the features of the same word. For example, the text feature information of bid A, bid B, and bid C are all N-dimensional feature information (N is an integer greater than or equal to 1), and the i-th dimension feature information of bid A, bid B, and bid C all represents the features of the same word (i is an integer greater than or equal to 1 and less than or equal to N).

[0091] In some embodiments, a text feature extraction model is used to determine feature information of each target word, including: determining the word frequency of each target word in the key information based on the number of occurrences of each target word in the key information and the total number of words in the key information using a word frequency determination algorithm in the text feature extraction model; determining the inverse document frequency of each target word based on the total number of files in the file library and the number of sub-files containing each target word in the file library using an inverse document frequency determination algorithm in the text feature extraction model; determining the feature information of each target word based on the word frequency of each target word and the inverse document frequency of each target word.

[0092] Exemplarily, the word frequency of each target word and the inverse document frequency of each target word may be correspondingly multiplied to obtain feature information of each target word.

[0093] Figure 3 A flow chart of a method for determining the degree of overlap of bids provided in the third embodiment is shown in FIG. Figure 3 As shown, the method is applied to a computer device, Figure 3 Example compared to Figure 1 The difference between the embodiments is that S102 includes S1021 to S1023:

[0094] S1021, performing histogram equalization and / or denoising processing on each original image information of the document to be checked for duplicate marking to obtain each converted image information.

[0095] Image histogram equalization is a classic image enhancement method that aims to improve the contrast of an image and make the brightness distribution of the image more uniform. This method improves the global contrast of an image by adjusting the grayscale distribution of the image, so that both the dark and bright areas of the image can be better displayed.

[0096] Image denoising is an important step to improve image quality. It mainly improves the clarity and visual effect of the image by removing random noise in the image. Image noise can be divided into additive noise and multiplicative noise. The former is irrelevant to the image signal, while the latter is related to the signal. Common denoising methods include filter-based methods, model-based methods, and learning-based methods.

[0097] S1022: Perform pixel size conversion processing on each converted image information to obtain each target image information of a preset pixel size.

[0098] S1023. Use an image feature extraction model to extract features from each target image information to obtain feature information of each image.

[0099] By performing pixel size conversion processing on each converted image information, each target image information of a preset pixel size is obtained, thereby obtaining each target image information with the same pixel size, and then using an image feature extraction model to extract features from each target image information to obtain each image feature information, which can ensure that the dimensions of the image feature information corresponding to each original image information obtained are the same.

[0100] Exemplarily, for a target image information, using an image feature extraction model to perform feature extraction on the target image information to obtain image feature information corresponding to the original image information can include the following steps: using the convolution layer and the pooling layer in the image feature extraction model to perform feature extraction on the target image information to obtain a feature map corresponding to the original image information, and flattening the feature map to obtain image feature information corresponding to the original image information.

[0101] Figure 4 A flowchart of a method for determining the degree of overlap of bids provided in the fourth embodiment is shown in FIG. Figure 4 As shown, the method is applied to a computer device, Figure 4 Example compared to Figure 1 The difference between the embodiments is that S103 includes S1031 to S1033:

[0102] S1031. Determine the enterprise similarity information between the attribute information of the enterprise information of the to-be-checked duplicate bid under each enterprise attribute and the attribute information of the enterprise information of each bid under each enterprise attribute in all bids collected in this bidding.

[0103] In some embodiments, the attribute information under each enterprise attribute of the bid to be checked for duplicates can be determined in the following manner: the text information under each enterprise attribute of the bid to be checked for duplicates is compared for identity with the text information under each enterprise attribute of the enterprise information in each bid among all the bids collected in this bidding; if the text information under a certain enterprise attribute of the bid to be checked for duplicates is the same as the text information under the enterprise attribute in any of all the bids collected in this bidding, the attribute information under the enterprise attribute of the bid to be checked for duplicates is determined to be a first value (for example, 1); otherwise, the attribute information under the enterprise attribute of the bid to be checked for duplicates is determined to be a second value (for example, 0).

[0104] S1032. Determine the degree of overlap of the enterprise information of the duplicate bids to be checked under the attributes of each enterprise based on the enterprise similarity information.

