Tender file intelligent review system fusing quantitative scoring and image recognition

Abstract

The application relates to the technical field of data analysis, in particular to a bid file intelligent evaluation system fusing quantitative scoring and image recognition, which comprises the following: a data acquisition unit used for acquiring hierarchical information corresponding to target qualification information corresponding to each bid file; a search analysis unit used for determining a regional division mode according to a judgment condition, and determining a regional type according to an associated hierarchical similarity and a linkage difference coefficient; an information acquisition unit used for determining a hierarchical state according to a linkage reference value and a linkage radiation coefficient, and determining a qualification search mode according to the hierarchical state to acquire public qualification information; an optimization unit used for determining an optimization mode according to an analysis condition; and an evaluation unit used for determining a file type of a bid file to be evaluated according to an associated threshold value and a similar reference value, and determining a file scoring mode according to the file type; the application can improve the accuracy of bid file evaluation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data analysis, and particularly relates to a bid file intelligent evaluation system fusing quantitative scoring and image recognition. BACKGROUND

[0002] A kind of AI bid automatic scoring method disclosed in Chinese patent publication No.CN114492368A includes: establishing the feature extraction model of natural language processing;Uploading tender documents, obtaining tender details information of tender documents, and using the feature extraction model of natural language processing, extracting specific tender requirements in tender details information;Specific tender requirements extracted after the feature extraction model of natural language processing is marked with sequence number;Establish a sub-database of the basic situation of multiple companies, and mark each sub-data in the database according to the order of Chinese pinyin;Establish a scoring model;Set the screening condition, filter out the relevant companies that meet the conditions, and match the sub-database of the basic situation of relevant companies with the specific tender requirement data marked with specified label, and call the scoring model to score the tender documents.It can be seen from the above technical solution that the following problems exist: only based on the content of the file, without considering the web page information related to the content of the bid file, resulting in the problem of deviation of the accuracy of file evaluation. SUMMARY

[0003] Therefore, the present application provides a bid file intelligent evaluation system fusing quantitative scoring and image recognition to overcome the problem that the web page information related to the content of the bid file is not considered in the prior art, resulting in deviation of the accuracy of file evaluation.

[0004] To achieve the above-mentioned purpose, the present application provides a bid file intelligent evaluation system fusing quantitative scoring and image recognition, which includes:

[0005] A data acquisition unit is used to obtain the hierarchical information of the target qualification information corresponding to each bid file;

[0006] A search analysis unit is connected to the data acquisition unit, used to determine the regional division method according to the judgment condition, and determine the region type according to the correlation hierarchical similarity and the linkage difference coefficient, the regional division method is uniform division according to the comprehensive evaluation value or correlation division according to the hierarchical threshold;

[0007] An information acquisition unit is connected to the search analysis unit, used to determine the hierarchical state according to the linkage reference value and the linkage radiation coefficient, and determine the qualification search method according to the hierarchical state to obtain public qualification information, the qualification search method is to determine the optimization method according to the proportion of one type of region and the distribution coefficient of one type of region, or to determine the search priority coefficient of the link according to the relevance of link information;

[0008] an optimization unit, connected with the information acquisition unit, configured to determine an optimization mode according to an analysis condition, the optimization mode being to determine a search area according to a link similarity and a region correlation degree, or to determine a region search duration according to a region threshold;

[0009] an evaluation unit, connected with the data acquisition unit, the information acquisition unit, and the optimization unit respectively, configured to mark a bid document corresponding to target qualification information with a qualification difference degree less than a preset qualification difference degree as an evaluated bid document, to determine a file type of the evaluated bid document according to an association threshold and a similarity reference value, and to determine a file evaluation coefficient determination mode according to the file type, the file evaluation coefficient determination mode being to determine a file evaluation coefficient according to an index fluctuation value and a bid classification value or according to the bid classification value.

[0010] Further, the search analysis unit responds to a judgment condition that a hierarchical quantity reference value is greater than or equal to a preset hierarchical quantity reference value or a link reference value is less than a preset link reference value, and the region division mode is uniform division according to a comprehensive evaluation value.

[0011] The search analysis unit responds to a judgment condition that a hierarchical quantity reference value is less than a preset hierarchical quantity reference value and a link reference value is greater than or equal to a preset link reference value, and the region division mode is associated division according to a hierarchical threshold.

[0012] Further, the search analysis unit performs associated division according to a hierarchical threshold, including:

[0013] performing associated analysis on each link in the target hierarchy in a preset order, and when performing associated analysis on a single link, marking the link as a target link, and marking links outside the target link and not recorded in a sub-region as reference links;

[0014] adding the hierarchical threshold corresponding to the target link and the hierarchical threshold corresponding to each link after the preset order one by one until the total value of the hierarchical thresholds is greater than the total value of the preset hierarchical thresholds, marking each reference link before the preset order of the last hierarchical threshold in the cumulative addition and the target link as an associated link, and marking the smallest rectangular region that can contain the associated link as a sub-region;

[0015] and continuing to perform associated analysis on the links not recorded in the sub-region until each link is recorded in the sub-region;

[0016] The preset order is an order of the associated distances corresponding to the links from small to large.

[0017] Further, the search analysis unit determines a region type according to an associated hierarchical similarity and a link difference coefficient, and the region type includes:

[0018] The one-type region has a correlation level similarity less than a preset correlation level similarity or a link difference coefficient greater than or equal to a preset link difference coefficient;

[0019] The two-type region has a correlation level similarity greater than or equal to a preset correlation level similarity and a link difference coefficient less than a preset link difference coefficient.

