An intelligent resume OCR parsing and automatic information extraction system and method based on deep learning

Abstract

The present invention discloses a system and method for intelligent resume OCR parsing and automatic information extraction based on deep learning, which specifically relates to the field of automated information technology, including a resume acquisition module, a resume preprocessing module, a resume text acquisition module, a resume text analysis module, a comprehensive resume analysis module, a discrimination module, and a feedback module. The present invention obtains the first resume text sent by the user, performs a preprocessing operation to obtain the first text, and then divides it into various sub-texts; analyzes and processes various types of data collected in the sub-texts, and then performs a comprehensive analysis through a mathematical model to calculate the information accuracy index of the target text, which helps to present a more intuitive effect, facilitates decision-making, and ultimately improves the comprehensiveness and security of the automatic information extraction of the intelligent resume.

Description

Technical Field

[0001] The present invention relates to the field of automated information technology, and more specifically, to a deep learning-based intelligent resume OCR parsing and automated information extraction system and method. Background Art

[0002] In today's era of rapid digital development, OCR analysis and automatic information extraction technology have emerged as a key means to improve information processing efficiency; the emergence of OCR (Optical Character Recognition) analysis technology stems from the need to quickly convert text information in a large number of paper documents and images into editable electronic text, and automatic information extraction is a further development based on OCR analysis. When faced with massive amounts of text data, simple OCR conversion is only the first step. Automatic information extraction technology can intelligently extract key information from text after OCR analysis. Therefore, OCR analysis and automatic information extraction technology are widely used in corporate document management, data entry, customer information processing, etc. It can quickly convert a large number of paper contracts, invoices, resumes and other documents into electronic data, and automatically extract key information for storage and analysis, thereby improving the company's operational efficiency and decision-making accuracy.

[0003] An existing intelligent resume OCR parsing and automatic information extraction system based on deep learning includes a resume data collection module, a resume data analysis module and a resume data evaluation module. The resume data collection module mainly collects relevant data of resumes through various collection tools; the resume data analysis module converts complex resumes into intuitive data through automatic recognition, making the system analysis clearer. The resume data evaluation module accurately evaluates the existing analysis results by comparing historical resume data.

[0004] However, it still has some shortcomings in actual use. For example, the resume data collection module collects biological sample data through various collection tools, which is easily affected by environmental factors and reduces the accuracy of the collected data; the resume data analysis module processes the collected resume data, but when the existing technology analyzes a large amount of resume data, there will be deviations in the analysis results and missing data, resulting in data gaps, which reduces the practical value of the resume data; the existing system lacks comprehensive data to measure the key indicators of resume reading, is not comprehensive enough, increases the difficulty and risk of decision-making, and brings troubles to resume reading. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide an intelligent resume OCR parsing and automatic information extraction system and system based on deep learning, which solves the problems raised in the above-mentioned background technology through the field of automated information technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an intelligent resume OCR parsing and automatic information extraction system based on deep learning, comprising:

[0007] A resume acquisition module, in response to the automatic identification operation of the browser target plug-in channel, acquires the first resume text sent by the user;

[0008] A resume preprocessing module performs a preprocessing operation on the first resume text sent by the user, obtains a first text corresponding to the first resume text, determines the first text as a target text, divides the target text into sub-texts, and numbers the sub-texts in sequence as i=1, 2, ..., n;

[0009] The resume text collection module is used to collect various types of data from any sub-text in the target text segmentation, and transmit the collected data to the resume text analysis module; the collected data includes image color feature data, image recognition data, brightness recognition feature data and recognition time consumption data;

[0010] The resume text analysis module is used to analyze various types of data collected by the resume text collection module and transmit the analysis results to the comprehensive resume analysis module; it includes an image color feature data calculation unit, an image recognition data calculation unit, a brightness recognition feature data calculation unit, and a recognition time-consuming data calculation unit;

[0011] Comprehensive resume analysis module: It is used to import the data transmitted by the resume text analysis module into the comprehensive resume analysis module, calculate the information accuracy index of the target text, and transmit it to the discrimination module; this module includes a comprehensive resume calculation unit;

[0012] Discrimination module: used to compare the information accuracy index of the target text with the information accuracy index of the preset target text in the actual process and transmit the judgment result to the feedback module;

[0013] The feedback module is used to send the signal sent by the discrimination module to the corresponding management personnel in a feedback manner.

[0014] Preferably, the specific operation of resume automatic recognition operation is:

[0015] A1: Plug-in design: The target plug-in is specially developed for the paid recruitment platform in the browser and is compatible with mainstream browsers. After the plug-in is installed, it will automatically run in the background to monitor the user's operations on the recruitment platform in real time. When the user browses to the target's resume page, the plug-in will automatically trigger the screenshot function and save the resume page to the database.