[0105] In some embodiments, the maximum value among multiple enterprise similarity information of the duplicate bid to be checked under each enterprise attribute can be determined as the attribute coincidence degree of the duplicate bid to be checked under each enterprise attribute.

[0106] S1033. Combine the attribute overlaps of each enterprise attribute to obtain enterprise characteristic information.

[0107] In some embodiments, the attribute overlaps under each enterprise attribute may be combined according to the order of multiple enterprise attributes to obtain enterprise characteristic information.

[0108] Figure 5 A flowchart of a method for determining the degree of overlap of bids provided in the fifth embodiment is shown in FIG. Figure 5 As shown, the method is applied to a computer device, Figure 5 Example compared to Figure 1 The difference between the embodiments is that S105 includes S1051 to S1054:

[0109] S1051. Use a bid overlap prediction model to predict the overlap of bid feature information, and obtain predetermined overlap information of the duplicate bids to be checked.

[0110] S1052. Determine the text overlap according to the text feature information, determine the image overlap according to the image feature information, and determine the enterprise overlap according to the enterprise feature information.

[0111] In some embodiments, the text overlap degree can be determined by the following method: determining the text similarity information between the text feature information of the bid to be checked for duplicates and the text feature information of each file in the file library, and determining the text overlap degree based on the text similarity information; wherein the file library includes files from historical bidding and files other than the bid to be checked for duplicates in the files of this bidding.

[0112] In some embodiments, the image coincidence degree can be determined by: determining the image feature information corresponding to each original image in the to-be-checked re-marked document, and the image similarity information between the image feature information corresponding to each image in each file in the file library, and determining the image coincidence degree based on the image similarity information.

[0113] In some embodiments, the enterprise overlap degree may be determined by performing weighted fusion on the attribute overlap degrees of various enterprise attributes included in the enterprise characteristic information to obtain the enterprise overlap degree.

[0114] S1053, integrating the text overlap, image overlap and enterprise overlap to obtain set overlap information.

[0115] In some embodiments, the text overlap, image overlap and enterprise overlap can be weighted and fused to obtain the set overlap information. In other embodiments, the maximum value among the text overlap, image overlap and enterprise overlap can be determined as the set overlap information.

[0116] S1054: The predetermined overlap information and the set overlap information are merged to obtain target overlap information.

[0117] In some embodiments, the predetermined overlap information and the set overlap information may be weighted and fused to obtain the target overlap information. In other embodiments, the maximum value of the predetermined overlap information and the set overlap information may be determined as the target overlap information.

[0118] The embodiment of the present application provides a method for calculating the overlap of bids that integrates multiple modes. The method for calculating the overlap of bids that integrates multiple modes is a complex and meticulous process. It combines multiple data types (such as text, pictures, etc.) to comprehensively evaluate the similarity or overlap of bids. Through feature fusion, data of different modes are respectively input into their respective sub-networks for feature extraction, and then the extracted features are fused to complement each other. By using artificial neural networks to perform multi-modal feature fusion, and on the basis of normalizing feature vectors (i.e., feature information), a BP neural network is constructed, which includes an input layer, two hidden layers, and an output layer, and the BP neural network is trained with collected sample data. Through quantification of output results: the BP neural network is obtained after training, and after extracting features from new samples, forward calculation is performed from the input layer to the output layer, and the similarity or overlap of the bids is directly obtained by calculating the activation function results of the output nodes. This quantified result ensures the comparability of the calculation results of the overlap of bids. In addition, this method can maintain the independence of each modality data while utilizing the complementarity between different modalities. At the same time, as the user usage increases, more bid samples are collected, and the trained artificial neural network will calculate the bid overlap more accurately.

[0119] In the process of preprocessing the text data of the bid document, the text content in the bid document is subjected to preprocessing steps such as word segmentation, stop word removal, and stem extraction to improve the accuracy of subsequent feature extraction. Exemplarily, industry standard, conventional, and effective natural language processing (NLP) toolkits can be used to assist in completing these preprocessing tasks. Exemplarily, the stop words can be filtered using the natural language processing toolkit (nltk) library in Python. Nltk already contains some commonly used stop word lists, which can be used directly. Exemplarily, the jieba library can be used to segment the text. Exemplarily, in stem extraction, affixes are removed from words so that the stems can be identified as variants of the same word.