[0020] Further, the correlation level similarity is determined according to a link information degree;

[0021] If the link information degree is greater than or equal to a preset link information degree, the correlation level similarity is determined according to a number of same keywords, and the correlation level similarity has a positive correlation with the number of same keywords;

[0022] If the link information degree is less than a preset link information degree, the correlation level similarity is determined according to a number of associated keywords, and the correlation level similarity has a positive correlation with the number of associated keywords.

[0023] Further, the information acquisition unit determines a level state according to a link reference value and a link radiation coefficient, and the level state includes:

[0024] The first level state has the link reference value greater than or equal to a preset link reference value or the link radiation coefficient greater than or equal to a preset link radiation coefficient;

[0025] The second level state has the link reference value less than a preset link reference value and the link radiation coefficient less than a preset link radiation coefficient.

[0026] Further, the information acquisition unit determines a qualification search mode in response to the level state;

[0027] The information acquisition unit determines an optimization mode according to a one-type region proportion and a one-type region distribution coefficient in response to the first level state;

[0028] The information acquisition unit determines a search priority coefficient of a link according to a link information correlation degree in response to the second level state;

[0029] The search priority coefficient has a positive correlation with the link information correlation degree.

[0030] Further, the optimization unit determines an optimization mode in response to an analysis condition;

[0031] The optimization unit determines a search region according to a link similarity and a region correlation degree in response to the analysis condition that the one-type region proportion is greater than or equal to a preset one-type region proportion or the one-type region distribution coefficient is greater than or equal to a preset one-type region distribution coefficient;

[0032] The analysis condition of the optimization unit response is that the proportion of a type of region is less than a preset proportion of a type of region and a distribution coefficient of a type of region is less than a preset distribution coefficient of a type of region, and the optimization mode is to determine the region search duration of each sub-region according to a region threshold value;

[0033] The search region is a type of region with a link similarity less than a preset link similarity and a region correlation degree less than a preset region correlation degree;

[0034] The region search duration of a single sub-region and the region threshold value are in a positive correlation relationship.

[0035] Further, the evaluation unit determines the file type of the to-be-evaluated bid document according to an association threshold value and a similarity reference value, and the file type includes:

[0036] A type of file with an association threshold value greater than or equal to a preset association threshold value or a similarity reference value greater than or equal to a preset similarity reference value;

[0037] A second type of file with an association threshold value less than a preset association threshold value and a similarity reference value less than a preset similarity reference value.

[0038] Further, the evaluation unit determines the file evaluation coefficient determination mode according to the file type;

[0039] For a type of file, the file evaluation coefficient determination mode is to determine the file evaluation coefficient according to an index fluctuation value and a bid classification value;

[0040] For a second type of file, the file evaluation coefficient determination mode is to determine the file evaluation coefficient according to a bid classification value.

[0041] Compared with the prior art, the beneficial effects of the present application are that in the technical scheme of the present application, the region division mode is determined according to the judgment condition, the judgment condition effectively reflects the link difference degree of the target qualification information corresponding to the target level and the link quantity and link distribution state in the sub-level corresponding to the target level, and then different region division modes are adaptively selected according to the judgment condition, so that the selection of the region division mode is more in line with the actual application scene, and the problem of poor correlation of the disclosed qualification information search result caused by unreasonable region division is avoided, which is beneficial to improve the accuracy of target qualification information review and further improve the accuracy of bid document evaluation.

[0042] Further, the application effectively reflects the link information of the corresponding associated hierarchy of the target hierarchy through the link information degree, and then determines the associated hierarchy similarity according to the link information degree, so that the confirmation method of the associated hierarchy similarity is more in line with the actual application scenario, and the link and the similarity degree of the associated hierarchy of the target hierarchy are effectively reflected through the associated hierarchy similarity and the link difference coefficient, and then the region type is determined according to the associated hierarchy similarity and the link difference coefficient, so that the determination of the region type is more in line with the actual application scenario, and the problem of poor search result relevance of the public qualification information caused by inaccurate region type determination is avoided, and the accuracy of the bid document evaluation is improved.

[0043] Further, the application determines the hierarchy state according to the link reference value and the link radiation coefficient, effectively reflects the number and distribution state of the links in the target hierarchy through the link reference value and the link radiation coefficient, and then adaptively selects different qualification search methods according to the hierarchy state, so that the selection of the qualification search method is more in line with the actual application scenario, and the problem of poor search result relevance of the public qualification information caused by too large search range is avoided, and the accuracy of the evaluation effect is improved.

[0044] Further, the application effectively reflects the distribution of a type of region through the analysis condition, and then adaptively selects different optimization methods according to the analysis condition, so that the selection of the optimization method is more in line with the actual application scenario, and the problem of poor search efficiency caused by large search range and large search difficulty of the public qualification information is avoided, which helps to improve the relevance of the search result, and then the accuracy of the bid document evaluation is improved.

[0045] Further, the application determines the file type of the bid document to be evaluated according to the association threshold and the similarity reference value, effectively reflects the similarity between the bid documents through the association threshold and the similarity reference value, and then adaptively selects different file evaluation coefficient determination methods according to the file type, so that the selection of the file evaluation coefficient determination method is more in line with the actual application scenario, and the problem of poor evaluation accuracy is avoided, and then the accuracy and fairness of the bidding are improved. The bid document to be evaluated with a file evaluation coefficient less than a preset file evaluation coefficient is manually reviewed, not only reduces the cost of manual review, but also can screen out the bid document to be evaluated that may have risks, and then improve the accuracy of the evaluation result. BRIEF DESCRIPTION OF DRAWINGS

[0046] Fig. 1 The unit connection diagram of the bid document intelligent evaluation system of the application fuses quantitative scoring and image recognition;

[0047] Fig. 2 The flowchart of the application for determining the region type according to the associated hierarchy similarity and the link difference coefficient;

[0048] Fig. 3 A flow chart for determining the qualification search mode according to the hierarchical state of the present application;

[0049] Fig. 4 A flow chart for determining the file type of the to-be-evaluated bid file according to the correlation threshold and the similar reference value of the present application. DETAILED DESCRIPTION

[0050] In order to make the objects and advantages of the present application clearer, the present application will be further described below in conjunction with embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application.

