A system and method for digital intelligent contract recognition

Abstract

This invention provides a system and method for intelligent digital contract recognition, belonging to the field of text recognition technology. The system comprises: a data acquisition module that uses an image acquisition device to capture an initial image of a target contract and pre-processes the initial image to obtain a standard image in the format of an electronic copy; an intelligent recognition module that uses an optical character recognition (OCR) tool to recognize the text content of the standard image and convert it into raw text data; a text processing module that performs text processing on the raw text data to obtain the standard text data of the target contract; and a text analysis module that uses natural language recognition technology to analyze the standard text data, extract key contract data, and integrate the key contract data into formatted contract data with a standard structure. This system improves contract recognition efficiency and accuracy and optimizes the system's recognition process.

Description

Technical Field

[0001] The present invention relates to the technical field of text recognition, and in particular to a system and method for digital intelligent contract recognition. Background Art

[0002] At present, with the development of image recognition text technology, contracts are important transaction documents for enterprises. However, the recognition of contract images and the extraction of key elements usually require a lot of manpower for extraction and identification, and the labor cost is also high, resulting in low recognition efficiency when processing contracts.

[0003] Therefore, the present invention proposes a system and method for digital intelligent contract identification. Summary of the Invention

[0004] The present invention provides a system and method for digital intelligent contract recognition, which is used to convert physical contracts into electronic format by utilizing image scanning technology, then recognize the text in the image using an OCR tool, and then analyze the pre-processed text to extract the contract terms and integrate them into structured data, thereby improving the efficiency and accuracy of contract recognition and optimizing the system recognition process.

[0005] In one aspect, the present invention provides a system for digitally intelligently identifying contracts, comprising:

[0006] Data acquisition module: uses image acquisition equipment to acquire the initial image of the target contract, and pre-processes the initial image to obtain a standard image in the electronic copy format;

[0007] Intelligent recognition module: uses OCR tools to recognize the text content of standard images and converts the image text content into original text data;

[0008] Text processing module: performs text processing on the original text data to obtain the standard text data of the target contract;

[0009] Text analysis module: uses natural language recognition technology to analyze standard text data, extract key contract data, and integrate the key contract data into formatted contract data with a standard structure.

[0010] On the other hand, the data acquisition module includes:

[0011] Equipment configuration unit: Based on the standard image requirements of the recognition contract, determine the type of image acquisition equipment, select a standard shooting environment based on the surrounding environment parameters obtained by the sensor, and perform installation and configuration based on the installation manual of the image acquisition equipment;

[0012] Image acquisition unit: used to acquire the initial image of the target contract using the installed image acquisition device.

[0013] On the other hand, the data acquisition module includes:

[0014] A pre-processing unit is configured to enhance the initial image to obtain a clear image in an electronic copy format, and adjust the clear image according to a preset standard ratio to obtain a first image;

[0015] Correction unit: performs horizontal correction processing on the text portion of the first image, and uses a blank area recognition algorithm to crop the area to be cropped, thereby obtaining a second image;

[0016] Compression unit: compresses the second image to a standard image size using a compression tool to obtain a standard image of the target contract.

[0017] On the other hand, the intelligent identification module includes:

[0018] The first processing unit is configured to extract features of the standard image by using an OCR tool after the standard image is loaded, thereby generating a first feature image.

[0019] The second processing unit is configured to rotate the first feature image by 8 angles to obtain directional feature images at 8 angles, and perform feature extraction on the 8 directional feature images to obtain 8 directional feature values;

[0020] Vector acquisition unit: converts each directional feature value into a corresponding directional vector according to a vector conversion standard, and at the same time, uses the feature vector of the first feature image as the center vector;

[0021] Matrix construction unit: constructs a covariance matrix based on all direction vectors and the center vector;

[0022] Projection unit: obtains a mixed eigenvector according to the covariance matrix, projects the mixed eigenvector onto each direction vector to obtain a projection coefficient, and uses the corresponding projection coefficient as a weight coefficient of the corresponding directional feature image.