[0120] In the preprocessing of the bidding document image data, the images in the bidding document may be subjected to preprocessing steps such as histogram equalization and denoising to ensure the consistency and comparability of the image data. Exemplarily, denoising may be performed using a Gaussian filter denoising method.

[0121] In text feature extraction, the bag-of-words model, TF-IDF, word embedding (such as Word2Vec, GloVe, etc.) or deep learning models (such as BERT, GPT, etc.) can be used to extract text features. These features can represent the key information, themes or semantic content in the bid.

[0122] In image feature extraction, deep learning models such as convolutional neural networks (CNNs) can be used to extract image features. These features can represent visual information such as texture, shape, and color in the image.

[0123] In the processing of legal person information, company qualifications, etc., a scoring method can be used to construct feature vectors of legal person information, company qualifications, etc. The scoring method is to score according to the similarity of each company's attributes.

[0124] By adopting the method of parallel feature fusion, the same comparable feature vector is constructed. The extracted text features, image features and file information are fused in parallel, that is, they are spliced ​​into a new feature vector. This fusion method can retain the independence of each modality data while utilizing their complementarity.

[0125] In the process of training the overlap neural network (i.e. the above-mentioned bid overlap prediction model), for the collected bids, the feature vector representing the bid is obtained through the above-mentioned operation steps. Combined with the evaluation expert scoring results corresponding to each bid (the scoring results are normalized to the interval [0,1]), a set of training data is formed. All collected bids constitute a training set and are sent to the overlap BP neural network for training. After the training is completed, the BP neural network is obtained. All nodes of the network and the connection weights of the nodes constitute the complete calculation formula for the bid overlap.

[0126] For the duplicate bids to be checked (also called bids to be processed), the feature vector representing the bid is obtained through the above-mentioned operation steps. After the features are extracted from the bid sample, forward calculation is performed from the input layer to the output layer of the BP neural network. At the nodes of the output layer, an activation function is applied to obtain a quantified result, which is the similarity or overlap of the bids. This quantified result ensures the comparability of the calculation results of the overlap of the bids. Exemplarily, a continuous and differentiable Sigmoid activation function can be used. Due to its monotonically increasing and inverse monotonically increasing properties, the Sigmoid function is often used as a threshold function of a neural network to map variables between 0 and 1. Among them, the activation function formula is .

[0127] In post-processing and actual application, a suitable threshold can be set according to actual needs to determine whether two bids are overlapped or similar. For example, the setting of the threshold can be adjusted according to factors such as specific application scenarios, data distribution, and calculation accuracy.

[0128] In the result interpretation, the calculated overlap results can be interpreted and analyzed to understand the similarities and differences between the bids. For example, the significance and value of the overlap results can be deeply interpreted in combination with knowledge and background information in a specific field.

[0129] In summary, a method for calculating the overlap of bids that integrates multiple modalities can be developed by comprehensively utilizing multiple steps, including data preprocessing, feature extraction, feature fusion, overlap calculation based on BP network, and post-processing.

[0130] In summary, the method for calculating the overlap of bids integrating multiple modalities in the technical solution provided in the embodiment of the present application is completed by comprehensively utilizing multiple steps such as data preprocessing, feature extraction, feature fusion, overlap calculation, post-processing and interpretation, thus overcoming the technical problems of traditional methods.

[0131] In practical applications, appropriate methods and strategies can be selected according to factors such as specific data types, task requirements, and computing resources, and practical applications have been obtained in the industry. In the technical solution provided in the embodiment of the present application, the technical defects of existing document duplicate checking (such as paper duplicate checking) are overcome, which are all based on semantic analysis and natural language processing, and accurate duplicate checking for bids can be achieved. In the technical solution provided in the embodiment of the present application, for the text, pictures, legal person information, company qualifications and other aspects that need to be focused on, special preprocessing and feature fusion devices can be constructed to achieve. In the technical solution provided in the embodiment of the present application, according to the design idea of ​​parallel fusion, the independence of features is guaranteed and complementary feature vectors are constructed. In the technical solution provided in the embodiment of the present application, a BP neural network is used. As the user usage increases, the more bid samples are collected, and the trained artificial neural network, the more accurate the bid overlap is calculated. In the technical solution provided in the embodiment of the present application, the quantitative bid duplicate check overlap calculation formula not only meets the specific purpose of bid duplicate checking, but also ensures the horizontal comparability and vertical comparability of data.