[0051] The preferred embodiments of the present application will be described below with reference to the accompanying drawings. It should be understood by those skilled in the art that the embodiments are only used to explain the technical principles of the present application and are not used to limit the protection scope of the present application.

[0052] It should be noted that, in the description of the present application, the terms of direction or position relationship such as "upper", "lower", "left", "right", "inner", "outer" and the like are based on the direction or position relationship shown in the drawings, which is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present application.

[0053] In addition, it should also be noted that, in the description of the present application, unless otherwise explicitly specified and limited, the terms "mounting", "connecting", "connection" should be understood in a broad sense, for example, can be fixed connection, can also be detachable connection, or integrally connected; can be mechanical connection, can also be electrical connection; can be directly connected, can also be indirectly connected through an intermediate medium, can be the communication inside two elements. Those skilled in the art can understand the specific meaning of the above terms in the present application according to the specific circumstances.

[0054] Please refer to Figs. 1 to 4 As shown in the drawings, the present application provides an intelligent bid file review system fusing quantitative scoring and image recognition, comprising:

[0055] A data acquisition unit is used to acquire hierarchical information of target qualification information corresponding to each bid file;

[0056] A search analysis unit is connected with the data acquisition unit, used to determine a region division mode according to a judgment condition, and determine a region type according to a correlation hierarchical similarity and a linkage difference coefficient, the region division mode is uniform division according to a comprehensive evaluation value or correlation division according to a hierarchical threshold;

[0057] An information acquisition unit connected with the search analysis unit, configured to determine a hierarchical state according to the link reference value and the link radiation coefficient, and determine a qualification search mode according to the hierarchical state to acquire the public qualification information, the qualification search mode being to determine an optimization mode according to a type area proportion and a type area distribution coefficient or determine a search priority coefficient of the link according to a link information correlation degree;

[0058] An optimization unit connected with the information acquisition unit, configured to determine an optimization mode according to the analysis condition, the optimization mode being to determine a search area according to a link similarity and a region correlation degree, or determine a region search duration according to a region threshold value;

[0059] An evaluation unit connected with the data acquisition unit, the information acquisition unit and the optimization unit respectively, configured to mark a target qualification information corresponding to a bid document with a qualification difference degree less than a preset qualification difference degree as a to-be-evaluated bid document, determine a file type of the to-be-evaluated bid document according to an association threshold value and a similarity reference value, and determine a file evaluation coefficient judgment mode according to the file type, the file evaluation coefficient judgment mode being to determine a file evaluation coefficient according to an index fluctuation value and a bid classification value or the bid classification value.

[0060] The application scenario of the application is bid document scoring, the hierarchical information includes but is not limited to a link position, a link quantity and a link name in a target hierarchy, the target hierarchy is a network page obtained by searching an enterprise name of a target qualification information, the target hierarchy contains a plurality of links, each link corresponds to a sub-hierarchy, the application extracts the target qualification information in the bid document through an image recognition technology, the image recognition technology can collect an OCR technology, and the specific implementation is not limited, the target qualification information includes but is not limited to an enterprise name, a business license number and financial report data, which are easy to understand for those skilled in the art, and will not be described in detail.

[0061] In the application, a plurality of historical records are correspondingly provided, each historical record records a hierarchical quantity reference value, a hierarchical threshold value total value, an associated hierarchical similarity, a link radiation coefficient and a type area distribution coefficient in a historical process of at least one bid document scoring, and each historical record corresponds to a qualified mark, the qualified mark records whether the accuracy of the bid document scoring meets user demand, and the qualified mark can be recorded manually.

[0062] The qualification difference degree = (the number of keywords in the target qualification information - the number of same keywords in the public qualification information and the target qualification information) / the number of keywords in the target qualification information.

[0063] The preset qualification difference degree value can be determined according to an actual application scene, and the greater the preset qualification difference degree value is, the greater the demand of the user for the score evaluation file without qualification fraud is, and a preset qualification difference degree value is provided, and the preset qualification difference degree is 30%.

[0064] Specifically, the search analysis unit determines the region division manner in response to a judgment condition;

[0065] The judgment condition responded by the search analysis unit is that the level quantity reference value is greater than or equal to a preset level quantity reference value or the link reference value is less than a preset link reference value, and the region division manner is uniform division according to the comprehensive evaluation value;

[0066] The judgment condition responded by the search analysis unit is that the level quantity reference value is less than a preset level quantity reference value and the link reference value is greater than or equal to a preset link reference value, and the region division manner is associated division according to the level threshold.