[0023] On the other hand, the intelligent identification module further includes:

[0024] Fusion unit: performs feature fusion on the eight directional vectors based on the weight coefficient of the directional feature image to obtain a fused feature vector, wherein the fused feature vector contains several element features, and each element feature contains a horizontal sub-feature and a vertical sub-feature;

[0025] Matching unit: Matching the fused feature vector with the feature-text library to obtain the text feature corresponding to each element feature in the fused feature vector;

[0026] Calculate the initial similarity between each element feature and the corresponding matching text feature;

[0027] ;

[0028] in, represents the initial similarity between the i-th element feature in the fused feature vector and the corresponding matching text feature; Represents the horizontal gradient difference between the i-th element feature in the fusion feature vector and the corresponding text feature, Represents the vertical gradient difference between the i-th element feature in the fusion feature vector and the corresponding text feature, ) represents the tangent function, and Respectively represent the i-th element feature and the corresponding text feature in the fusion feature vector, Indicates based on The dot product function of express norm ratio;

[0029] Correcting the initial similarity to obtain a final similarity of corresponding element features;

[0030] ;

[0031] in, Represents the final similarity of the i-th element feature in the fusion feature vector, ) represents the exponential function, Indicates the maximum matching similarity of all element features in the fusion feature vector. Indicates the minimum matching similarity of all element features in the fusion feature vector, Represents the mean similarity of all element features in the fused feature vector; Indicates that based on all The variance of The mean of the variances;

[0032] Based on the final similarity of each element feature in the fused feature vector, a corresponding text conversion scheme is matched from a similarity-text conversion mapping table to convert the corresponding image text content into original text data.

[0033] On the other hand, the text processing module includes:

[0034] Cleaning unit: traverses the original text data, marks unrecognized words in the original text data as target words, and compares the target words with the standard dictionary;

[0035] If a standard similar word exists in the target word, the target word is converted into a standard similar word with the highest similarity; otherwise, the target word is determined to be an unknown word and deleted, and the first contract data is output after traversal.

[0036] Other format unit: Filters and converts unstructured data in the original text data in combination with a preset target text format, and outputs the second contract data;

[0037] The text proofreading unit corrects misspelled words in the first contract data and the second contract data based on a proofreader to obtain standard text data.

[0038] On the other hand, the text analysis module includes:

[0039] Word segmentation unit: used to segment standard text data into word sequences, and filter out useless stop words based on a general word dictionary to obtain a standard word sequence;

[0040] Entity recognition unit: Based on the contract entity library, the named entity recognition system is used to identify entities in standard word sequences and obtain entity sequences;

[0041] Keyword extraction unit: Based on the contract keyword recognition library, perform keyword extraction on the standard word sequence to obtain a keyword sequence;

[0042] Judgment unit: Take the union of the entity sequence and the keyword sequence, filter out the useless words in the union, and obtain the key contract data.

[0043] On the other hand, the text analysis module includes:

[0044] Syntax extraction unit: performs grammatical analysis on key contract data, determines the relationship and hierarchical structure between terms, and adds grammatical terms in a grammatical logical order to generate contract terms;

[0045] Structured data unit: standardizes the contract terms according to the format of the standard preset contract structure to obtain the formatted contract data of the target contract.

[0046] On the other hand, a method for digitally intelligently identifying contracts includes:

[0047] Step 1: Use an image acquisition device to acquire an initial image of the target contract, and pre-process the initial image to obtain a standard image in the electronic copy format;

[0048] Step 2: Use OCR tools to recognize the text content of the standard image and convert it into original text data;

[0049] Step 3: Perform text processing on the original text data to obtain the standard text data of the target contract;

[0050] Step 4: Use natural language recognition technology to analyze standard text data, extract key contract data, and integrate the key contract data into formatted contract data with a standard structure.

[0051] The present invention provides a system and method for digital intelligent contract recognition, which is used to convert physical contracts into electronic format by utilizing image scanning technology, then recognize the text in the image using OCR tools, and finally use NLP technology to analyze the pre-processed text, extract the contract terms and integrate them into structured data, thereby improving the efficiency and accuracy of contract recognition and optimizing the system recognition process. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0053] Figure 1 This is a schematic diagram of the system structure of a digital intelligent contract recognition system provided by an embodiment of the present invention;

[0054] Figure 2 This is a flow chart of a method for digital intelligent contract identification provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0055] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0056] Example 1:

[0057] like Figure 1 As shown, an embodiment of the present invention provides a digital intelligent contract recognition system, including:

[0058] Data acquisition module: uses image acquisition equipment to acquire the initial image of the target contract, and pre-processes the initial image to obtain a standard image in the electronic copy format;

[0059] Intelligent recognition module: uses OCR tools to recognize the text content of standard images and converts the image text content into original text data;

[0060] Text processing module: performs text processing on the original text data to obtain the standard text data of the target contract;

[0061] Text analysis module: uses natural language recognition technology to analyze standard text data, extract key contract data, and integrate the key contract data into formatted contract data with a standard structure.