[0132] In an exemplary embodiment, Figure 6A schematic diagram of the structure of a computer device provided for some embodiments, the computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through wireless fidelity (Wireless Fidelity, WIFI), a mobile cellular network, near field communication (Near Field Communication, NFC) or other technologies. When the computer program is executed by the processor, a method for determining the overlap of bid documents is implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0133] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0134] For example, a computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the method of any of the above embodiments when executing the computer program.

[0135] For example, in an exemplary embodiment, the processor is used to execute a computer program to implement: using a text feature extraction model to perform feature extraction on the acquired text information of the duplicate bid to be checked to obtain text feature information, using an image feature extraction model to perform feature extraction on the original image information of the duplicate bid to be checked to obtain image feature information, and performing feature extraction on the enterprise information of the duplicate bid to be checked according to preset enterprise attributes to obtain enterprise feature information; fusing the text feature information, image feature information and enterprise feature information to obtain bid feature information; using a bid overlap prediction model to predict the overlap of the bid feature information to obtain target overlap information of the duplicate bid to be checked.

[0136] In one embodiment, a computer-readable storage medium is provided, and when a computer program is executed by a processor, the steps of the method provided in any of the above embodiments are implemented.

[0137] For example, in an exemplary embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented: using a text feature extraction model to perform feature extraction on the acquired text information of the duplicate bid to be checked to obtain text feature information, using an image feature extraction model to perform feature extraction on the original image information of the duplicate bid to be checked to obtain image feature information, and performing feature extraction on the enterprise information of the duplicate bid to be checked according to preset enterprise attributes to obtain enterprise feature information; fusing the text feature information, image feature information and enterprise feature information to obtain bid feature information; using a bid overlap prediction model to predict the overlap of the bid feature information to obtain target overlap information of the duplicate bid to be checked.

[0138] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing related hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods.

[0139] The processor, each functional module or each functional unit in any embodiment of the present application may include any one or more of the following integrations: general-purpose processor, application-specific integrated circuit (ASIC), digital signal processor (DSP), digital signal processing device (DSPD), programmable logic device (PLD), field programmable gate array (FPGA), central processing unit (CPU), graphics processing unit (GPU), embedded neural network processor (neural-network processing units, NPU), controller, microcontroller, microprocessor, programmable logic device, discrete gate or transistor logic device, discrete hardware component, data processing logic based on quantum computing, artificial intelligence (AI) processor, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0140] The memory or computer-readable storage medium in any embodiment of the present application may include at least one of a non-volatile memory and a volatile memory. Non-volatile memory includes the integration of one or more of the following: Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Ferromagnetic Random Access Memory (FRAM), Flash Memory, Magnetic Surface Storage, Optical Disc, Compact Disc Read-Only Memory (CD-ROM), Tape, Floppy Disk, Flash Memory, Optical Storage, High-density Embedded Non-volatile Memory, Resistive Random Access Memory (ReRAM), Magnetoresistive Random Access Memory (MRAM), Ferroelectric Random Access Memory (FRAM), Phase Change Memory (PCM), Graphene Memory, Volatile Memory, etc. Volatile memory includes an integration of one or more of the following: Random Access Memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0141] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0142] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for determining the degree of overlap of bid documents, characterized in that: The method comprises: Using a text feature extraction model to extract features from the acquired text information of the duplicate bid to be checked, to obtain text feature information; using an image feature extraction model to extract features from the original image information of the duplicate bid to be checked, to obtain image feature information; and according to preset attributes of each enterprise, extracting features from the enterprise information of the duplicate bid to be checked, to obtain enterprise feature information; Merging the text feature information, the image feature information and the enterprise feature information to obtain the bid feature information; The bid overlap prediction model is used to predict the overlap of the bid feature information to obtain the target overlap information of the duplicate bid to be checked.