[0067] In the uniform division according to the comprehensive evaluation value, the target region is divided into rectangular regions with a preset number and equal size, and the preset number and the comprehensive evaluation value are in a positive correlation;

[0068] The judgment condition includes a first judgment condition and a second judgment condition, the first judgment condition is that the level quantity reference value is greater than or equal to a preset level quantity reference value or the link reference value is less than a preset link reference value, and the second judgment condition is that the level quantity reference value is less than a preset level quantity reference value and the link reference value is greater than or equal to a preset link reference value;

[0069] The target region is divided into rectangular regions with a preset number and satisfying a preset condition, wherein each rectangular region is a sub-region obtained by division;

[0070] A tree structure diagram is established with the target level as a reference, node analysis is performed on the target level, all sub-levels corresponding to the target level are extracted, a plurality of new child nodes are created for each sub-level, the child nodes are connected with the root node, node analysis is performed on each sub-level, and each sub-level is connected with the corresponding child node, and the node analysis is stopped when the level threshold of the sub-level is 0, and finally a tree structure diagram is established with the target level as the root node and each sub-level as the child node, which is easily understood by those skilled in the art and will not be described in detail.

[0071] The level quantity reference value is the number of all child nodes in the tree structure diagram established with the target level as a reference, and the link reference value is the total amount of links in the target level; the comprehensive evaluation value = level quantity reference value + link reference value, and the level threshold is the number of links in a single sub-level.

[0072] The values of the preset level quantity reference value and the preset link reference value can be determined by the user according to an actual application scenario. The greater the values of the preset level quantity reference value and the preset link reference value, the greater the demand of the user for uniform division according to the comprehensive evaluation value. A value of the preset level quantity reference value and the preset link reference value is provided. The preset link reference value is 15. According to historical records of uniform division according to the comprehensive evaluation value, an average value of the level quantity reference value corresponding to the historical record that can meet the demand of the user is taken as the preset level quantity reference value.

[0073] Specifically, the search analysis unit performs associated division according to a level threshold, including:

[0074] According to a preset order, associated analysis is performed on each link in the target level. When associated analysis is performed on a single link, the link is recorded as a target link, and a link outside the target link and not recorded in a sub-region is recorded as a reference link.

[0075] The level threshold corresponding to the target link and the level threshold corresponding to each link after the preset order are added one by one until the total value of the level threshold is greater than the total value of the preset level threshold. Each reference link before the preset order of the reference link corresponding to the last level threshold in the addition and the target link are recorded as associated links. The smallest rectangular region that can contain the associated links is recorded as a sub-region.

[0076] And the associated analysis is continued for the links not recorded in the sub-region until each link is recorded in the sub-region.

[0077] The preset order is an order of the associated distance corresponding to each link from small to large.

[0078] The links after the preset order of the target link are each link before the order of the target link after the links are sorted in the associated order from small to large. The reference link corresponding to the last level value in the addition is recorded as an end reference link. Each reference link before the preset order of the end reference link is each reference link before the order of the end reference link.

[0079] The total value of the level threshold is the sum of the level threshold corresponding to each link in the addition. The associated distance is the shortest distance from the link center point to the reference point. The link center point of a single link is the intersection point of the diagonal of the smallest rectangle that can contain the link. The reference point is the vertex at the top left corner of the target level.

[0080] The preset total value of the hierarchy threshold is determined according to an actual application scene. The greater the demand of the user for the richness of the information contained in the sub-regions, the greater the value of the preset total value of the hierarchy threshold. A value of the preset total value of the hierarchy threshold is provided. Historical records of association division according to the hierarchy threshold are detected. An average value of the total value of the hierarchy threshold corresponding to the historical records that can meet the demand of the user is recorded as the preset total value of the hierarchy threshold.

[0081] Specifically, the search analysis unit determines the region type according to the association hierarchy similarity and the link difference coefficient. The region type includes:

[0082] a first type of region in which the association hierarchy similarity is less than a preset association hierarchy similarity or the link difference coefficient is greater than or equal to a preset link difference coefficient;

[0083] a second type of region in which the association hierarchy similarity is greater than or equal to the preset association hierarchy similarity and the link difference coefficient is less than the preset link difference coefficient.

[0084] The association hierarchy is determined as follows. A sub-hierarchy corresponding to a link in a single sub-region is recorded as a first sub-hierarchy. Association analysis is performed on each first sub-hierarchy. When the association analysis is performed on a single first sub-hierarchy, the first sub-hierarchy is recorded as a target first sub-hierarchy. Association analysis is performed on the association hierarchy corresponding to the target first sub-hierarchy. A sub-hierarchy corresponding to a node adjacent to the left of the node corresponding to the target first sub-hierarchy in the tree structure diagram and a sub-hierarchy corresponding to a node adjacent to the right of the node corresponding to the target first sub-page in the tree structure diagram are recorded as association hierarchies. The association analysis is continued on each association hierarchy until the hierarchy threshold of the association hierarchy is 0.

[0085] The sum of the number of links of the matching hierarchies corresponding to a single first sub-hierarchy is recorded as a total sum. The maximum value of the total sum of the links corresponding to each first sub-hierarchy is recorded as b1. The minimum value of the total sum of the links corresponding to each first sub-hierarchy is recorded as b2. The link difference coefficient = (b1-b2) / b1.

[0086] The values of the preset association level similarity and the preset link difference coefficient can be determined by the user according to an actual application scenario. The greater the value of the preset association level similarity and the smaller the value of the preset link difference coefficient, the greater the demand of the user for dividing the sub-regions into one type of region. A value of the preset association level similarity and the preset link difference coefficient is provided. The preset link difference coefficient is 50%. The historical record that can meet the demand of the user and determine the sub-region as a second type of region is recorded as a reference record. In a first confirmation condition, the preset association level similarity is an average value of the association level similarity corresponding to the historical record in the reference record, which is determined according to the number of same keywords. In a second confirmation condition, the preset association level similarity is an average value of the association level similarity corresponding to the historical record in the reference record, which is determined according to the number of associated keywords. The first confirmation condition is that the link information degree is greater than or equal to a preset link information degree. The second confirmation condition is that the link information degree is less than the preset link information degree.