[0062] In this embodiment, the image acquisition device is any device used to photograph or scan documents, such as a scanner, a camera, or a webcam.

[0063] In this embodiment, the target contract refers to a specific contract document that needs to be processed and analyzed, and may be various types of legal contracts, commercial contracts, or other types of agreement documents.

[0064] In this embodiment, the initial image refers to a basic image of the target contract acquired from an image acquisition device, and the image includes text, graphics, and other elements.

[0065] In this embodiment, preprocessing is a series of operations performed on the initial image, including removing noise, adjusting the contrast and brightness of the image, and cropping or rotating the image.

[0066] In this embodiment, the electronic copy format refers to a standard format for converting a paper document into an electronic form, such as PDF, image file (such as JPEG, PNG), etc.

[0067] In this embodiment, the standard image refers to an image that has been pre-processed and meets specific format and quality standards, and is used for subsequent OCR processing.

[0068] In this embodiment, the OCR tool is a technical tool that scans an image and converts the text therein into editable text.

[0069] In this embodiment, the image text content refers to all text information displayed in the image. In a contract scan image, the text content includes all printed or handwritten text such as contract terms, title, signature, date, etc.

[0070] In this embodiment, the raw text data refers to the unprocessed text content recognized from the image, including all text in the contract, including clauses, titles, dates, signatures, etc.

[0071] In this embodiment, text processing refers to various operations and processing performed on original text data, including text cleaning, text segmentation, and removal of stop words.

[0072] In this embodiment, standard text data is text data processed by a text processing module, has a certain format and structure, and is easy to further analyze and process. Characteristics include: clear readability, semantic accuracy, noise removal, etc.

[0073] In this embodiment, natural language recognition technology includes: text analysis, named entity recognition, sentiment analysis, semantic understanding and other technical contents.

[0074] In this embodiment, key contract data refers to data of great significance in the contract, including contract terms and conditions, date and time, information of participating parties, amount and fees, agreed matters, etc.

[0075] In this embodiment, the standard structure refers to a unified format preset for the contract identification text.

[0076] In this embodiment, the formatted contract data is data obtained by arranging and processing the contract text data according to a standard structure.

[0077] The working principle and beneficial effects of the above technical solution are: using image acquisition equipment to obtain contract images, using OCR tools to recognize text, and then performing text processing and analysis to convert contract data into structured formatted data, thereby improving the efficiency and accuracy of contract text recognition and realizing the automation and intelligence of contract management.

[0078] Example 2:

[0079] Based on the above embodiment 1, the data acquisition module includes:

[0080] Equipment configuration unit: Based on the standard image requirements of the recognition contract, determine the type of image acquisition equipment, select a standard shooting environment based on the surrounding environment parameters obtained by the sensor, and perform installation and configuration based on the installation manual of the image acquisition equipment;

[0081] Image acquisition unit: used to acquire the initial image of the target contract using the installed image acquisition device.

[0082] In this embodiment, the standard image requirements include: clarity, resolution, contrast, angle, direction and other requirements.

[0083] In this embodiment, the sensor is a sensor for acquiring various parameters in the image acquisition device, including: a light sensor, a temperature sensor, a humidity sensor, a positioning sensor and other sensors.

[0084] In this embodiment, the ambient environment parameters refer to environmental factors that affect the quality and stability of image acquisition, including parameters such as light intensity, light source type, temperature, and ambient color.

[0085] In this embodiment, the standard shooting environment refers to a set of environmental conditions that meet the working requirements of the image acquisition device, including: appropriate lighting conditions, stable background, suitable temperature and humidity, appropriate shooting angle and distance, etc.

[0086] In this embodiment, the installation manual refers to the instruction manual or guide of the image acquisition device, including: device overview, installation requirements, installation steps, environment configuration, device calibration, etc.

[0087] The working principle and beneficial effects of the above technical solution are: by configuring appropriate image acquisition equipment and environment, the accuracy and efficiency of contract recognition are improved, the image processing process is optimized, and the error rate is reduced.

[0088] Example 3:

[0089] Based on the above embodiment 1, the data acquisition module includes:

[0090] A pre-processing unit is configured to enhance the initial image to obtain a clear image in an electronic copy format, and adjust the clear image according to a preset standard ratio to obtain a first image;

[0091] Correction unit: performs horizontal correction processing on the text portion of the first image, and uses a blank area recognition algorithm to crop the area to be cropped, thereby obtaining a second image;

[0092] Compression unit: compresses the second image to a standard image size using a compression tool to obtain a standard image of the target contract.