2. The method according to claim 1, characterized in that The method of using a text feature extraction model to extract features from the text information of the duplicate bid to be checked includes: Perform at least one of filtering stop words, word segmentation, and stem extraction on the text information of the to-be-checked duplicate bid to obtain key information of the to-be-checked duplicate bid; The text feature extraction model is used to extract the key information of the duplicate bid to be checked to obtain the text feature information.

3. The method according to claim 2, characterized in that The method of using the text feature extraction model to extract features of key information of the to-be-checked duplicate bid to obtain the text feature information includes: Obtain a key word set of files in a file library; wherein the file library includes files of historical bidding and files other than the bid to be checked for duplicates in the current bidding; According to each target word in the key information belonging to the key word set, the text feature extraction model is used to determine the feature information of each target word, and the target value is determined as the feature information of each designated word other than each target word in the key word set; The feature information of each of the target words and the feature information of each of the designated words are combined according to the order of the words in the keyword set to obtain the text feature information.

4. The method according to claim 3, characterized in that The step of using the text feature extraction model to determine the feature information of each target word includes: According to the number of occurrences of each of the target words in the key information and the total number of words in the key information, the word frequency determination algorithm in the text feature extraction model is used to determine the word frequency of each of the target words in the key information; According to the total number of files in the file library and the number of sub-files in the file library containing each target word, the inverse document frequency determination algorithm in the text feature extraction model is used to determine the inverse document frequency of each target word; The feature information of each target word is determined according to the word frequency of each target word and the inverse document frequency of each target word.

5. The method according to claim 1, characterized in that The image feature extraction model is used to extract features from the original image information of the duplicated document to be checked to obtain image feature information, including: Performing histogram equalization and / or denoising processing on each of the original image information of the duplicated documents to be checked to obtain each converted image information; Performing pixel size conversion processing on each of the converted image information to obtain each target image information of a preset pixel size; The image feature extraction model is used to extract features from each target image information to obtain each image feature information.

6. The method according to claim 1, characterized in that The feature extraction of the enterprise information of the to-be-checked duplicate bid according to the preset attributes of each enterprise to obtain the enterprise feature information includes: Determine the enterprise similarity information between the attribute information of the enterprise information of the to-be-checked duplicate bid under each of the enterprise attributes and the attribute information of the enterprise information of each of the bids collected in this bidding under each of the enterprise attributes; Determine the degree of overlap of the enterprise information of the to-be-checked duplicate bid under each of the enterprise attributes according to the enterprise similarity information; The attribute overlap degrees under each of the enterprise attributes are combined to obtain the enterprise characteristic information.

7. The method according to any one of claims 1 to 6, characterized in that: The method of using a bid overlap prediction model to predict the overlap of the bid feature information to obtain target overlap information of the duplicate bid to be checked includes: Using a bid overlap prediction model to predict the overlap of the bid feature information, and obtaining predetermined overlap information of the duplicate bid to be checked; Determine the text overlap according to the text feature information, determine the image overlap according to the image feature information, and determine the enterprise overlap according to the enterprise feature information; The text overlap, the image overlap and the enterprise overlap are integrated to obtain set overlap information; The predetermined overlap information and the set overlap information are merged to obtain the target overlap information.

8. The method according to claim 7, characterized in that The determining of the text overlap according to the text feature information, the determining of the image overlap according to the image feature information, and the determining of the enterprise overlap according to the enterprise feature information includes: Determine the text similarity information between the text feature information of the to-be-checked duplicate bid and the text feature information of each file in the file library, and determine the text coincidence degree according to the text similarity information; wherein the file library includes the files of the historical bidding and the files of the current bidding other than the to-be-checked duplicate bid; Determine image similarity information between image feature information corresponding to each of the original images in the to-be-checked duplicated document and image feature information corresponding to each of the images in each of the files in the file library, and determine the image coincidence degree based on the image similarity information; The attribute overlap degrees under the enterprise attributes included in the enterprise characteristic information are weightedly fused to obtain the enterprise overlap degree.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images