[0087] Specifically, the association level similarity is determined according to the link information degree.

[0088] If the link information degree is greater than or equal to the preset link information degree, the association level similarity is determined according to the number of same keywords. The association level similarity and the number of same keywords are in a positive correlation.

[0089] If the link information degree is less than the preset link information degree, the association level similarity is determined according to the number of associated keywords. The association level similarity and the number of associated keywords are in a positive correlation.

[0090] The link information degree is the number of characters in a single link. The value of the preset link information degree can be determined by the user according to an actual application scenario. The greater the value of the preset link information degree, the greater the demand of the user for determining the association level similarity according to the number of associated keywords. A value of the preset link information degree is provided. The preset link information degree is 20.

[0091] For a single first sub-level, the first sub-level is recorded as a target sub-level. Each association level and the target sub-level obtained by performing association analysis on the target sub-level until the level threshold of the association level is 0 is recorded as an analysis level corresponding to the target sub-level. The link corresponding to each analysis level is recorded as a reference link corresponding to the target sub-level. The name of the reference link corresponding to the target sub-level is recorded as reference text corresponding to the target sub-level. The characters appearing in the analysis level corresponding to the target sub-level are recorded as associated text corresponding to the target sub-level. The number of same keywords is the number of keywords existing in the reference text corresponding to each first sub-level. The number of associated keywords is the number of keywords existing in the associated text corresponding to each first sub-level.

[0092] Specifically, the information acquisition unit determines a level state according to the link reference value and the link radiation coefficient, and the level state includes:

[0093] a first level state in which the link reference value is greater than or equal to a preset link reference value or the link radiation coefficient is greater than or equal to a preset link radiation coefficient;

[0094] a second level state in which the link reference value is less than the preset link reference value and the link radiation coefficient is less than the preset link radiation coefficient.

[0095] The link radiation coefficient is the standard deviation of the associated distances corresponding to each link in the target level. The value of the preset link radiation coefficient can be determined by the user according to the actual application scenario. The greater the value of the preset link radiation coefficient, the greater the user's demand for determining the search priority coefficient according to the link information relevance. A value of the preset link radiation coefficient is provided. The historical records of determining the search priority coefficient according to the link information relevance are detected. The average value of the link radiation coefficients corresponding to the historical records that meet the user's demand is taken as the preset link radiation coefficient.

[0096] Specifically, the information acquisition unit determines a qualification search mode in response to the level state;

[0097] In response to the first level state, the information acquisition unit determines an optimization mode according to a first-type area proportion and a first-type area distribution coefficient.

[0098] In response to the second level state, the information acquisition unit determines a search priority coefficient of a link according to link information relevance.

[0099] The search priority coefficient and the link information relevance are in a positive correlation.

[0100] The first-type area proportion is the number of first-type areas in the target level divided by the number of sub-areas in the target level. The first-type area distribution coefficient is the average value of the reference distances corresponding to each first-type point. For a single first-type point, the first-type point is taken as a target first-type point, and other first-type points other than the target first-type point are taken as reference first-type points. The average value of the shortest distances from the target first-type point to each reference first-type point is taken as the reference distance. For a single link, the link is taken as a target link, the name of the target link and the text of the sub-level corresponding to the target link are taken as the link text corresponding to the target link, and the link information relevance is the number of keywords in the link text corresponding to a single link that are the same as the target qualification information.

[0101] Specifically, the optimization unit determines an optimization mode in response to an analysis condition.

[0102] The analysis condition for which the optimization unit responds is that the proportion of the first type of region is greater than or equal to a preset proportion of the first type of region or the distribution coefficient of the first type of region is greater than or equal to a preset distribution coefficient of the first type of region, and the optimization manner is to determine the search region according to the link similarity and the region correlation degree;

[0103] The analysis condition for which the optimization unit responds is that the proportion of the first type of region is less than a preset proportion of the first type of region and the distribution coefficient of the first type of region is less than a preset distribution coefficient of the first type of region, and the optimization manner is to determine the region search duration of each sub-region according to the region threshold value;

[0104] The search region is a first type of region with a link similarity less than a preset link similarity and a region correlation degree less than a preset region correlation degree;

[0105] The region search duration of a single sub-region and the region threshold value are in a positive correlation relationship.

[0106] The preset proportion of the first type of region and the preset distribution coefficient of the first type of region can be determined by the user according to the actual application scenario. The smaller the preset proportion of the first type of region and the preset distribution coefficient of the first type of region, the greater the demand of the user for determining the search region according to the link similarity and the region correlation degree. A preset proportion of the first type of region is 60%, and a preset distribution coefficient of the first type of region is determined by detecting historical records of determining the region search duration according to the region threshold value, and the average value of the distribution coefficient of the first type of region corresponding to the historical records that can meet the user's demand is recorded as the preset distribution coefficient of the first type of region.

[0107] The confirmation manner of the link similarity and the region correlation degree is that, for a single first type of region, the first type of region is recorded as a target first type of region, other first type of regions outside the target first type of region are recorded as reference regions, and the average value of the similarity coefficients of the target first type of region and each reference region is recorded as the link similarity. The similarity coefficient is the cosine value of the angle between the text vector corresponding to the target first type of region and the text vector corresponding to a single reference region. The region correlation degree is the average value of the hierarchical similarity corresponding to each link in the target first type of region. For a single link in the target first type of region, the link is recorded as a target link, other links outside the target link are recorded as reference links, and the cosine value of the angle between the text vector of the sub-hierarchy corresponding to the target link and the text vector of the sub-hierarchy corresponding to a single reference link is recorded as the hierarchical similarity corresponding to the target link. The text vectors can be obtained by converting the text and the text corresponding to each sub-hierarchy using the TF-IDF technology, which is easily understood by those skilled in the art and will not be described in detail.