[0093] In this embodiment, image enhancement refers to a series of image processing technologies, including operations such as improving image quality, enhancing image features, and removing noise from images.

[0094] In this embodiment, a clear image refers to an image with high quality, clear details and vivid colors after image enhancement processing.

[0095] In this embodiment, the preset standard ratio is a pre-set size ratio of the contract image, such as the aspect ratio of the image.

[0096] In this embodiment, the first image is an image obtained after being processed by the pre-processing unit and adjusted according to a preset standard ratio.

[0097] In this embodiment, the horizontal correction processing refers to adjusting the text portion in the image so that it is arranged or aligned horizontally, including operations such as rotation and text distortion adjustment.

[0098] In this embodiment, the blank area recognition algorithm is an algorithm for detecting blank areas in an image. A blank area is an area without text or content. The algorithm is implemented based on image pixel values, edge detection, region segmentation and other technologies.

[0099] In this embodiment, the area to be cropped refers to an area that is identified as needing to be cropped during processing by the correction unit.

[0100] In this embodiment, the second image is the first image processed by the correction unit.

[0101] In this embodiment, the compression tool is various image compression software or tools, and common compression tools include Adobe Photoshop, GIMP, etc.

[0102] The working principle and beneficial effects of the above technical solution are: through the three steps of preprocessing, correction and compression, the quality and clarity of the contract image are improved, irrelevant information is removed, and finally a standardized image is obtained, which helps to improve the accuracy of text recognition and the efficiency of subsequent processing.

[0103] Example 4:

[0104] Based on the above embodiment 1, the intelligent identification module includes:

[0105] The first processing unit is configured to extract features of the standard image by using an OCR tool after the standard image is loaded, thereby generating a first feature image.

[0106] The second processing unit is configured to rotate the first feature image by 8 angles to obtain directional feature images at 8 angles, and perform feature extraction on the 8 directional feature images to obtain 8 directional feature values;

[0107] Vector acquisition unit: converts each directional feature value into a corresponding directional vector according to a vector conversion standard, and at the same time, uses the feature vector of the first feature image as the center vector;

[0108] Matrix construction unit: constructs a covariance matrix based on all direction vectors and the center vector;

[0109] Projection unit: obtains a mixed eigenvector according to the covariance matrix, projects the mixed eigenvector onto each direction vector to obtain a projection coefficient, and uses the corresponding projection coefficient as a weight coefficient of the corresponding directional feature image.

[0110] In this embodiment, the recognition algorithm is an algorithm for extracting image features, including: scale-invariant feature transformation, oriented gradient histogram, convolutional neural network, principal component analysis and other algorithmic technologies.

[0111] In this embodiment, the features of the image refer to relevant information in the image extracted and processed by the processing unit, including the texture, shape, edge, color distribution, etc. of the image.

[0112] In this embodiment, the first feature image is an image feature representation extracted by a recognition algorithm after a standard image is loaded through an OCR tool.

[0113] In this embodiment, the directional characteristic image refers to an image obtained by rotating the first characteristic image to 8 different angles. Each directional characteristic image represents the characteristic performance of the original image at a different angle.

[0114] In this embodiment, feature extraction is a technology for extracting image features in image processing, such as edge detection, histogram statistics, Fourier transform, etc.

[0115] In this embodiment, the directional feature value refers to the feature value extracted from the feature image after each angle rotation, including features that can be represented by numerical values, such as edges, textures, and corner points.

[0116] In this embodiment, the vector conversion standard refers to the standard used in the process of converting the eigenvalue into the direction vector. The conversion method may, for example, use a simple normalization method to map the eigenvalue into a unit vector.

[0117] In this embodiment, the center vector is the feature vector extracted from the first feature image. For example, the center image A is , the vector is .

[0118] In this embodiment, the direction vector is a vector that represents the image feature in a specific direction. For example, the image obtained by rotating the center vector A by 45 degrees is: , then the direction vector is .

[0119] In this embodiment, the covariance matrix is ​​a matrix used to describe the relationship between multi-dimensional random variables, as follows:

[0120] ;

[0121] Every element Represents the covariance between the corresponding direction vector and the center vector, reflecting the linear relationship between them.

[0122] In this embodiment, the mixed feature vector is generated by the covariance matrix and is used to represent the mixing of image features in different directions. For example, the mixed feature vector obtained by the above covariance matrix A is , is Each element is obtained by eigenvalue extraction Represents the mixed features of the corresponding direction feature vector and the center vector.