[0108] The values of the preset link similarity and the preset area correlation degree can be determined by the user according to an actual application scene. The greater the user's demand for narrowing the search range, the smaller the values of the preset link similarity and the preset area correlation degree. A value of a preset link similarity and a value of a preset area correlation degree are provided. Historical records of determining a search area according to a link similarity and an area correlation degree are detected. An average value of a link similarity corresponding to the historical records that can meet the user's demand is recorded as the preset link similarity. An average value of an area correlation degree corresponding to the historical records that can meet the user's demand is recorded as the preset area correlation degree.

[0109] The area threshold value = the link difference coefficient - the correlation level similarity. The area search duration is the time for searching a single sub-area.

[0110] Specifically, the evaluation unit determines the file type of the to-be-evaluated bid file according to the correlation threshold value and the similarity reference value. The file type includes:

[0111] A first type of file, for which the correlation threshold value is greater than or equal to a preset correlation threshold value or the similarity reference value is greater than or equal to a preset similarity reference value.

[0112] A second type of file, for which the correlation threshold value is less than the preset correlation threshold value and the similarity reference value is less than the preset similarity reference value.

[0113] The correlation threshold value and the similarity reference value are confirmed by the following method. For a single to-be-scored file, the to-be-scored file is recorded as a target to-be-scored file, and other to-be-scored files other than the target to-be-scored file are recorded as reference to-be-scored files. The maximum value of the correlation average values of the target to-be-scored file and each reference to-be-scored file is recorded as the correlation threshold value. The correlation average value is the average value of the correlation coefficients corresponding to each bid index in the two to-be-scored files. The bid index includes but is not limited to material cost, labor cost, management cost, and tax.

[0114] For a single bid index in any two to-be-scored files, the bid index is recorded as a target index. The calculation formula of the correlation coefficient w of the target index in the two to-be-scored files is as follows:

[0115]

[0116] r is the number of data values of the target index in a single to-be-scored file. k and y k are the kth data values of the target index in the two to-be-scored files, respectively. is the average value of each data value in the to-be-scored file corresponding to x k . is the average value of each data value in the to-be-scored file corresponding to y k , k = 1, 2, 3,..., r. ​

[0117] The maximum value of the similar quantity of the target file to be scored and each reference file to be scored is recorded as a similar reference value, and the similar quantity is the number of the same keywords in the target file to be scored and a single reference file to be scored;

[0118] The values of the preset association threshold and the preset similar reference value can be determined by the user according to the actual application scenario. The smaller the values of the preset association threshold and the preset similar reference value are, the greater the demand of the user for scoring according to the bidding classification value is. A value of the preset association threshold and the preset similar reference value is provided. The historical records of scoring according to the bidding classification value are detected. The average value of the association threshold corresponding to the historical records that can meet the user's demand is recorded as the preset association threshold. The average value of the similar reference value corresponding to the historical records that can meet the user's demand is recorded as the preset similar reference value.

[0119] Specifically, the evaluation unit determines a file evaluation coefficient determination mode according to the file type;

[0120] For a type of file, the file evaluation coefficient determination mode is to determine the file evaluation coefficient according to the index fluctuation value and the bidding classification value;

[0121] For a type of file, the file evaluation coefficient determination mode is to determine the file evaluation coefficient according to the bidding classification value.

[0122] For a type of file, the file evaluation coefficient = the bidding classification value - the index fluctuation value. The index fluctuation value is confirmed in the following manner. For a single type of file, the type of file is recorded as a target type of file. The file to be scored with the largest association threshold of the target type of file is recorded as a reference file. The association coefficients corresponding to each bid index of the target type of file and the reference file are obtained. The standard deviation of each association coefficient is recorded as the index fluctuation value.

[0123] For a type of file, the file evaluation coefficient and the bidding classification value have a positive correlation;

[0124] The bidding classification value = the service completeness + the scheme completeness + the price reference value. In the present application, a deep learning network is applied. The deep learning network includes but is not limited to a feedforward neural network, a convolutional neural network, and a recurrent neural network. The user can select according to actual needs. Then, the deep learning network is learned for the to-be-evaluated bid file to obtain the service completeness, the scheme completeness, and the price reference value corresponding to the to-be-evaluated bid file. This is easily understood by those skilled in the art, and specific details are not described.

[0125] After obtaining the file evaluation coefficients corresponding to each to-be-evaluated bidding document, the to-be-evaluated bidding document with a file evaluation coefficient less than the preset file evaluation coefficient is manually reviewed. It can be understood that, by manually reviewing the to-be-evaluated bidding document with a file evaluation coefficient less than the preset file evaluation coefficient, not only the cost of manual review is reduced, but also the to-be-evaluated bidding document that may have risks is screened out, thereby improving the accuracy of the evaluation result.

[0126] The value of the preset file evaluation coefficient can be determined by the user according to the actual application scenario. The greater the user's demand for improving the accuracy of patent review, the greater the value of the preset file evaluation coefficient. A value of the preset file evaluation coefficient is provided, and the historical records of the to-be-evaluated bidding document that is not manually reviewed are detected. The average value of the file evaluation coefficients of the historical records that can meet the user's demand is recorded as the preset file evaluation coefficient.

[0127] So far, the technical solutions of the present application have been described in combination with the preferred embodiments shown in the drawings, but those skilled in the art can easily understand that the protection scope of the present application is obviously not limited to these specific embodiments. Those skilled in the art can make equivalent changes or replacements to the related technical features without departing from the principles of the present application, and the technical solutions after the changes or replacements will fall within the protection scope of the present application.