[0123] In this embodiment, projection refers to a process of mapping one vector onto another vector.

[0124] In this embodiment, the projection coefficient represents the degree of mapping of the mixed feature vector on each direction feature image. For example, The mixed eigenvector is projected onto the kth direction vector The final projection coefficient is .

[0125] In this embodiment, the weight coefficient is a coefficient indicating the contribution of the mixed feature vector to each directional feature image.

[0126] The working principle and beneficial effects of the above technical solution are: through multi-angle feature extraction and weight adjustment, the expressive ability and classification performance of image features are improved, and the accuracy of system text recognition is enhanced.

[0127] Example 5:

[0128] Based on the above embodiment 4, the intelligent identification module further includes:

[0129] Fusion unit: performs feature fusion on the eight directional vectors based on the weight coefficient of the directional feature image to obtain a fused feature vector, wherein the fused feature vector contains several element features, and each element feature contains a horizontal sub-feature and a vertical sub-feature;

[0130] Matching unit: Matching the fused feature vector with the feature-text library to obtain the text feature corresponding to each element feature in the fused feature vector;

[0131] Calculate the initial similarity between each element feature and the corresponding matching text feature;

[0132] ;

[0133] in, represents the initial similarity between the i-th element feature in the fused feature vector and the corresponding matching text feature; Represents the horizontal gradient difference between the i-th element feature in the fusion feature vector and the corresponding text feature, Represents the vertical gradient difference between the i-th element feature in the fusion feature vector and the corresponding text feature, ) represents the tangent function, and Respectively represent the i-th element feature and the corresponding text feature in the fusion feature vector, Indicates based on The dot product function of express norm ratio;

[0134] Correcting the initial similarity to obtain a final similarity of corresponding element features;

[0135] ;

[0136] in, Represents the final similarity of the i-th element feature in the fusion feature vector, ) represents the exponential function, Indicates the maximum matching similarity of all element features in the fusion feature vector. Indicates the minimum matching similarity of all element features in the fusion feature vector, Represents the mean similarity of all element features in the fused feature vector; Indicates that based on all The variance of The mean of the variances;

[0137] Based on the final similarity of each element feature in the fused feature vector, a corresponding text conversion scheme is matched from a similarity-text conversion mapping table to convert the corresponding image text content into original text data.

[0138] In this embodiment, feature fusion refers to combining feature information from different directions to form a comprehensive feature representation. For example, the feature of vector 1 is , the characteristics of vector 2 are , then their fusion vector is: ,in is the weight coefficient of vector 1, is the weight coefficient of vector 2.

[0139] In this embodiment, the fused feature vector is a result generated during the feature fusion process, and is a collection of feature information extracted from multiple direction vectors.

[0140] In this embodiment, the element feature refers to each part in the fused feature vector, and these parts represent different information in the image.

[0141] In this embodiment, the horizontal sub-feature refers to the horizontal component of each element feature in the fused feature vector.

[0142] In this embodiment, the vertical sub-feature refers to the vertical component of each element feature in the fused feature vector.

[0143] In this embodiment, the feature-text library is a database that stores the correspondence between features and texts.

[0144] In this embodiment, the text feature is text information used to describe the image content, including text descriptions of the image content, keywords, classification labels, etc.

[0145] In this embodiment, the initial similarity refers to the initial similarity between each element feature and the corresponding text feature calculated in the matching unit.

[0146] In this embodiment, the horizontal gradient difference refers to the difference between the horizontal gradients of the i-th element feature in the fused feature vector and the corresponding text feature.

[0147] In this embodiment, the vertical gradient difference is the difference between the vertical gradient of the i-th element feature in the fused feature vector and the corresponding text feature.

[0148] In this embodiment, the dot product refers to the result obtained by multiplying the elements at corresponding positions in two vectors and then adding them together. For example, the dot product of A and B is:

[0149]

[0150] In this embodiment, the norm refers to a measure of the size of a vector, which is the square root of the sum of the squares of all elements in the vector.

[0151] In this embodiment, the final similarity refers to the final similarity between each element feature and the corresponding matching text feature calculated based on the modified initial similarity.

[0152] In this embodiment, the matching similarity refers to the similarity between the element feature and the corresponding matching text feature.

[0153] In this embodiment, the similarity-text conversion mapping table is a table that maps similarity values ​​to corresponding text conversion solutions.

[0154] In this embodiment, the text conversion scheme is obtained based on a similarity-text conversion mapping table, and the content types include: mapping feature similarity to corresponding text content, identifiers, etc. according to predefined rules.

[0155] The working principle and beneficial effects of the above technical solution are: using directional feature images to fuse feature vectors, and realizing image-to-text conversion through similarity calculation and correction, thereby improving matching accuracy and data conversion efficiency.

[0156] Example 6:

[0157] Based on the above embodiment 1, the text processing module includes:

[0158] Cleaning unit: traverses the original text data, marks unrecognized words in the original text data as target words, and compares the target words with the standard dictionary;

[0159] If a standard similar word exists in the target word, the target word is converted into a standard similar word with the highest similarity; otherwise, the target word is determined to be an unknown word and deleted, and the first contract data is output after traversal.

[0160] Other format unit: Filters and converts unstructured data in the original text data in combination with a preset target text format, and outputs the second contract data;

[0161] The text proofreading unit corrects misspelled words in the first contract data and the second contract data based on a proofreader to obtain standard text data.

[0162] In this embodiment, unrecognized words refer to words for which no matching items can be found in the vocabulary library during text processing.

[0163] In this embodiment, the target vocabulary refers to the vocabulary marked as unrecognized in the original text data.

[0164] In this embodiment, the standard dictionary refers to a dictionary that includes a vocabulary of terms in the contract field and their corresponding definitions or explanations.

[0165] In this embodiment, vocabulary comparison refers to the process of comparing and matching the vocabulary in the original text data with the vocabulary in the standard dictionary.

[0166] In this embodiment, the standard similar vocabulary refers to known vocabulary similar to the target vocabulary found in a standard dictionary.

[0167] In this embodiment, unknown words refer to words for which no similar words can be found in a standard dictionary when processing text data.

[0168] In this embodiment, the first contract data refers to data processed by the vocabulary cleaning unit.

[0169] In this embodiment, unstructured data refers to data that does not have a clearly defined format or pattern in the text, and types include: free text, unordered lists, free-format tables, etc.

[0170] In this embodiment, the preset target text format refers to the text format structure defined when processing unstructured data, including the layout of the document, the arrangement of fields, format requirements of specific data, etc.

[0171] In this embodiment, filtering refers to removing portions that do not conform to a preset target text format from the original text data.

[0172] In this embodiment, the transformation converts the unstructured data into data that is recognizable and conforms to a preset target text format.

[0173] In this embodiment, the second contract data refers to text data obtained by processing unstructured data.

[0174] The working principle and beneficial effects of the above technical solution are: through the cleaning unit to identify and convert unrecognized words, other format units to convert unstructured data, and the text proofreading unit to correct spelling errors, automated contract text cleaning and format unification are achieved, thereby improving text accuracy and readability.

[0175] Example 7:

[0176] Based on the above embodiment 1, the text analysis module includes:

[0177] Word segmentation unit: used to segment standard text data into word sequences, and filter out useless stop words based on a general word dictionary to obtain a standard word sequence;

[0178] Entity recognition unit: Based on the contract entity library, the named entity recognition system is used to identify entities in standard word sequences and obtain entity sequences;

[0179] Keyword extraction unit: Based on the contract keyword recognition library, perform keyword extraction on the standard word sequence to obtain a keyword sequence;

[0180] Judgment unit: Take the union of the entity sequence and the keyword sequence, filter out the useless words in the union, and obtain the key contract data.

[0181] In this embodiment, the word sequence refers to a sequence of words obtained by segmenting the text according to spaces or other symbols.

[0182] In this embodiment, the general word dictionary refers to a dictionary containing common words.

[0183] In this embodiment, useless stop words refer to words that lack practical meaning or carry little information in the context of text processing tasks, such as "is", "of", "and", etc.

[0184] In this embodiment, the standard word sequence refers to a word sequence in text data obtained after word segmentation and filtering.

[0185] In this embodiment, the contract entity library refers to a database containing contract-related entities, including specific information such as names of people, organizations, dates, and locations in the contract.

[0186] In this embodiment, an entity sequence refers to a word sequence in a text that is recognized as having a specific entity type.

[0187] In this embodiment, the contract keyword recognition database is a database containing contract-related keywords, including: contract type, terms, amount, date, responsibility, etc.

[0188] In this embodiment, the keyword sequence refers to a sequence of words that are identified as having importance in the text.

[0189] The working principle and beneficial effects of the above technical solution are: through word segmentation, entity recognition and keyword extraction, useless words are filtered out, the accuracy and efficiency of contract data processing are improved, and more accurate key contract data is obtained.