[0128] The above description is only the preferred embodiments of the present application and is not used to limit the present application; for those skilled in the art, the present application can have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A bid document intelligent review system that fuses quantitative scoring and image recognition, characterized by, The method comprises the following steps: a data acquisition unit is used to obtain hierarchical information of target qualification information corresponding to each bid document; the hierarchical information comprises a link position, a link quantity and a link name in a target hierarchy, and the target hierarchy is a web page obtained by searching for an enterprise name of the target qualification information; a search analysis unit is connected to the data acquisition unit, and is used to determine a region division mode according to a judgment condition, and determine a region type according to a correlation hierarchical similarity and a link difference coefficient, wherein the region division mode is uniform division according to a comprehensive evaluation value or correlation division according to a hierarchical threshold value; the judgment condition comprises a first judgment condition and a second judgment condition, the first judgment condition is that a hierarchical quantity reference value is greater than or equal to a preset hierarchical quantity reference value or a link reference value is less than a preset link reference value, and the second judgment condition is that the hierarchical quantity reference value is less than the preset hierarchical quantity reference value and the link reference value is greater than or equal to the preset link reference value; the correlation hierarchical similarity is determined according to a link information degree; if the link information degree is greater than or equal to a preset link information degree, the correlation hierarchical similarity is determined according to a same keyword quantity, wherein the correlation hierarchical similarity and the same keyword quantity are in a positive correlation relationship; if the link information degree is less than the preset link information degree, the correlation hierarchical similarity is determined according to a correlation keyword quantity, wherein the correlation hierarchical similarity and the correlation keyword quantity are in a positive correlation relationship; the link information degree is a quantity of characters in a single link; a sum of link quantities of a matching hierarchy corresponding to a single first sub-hierarchy is recorded as a link total sum, a maximum value in the link total sum corresponding to each first sub-hierarchy is recorded as b1, and a minimum value in the link total sum corresponding to each first sub-hierarchy is recorded as b2, and a link difference coefficient = (b1-b2) / b1; the hierarchical quantity reference value is a quantity of all child nodes in a tree structure diagram established based on the target hierarchy, the link reference value is a total quantity of links in the target hierarchy, and the comprehensive evaluation value = the hierarchical quantity reference value + the link reference value; the hierarchical threshold value is a quantity of links in a single sub-hierarchy; in the uniform division according to the comprehensive evaluation value, the target region is divided into rectangular regions with a preset quantity and equal sizes, and the preset quantity and the comprehensive evaluation value are in a positive correlation relationship; the search analysis unit performs correlation division according to the hierarchical threshold value, which comprises the following steps: correlation analysis is performed on each link in the target hierarchy in a preset order, and when a single link is subjected to correlation analysis, the link is recorded as a target link, and a link not recorded in a sub-region outside the target link is recorded as a reference link; a hierarchical threshold value corresponding to the target link and hierarchical threshold values corresponding to each link after the preset order are added one by one until a hierarchical threshold value total value is greater than a preset hierarchical threshold value total value, each reference link before the preset order of the reference link corresponding to the last added hierarchical threshold value and the target link are recorded as correlation links, and the smallest rectangular region capable of containing the correlation links is recorded as a sub-region; and the correlation analysis is continued on the links not recorded in the sub-region until each link is recorded in the sub-region; the preset order is an order of correlation distances corresponding to each link from small to large; The information acquisition unit is connected with the search analysis unit, and is configured to determine a level state according to the link reference value and the link radiation coefficient, and determine a qualification search mode according to the level state to acquire the public qualification information, the qualification search mode comprising determining an optimization mode according to a first-class area proportion and a first-class area distribution coefficient, and determining a search priority coefficient of the link according to a link information correlation degree; The link radiation coefficient is a standard deviation of the correlation distance corresponding to each link in the target level, and the correlation distance is the shortest distance from the link center point to the reference point, the link center point of a single link being the intersection point of the diagonal of the smallest rectangle capable of containing the link, and the reference point being the vertex located at the top left corner of the target level; The first-class area distribution coefficient is the average value of the reference distances corresponding to each first-class point, for a single first-class point, the first-class point is denoted as a target first-class point, other first-class points except the target first-class point are denoted as reference first-class points, and the average value of the shortest distances from the target first-class point to each reference first-class point is denoted as a reference distance; for a single link, the link is denoted as a target link, the name of the target link and the text of the sub-level corresponding to the target link are denoted as the link text corresponding to the target link, and the link information correlation degree is the number of the same keywords in the link text corresponding to a single link and in the target qualification information; The optimization unit is connected with the information acquisition unit, and is configured to determine an optimization mode according to an analysis condition, the optimization mode comprising determining a search area according to a link similarity and a region correlation degree, and determining a region search duration according to a region threshold value; The analysis condition to which the optimization unit responds is that the first-class area proportion is greater than or equal to a preset first-class area proportion or the first-class area distribution coefficient is greater than or equal to a preset first-class area distribution coefficient, and the optimization mode is to determine the search area according to the link similarity and the region correlation degree; The analysis condition to which the optimization unit responds is that the first-class area proportion is less than a preset first-class area proportion and the first-class area distribution coefficient is less than a preset first-class area distribution coefficient, and the optimization mode is to determine the region search duration of each sub-region according to the region threshold value; The search area is a first-class area in which the link similarity is less than a preset link similarity and the region correlation degree is less than a preset region correlation degree; The region search duration of a single sub-region and the region threshold value are in a positive correlation relationship; The confirmation mode of the link similarity and the region correlation degree is that, for a single first-class area, the first-class area is denoted as a target first-class area, other first-class areas outside the target first-class area are denoted as reference areas, and the average value of the similarity coefficients of the target first-class area and each reference area is denoted as the link similarity, the similarity coefficient being the cosine value of the included angle between the text vector corresponding to the target first-class area and the text vector corresponding to a single reference area; the region correlation degree being the average value of the level similarities corresponding to each link in the target first-class area, for a single link in the target first-class area, the link is denoted as a target link, other links outside the target link are denoted as reference links, and the cosine value of the included angle between the text vector of the sub-level corresponding to the target link and the text vector of the sub-level corresponding to a single reference link is denoted as the level similarity corresponding to the target link. An evaluation unit connected with the data collection unit, the information acquisition unit and the optimization unit respectively, used to mark the bidding document corresponding to the target qualification information with a qualification difference degree less than a preset qualification difference degree as a to-be-evaluated bidding document, determine the file type of the to-be-evaluated bidding document according to the correlation threshold and the similarity reference value, and determine the file evaluation coefficient determination mode according to the file type, the file evaluation coefficient determination mode being to determine the file evaluation coefficient according to the index fluctuation value and the bidding classification value or according to the bidding classification value; The qualification difference degree = (the number of keywords in the target qualification information - the number of the same keywords in the public qualification information and the target qualification information) / the number of keywords in the target qualification information; For a single to-be-scored file, the to-be-scored file is marked as a target to-be-scored file, other to-be-scored files other than the target to-be-scored file are marked as reference to-be-scored files, and the maximum value of the correlation average of the target to-be-scored file and each reference to-be-scored file is marked as a correlation threshold, the correlation average being the average of the correlation coefficients corresponding to each bid index in the two to-be-scored files, the bid index including material cost, labor cost, management cost and tax; For a single bid index in any two to-be-scored files, the bid index is marked as a target index, and the calculation formula of the correlation coefficient w corresponding to the target index in the two to-be-scored files is: ； r represents the number of data values ​​in the target metric within a single file to be scored; x k and y k These are the k-th data values ​​in the target metrics of the two files to be scored. For x k The average value of each data point in the corresponding scoring file. For y k The average value of each data value in the corresponding file to be scored, k = 1, 2, 3, ..., r; The maximum value of the similarity number of the target to-be-scored file and each reference to-be-scored file is marked as a similarity reference value, and the similarity number is the number of the same keywords in the target to-be-scored file and a single reference to-be-scored file; The confirmation mode of the index fluctuation value is that, for a single one-type file, the one-type file is marked as a target one-type file, the to-be-scored file with the maximum correlation threshold of the target one-type file is marked as a reference file, the correlation coefficients corresponding to each bid index of the target one-type file and the reference file are obtained, and the standard deviation of each correlation coefficient is marked as an index fluctuation value; For a one-type file, the file evaluation coefficient determination mode is to determine the file evaluation coefficient according to the index fluctuation value and the bidding classification value, and the file evaluation coefficient = the bidding classification value - the index fluctuation value; For a two-type file, the file evaluation coefficient determination mode is to determine the file evaluation coefficient according to the bidding classification value, the file evaluation coefficient and the bidding classification value are in a positive correlation relationship, and the bidding classification value = service completeness + scheme completeness + price reference value; The file type includes a one-type file with a correlation threshold greater than or equal to a preset correlation threshold or a similarity reference value greater than or equal to a preset similarity reference value, and a two-type file with a correlation threshold less than a preset correlation threshold and a similarity reference value less than a preset similarity reference value; The service completeness, the scheme completeness and the price reference value are obtained by learning the to-be-evaluated bidding document through a feedforward neural network, a convolutional neural network or a recurrent neural network.