[0190] Example 8:

[0191] Based on the above embodiment 1, the text analysis module includes:

[0192] Syntax extraction unit: performs grammatical analysis on key contract data, determines the relationship and hierarchical structure between terms, and adds grammatical terms in a grammatical logical order to generate contract terms;

[0193] Structured data unit: standardizes the contract terms according to the format of the standard preset contract structure to obtain the formatted contract data of the target contract.

[0194] In this embodiment, grammatical analysis is a technology for processing natural language, which determines the relationship and hierarchical structure between words by analyzing the structure and grammatical rules of sentences.

[0195] In this embodiment, the hierarchical structure refers to the organizational structure between contract clauses, that is, the relationship and arrangement order between different clauses. For example, the main clauses, sub-clauses, additional clauses, etc. in a contract constitute the hierarchical structure of the contract.

[0196] In this embodiment, the grammatical logical order refers to the order of organizing text data according to grammatical rules and logical order.

[0197] In this embodiment, the contract terms refer to specific clauses or items stipulated in the contract, including payment terms, delivery terms, confidentiality terms, dispute resolution terms, etc.

[0198] In this embodiment, the standard preset contract structure refers to a contract structure and format developed according to a specific industry, legal requirements or practices.

[0199] In this embodiment, the modification and standardization process refers to the process of modifying and standardizing the contract terms. Modification is to make necessary modifications and amendments to the contract terms based on the results of grammatical analysis and the grammatical logical order; standardization is to adjust the contract terms according to the preset contract standard structure.

[0200] The working principle and beneficial effects of the above technical solution are: automatic extraction, analysis and formatting of contract terms are achieved through natural language processing, which improves the efficiency and accuracy of contract processing and enhances the efficiency of contract management.

[0201] Example 9:

[0202] A method for digital intelligent identification of contracts, such as Figure 2 Shown, including:

[0203] Step 1: Use an image acquisition device to acquire an initial image of the target contract, and pre-process the initial image to obtain a standard image in the electronic copy format;

[0204] Step 2: Use OCR tools to recognize the text content of the standard image and convert it into original text data;

[0205] Step 3: Perform text processing on the original text data to obtain the standard text data of the target contract;

[0206] Step 4: Use natural language recognition technology to analyze standard text data, extract key contract data, and integrate the key contract data into formatted contract data with a standard structure.

[0207] The working principle and beneficial effects of the above technical solution are: using image acquisition equipment to obtain contract images, using OCR tools to recognize text, and then performing text processing and analysis to convert contract data into structured formatted data, thereby improving the efficiency and accuracy of contract text recognition and realizing the automation and intelligence of contract management.

[0208] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A digital intelligent contract recognition system, characterized by: include: Data acquisition module: uses image acquisition equipment to acquire the initial image of the target contract, and pre-processes the initial image to obtain a standard image in the electronic copy format; Intelligent recognition module: uses OCR tools to recognize the text content of standard images and converts the image text content into original text data; Text processing module: performs text processing on the original text data to obtain the standard text data of the target contract; Text analysis module: uses natural language recognition technology to analyze standard text data, extract key contract data, and integrate the key contract data into formatted contract data with a standard structure; The intelligent recognition module includes: The first processing unit is configured to extract features of the standard image by using an OCR tool after the standard image is loaded, thereby generating a first feature image. The second processing unit is configured to rotate the first feature image by 8 angles to obtain directional feature images at 8 angles, and perform feature extraction on the 8 directional feature images to obtain 8 directional feature values; Vector acquisition unit: converts each directional feature value into a corresponding directional vector according to a vector conversion standard, and at the same time, uses the feature vector of the first feature image as the center vector; Matrix construction unit: constructs a covariance matrix based on all direction vectors and the center vector; Projection unit: obtains a mixed eigenvector according to the covariance matrix, projects the mixed eigenvector onto each direction vector to obtain a projection coefficient, and uses the corresponding projection coefficient as a weight coefficient of the corresponding directional feature image; Fusion unit: performs feature fusion on the eight directional vectors based on the weight coefficient of the directional feature image to obtain a fused feature vector, wherein the fused feature vector contains several element features, and each element feature contains a horizontal sub-feature and a vertical sub-feature; Matching unit: Based on the feature-text library and the fused feature vector, the text feature corresponding to each element feature in the fused feature vector is obtained; the similarity between each element feature and the corresponding matching text feature is calculated respectively; based on the similarity of each element feature in the fused feature vector, the corresponding text conversion scheme is matched from the similarity-text conversion mapping table to convert the corresponding image text content into original text data.