2. The intelligent bid evaluation system of fusion of quantitative score and image recognition according to claim 1, characterized in that, The judgment condition for which the search analysis unit responds is that the hierarchical quantity reference value is greater than or equal to a preset hierarchical quantity reference value or the link reference value is less than a preset link reference value, and the area division mode is to divide uniformly according to the comprehensive evaluation value. The judgment condition responded by the search analysis unit is that the level quantity reference value is less than a preset level quantity reference value and the link reference value is greater than or equal to a preset link reference value, and the area division mode is associated division according to a level threshold.

3. The intelligent bid evaluation system of fusion of quantitative score and image recognition according to claim 1, characterized in that, The search analysis unit determines an area type according to an associated level similarity and a link difference coefficient, and the area type includes: a first type of area with the associated level similarity being less than a preset associated level similarity or the link difference coefficient being greater than or equal to a preset link difference coefficient; a second type of area with the associated level similarity being greater than or equal to the preset associated level similarity and the link difference coefficient being less than a preset link difference coefficient.

4. The intelligent bid evaluation system for fusion of quantitative score and image recognition according to claim 1, characterized in that, The information acquisition unit determines a level state according to a link reference value and a link radiation coefficient, and the level state includes: a first level state with the link reference value being greater than or equal to a preset link reference value or the link radiation coefficient being greater than or equal to a preset link radiation coefficient; a second level state with the link reference value being less than a preset link reference value and the link radiation coefficient being less than a preset link radiation coefficient.

5. The intelligent bid evaluation system of fusion quantitative score and image recognition according to claim 4, characterized in that, The information acquisition unit determines a qualification search mode in response to the level state. The information acquisition unit determines an optimization mode according to a proportion of the first type of area and a distribution coefficient of the first type of area in response to the first level state. The information acquisition unit determines a search priority coefficient of a link according to a link information correlation in response to the second level state. The search priority coefficient and the link information correlation are in a positive correlation.

Images