2. A digital intelligent contract recognition system according to claim 1, characterized in that: The data acquisition module includes: Equipment configuration unit: Based on the standard image requirements of the recognition contract, determine the type of image acquisition equipment, select a standard shooting environment based on the surrounding environment parameters obtained by the sensor, and perform installation and configuration based on the installation manual of the image acquisition equipment; Image acquisition unit: used to acquire the initial image of the target contract using the installed image acquisition device.

3. A digital intelligent contract recognition system according to claim 1, characterized in that: The data acquisition module includes: A pre-processing unit is configured to enhance the initial image to obtain a clear image in an electronic copy format, and adjust the clear image according to a preset standard ratio to obtain a first image; Correction unit: performs horizontal correction processing on the text portion of the first image, and uses a blank area recognition algorithm to crop the area to be cropped, thereby obtaining a second image; Compression unit: compresses the second image to a standard image size using a compression tool to obtain a standard image of the target contract.

4. A digital intelligent contract recognition system according to claim 1, characterized in that: The text processing module includes: Cleaning unit: traverses the original text data, marks unrecognized words in the original text data as target words, and compares the target words with the standard dictionary; If a standard similar word exists in the target word, the target word is converted into a standard similar word with the highest similarity; otherwise, the target word is determined to be an unknown word and deleted, and the first contract data is output after traversal. Other format unit: Filters and converts unstructured data in the original text data in combination with a preset target text format, and outputs the second contract data; The text proofreading unit corrects misspelled words in the first contract data and the second contract data based on a proofreader to obtain standard text data.

5. The digital intelligent contract recognition system according to claim 1 is characterized in that: The text analysis module includes: Word segmentation unit: used to segment standard text data into word sequences, and filter out useless stop words based on a general word dictionary to obtain a standard word sequence; Entity recognition unit: Based on the contract entity library, the named entity recognition system is used to identify entities in standard word sequences and obtain entity sequences; Keyword extraction unit: Based on the contract keyword recognition library, perform keyword extraction on the standard word sequence to obtain a keyword sequence; Judgment unit: Take the union of the entity sequence and the keyword sequence, filter out the useless words in the union, and obtain the key contract data.

6. A digital intelligent contract recognition system according to claim 1, characterized in that: The text analysis module includes: Syntax extraction unit: performs grammatical analysis on key contract data, determines the relationship and hierarchical structure between terms, and adds grammatical terms in a grammatical logical order to generate contract terms; Structured data unit: standardizes the contract terms according to the format of the standard preset contract structure to obtain the formatted contract data of the target contract.

7. A method for digital intelligent contract identification, characterized in that: include: Step 1: Use an image acquisition device to acquire an initial image of the target contract, and pre-process the initial image to obtain a standard image in the electronic copy format; Step 2 uses an OCR tool to recognize the image text content of the standard image and converts the image text content into original text data, including: after the standard image is loaded using the OCR tool, the features of the standard image are extracted through a recognition algorithm to generate a first feature image; the first feature image is rotated at 8 angles to obtain directional feature images of 8 angles, and the features of the 8 directional feature images are extracted to obtain 8 directional feature values; each directional feature value is converted into a corresponding directional vector according to a vector conversion standard, and at the same time, the feature vector of the first feature image is used as a center vector; a covariance matrix is ​​constructed based on all directional vectors and the center vector; a mixed feature vector is obtained according to the covariance matrix, and the mixed feature vector is projected onto each Obtaining projection coefficients on the direction vectors, and using the corresponding projection coefficients as weight coefficients of the corresponding direction feature images; performing feature fusion on the eight direction vectors based on the weight coefficients of the direction feature images to obtain a fused feature vector, wherein the fused feature vector includes a number of element features, and each element feature includes a horizontal sub-feature and a vertical sub-feature; matching the fused feature vector based on a feature-text library to obtain text features corresponding to each element feature in the fused feature vector; calculating the similarity between each element feature and the corresponding matched text feature; and matching the corresponding text conversion scheme from a similarity-text conversion mapping table based on the similarity of each element feature in the fused feature vector to convert the corresponding image text content into original text data; Step 3: Perform text processing on the original text data to obtain the standard text data of the target contract; Step 4: Use natural language recognition technology to analyze standard text data, extract key contract data, and integrate the key contract data into formatted contract data with a standard structure.

Images