[0016] A2: Paid recruitment platform: The system's plug-in can support multiple paid recruitment platforms, including domestic recruitment websites and foreign talent recruitment platforms;

[0017] A3: Screenshot saving and recognition: After the screenshot is taken, the plug-in will automatically save the screenshot file to the PC storage database; users can set the save path and file format according to their needs to facilitate subsequent management and viewing; the system will automatically recognize and process the saved screenshot file;

[0018] A4: Through advanced OCR technology, the system can accurately recognize the text information in the screenshot and convert it into an editable electronic text format, which is the first resume text sent by the user.

[0019] Preferably, the image color feature data include the high saturation distribution rate in the image, the cold-toned character recognition accuracy, and the warm-toned character recognition error rate, which are denoted as Tb, Tl, and Tn, respectively; the image recognition data include the number of correctly recognized characters in high-resolution images, the recognition accuracy of specific fonts in clear images, and the recognition speed of noise-free images, which are denoted as Sg, Sq, and Sd, respectively; the brightness recognition feature data include the number of recognized characters in over-bright images, the increase in recognition time for over-dark images, and the recognition accuracy of special characters, which are denoted as Lt, Lz, and Lq, respectively; the recognition time data include the speed of single-page character recognition, the time difference value of text recognition in different formats, and the special character recognition time, which are denoted as Bs, Bc, and Bt, respectively.

[0020] Preferably, the image color feature data calculation unit is used to establish a mathematical model of the image color feature data, import the image color feature data collected by the resume text collection module into the mathematical model of the image color feature data, and calculate the image color coefficient value of each sub-text, which is specifically expressed as:

[0021]

[0022] TA i Indicates the image color coefficient value of the i-th subtext, Tl i represents the cold tone character recognition accuracy of the i-th subtext, Tb i represents the high saturation distribution rate in the image of the i-th subtext, Tn i represents the warm-tone character recognition error rate of the i-th subtext, and n represents the number of subtexts.

[0023] Preferably, the image recognition data calculation unit is used to establish a mathematical model of image recognition data, import the image recognition data collected by the resume text collection module into the mathematical model of image recognition data, and calculate the image recognition coefficient value of each sub-text, which is specifically expressed as:

[0024]

[0025] S i Represents the image recognition coefficient value of the i-th subtext, Sg i-1 represents the number of correctly recognized characters in the high-resolution image of the i-1th subtext, Sg i represents the number of correctly recognized characters in the high-resolution image of the i-th subtext, Sq i represents the recognition accuracy of a specific font in the clear image of the i-th subtext, Sd i represents the noise-free image recognition speed of the i-th subtext, and n represents the number of subtexts.

[0026] Preferably, the brightness recognition feature data calculation unit is used to establish a mathematical model of brightness recognition feature data, import the brightness recognition feature data collected by the resume text collection module into the mathematical model of brightness recognition feature data, and calculate the brightness recognition coefficient value of each sub-text, which is specifically expressed as:

[0027]

[0028] LA i Indicates the brightness recognition coefficient value of the i-th subtext, Lt i represents the number of over-brightness image recognition characters of the i-th subtext, Lz i represents the increase in time consumption for dark image recognition of the i-th subtext, Lq i represents the special character recognition accuracy of the i-th subtext, and n represents the number of subtexts.

[0029] Preferably, the recognition time-consuming data calculation unit is used to establish a mathematical model for recognition time-consuming data, import the recognition time-consuming data collected by the resume text collection module into the mathematical model for recognition time-consuming data, and calculate the recognition time-consuming coefficient value of each sub-text, which is specifically expressed as:

[0030]

[0031] BA i Indicates the recognition time coefficient value of the i-th subtext, Bs i represents the speed of single-page character recognition of the i-th subtext, Bs max Indicates the maximum speed of single-page character recognition of target text, Bc iIndicates the time difference of text recognition in different formats of the i-th subtext, Bt i represents the special character recognition time of the i-th subtext, and n represents the number of subtexts.

[0032] Preferably, the comprehensive resume calculation unit is used to establish a mathematical model of the target text, import the coefficient values ​​calculated by the resume text analysis module into the mathematical model of the target area, and calculate the information accuracy index of the target text, which is specifically expressed as:

[0033]

[0034] [α i =TA i ×λ1+SA i ×λ2+LA i ×λ3+BA i ×λ4],

[0035] Where η represents the information accuracy index of the target text, α i Represents the information accuracy index of the i-th subtext, TA i Represents the image color coefficient value of the i-th subtext, SA i Represents the image recognition coefficient value of the i-th subtext, LA i Indicates the brightness recognition coefficient value of the i-th subtext, BA i represents the recognition time coefficient value of the i-th sub-text, λ1, λ2, λ3, λ4 represent the weight coefficients of the information accuracy index, λ1 ranges from 0.15 to 0.25, λ2 ranges from 0.20 to 0.24, λ3 ranges from 0.15 to 0.26, λ4 ranges from 0.14 to 0.23, λ5 ranges from 0.14 to 0.22, where β represents other influencing factors of the information accuracy index of the target text, where i = 1, 2, …, n.

[0036] Preferably, the information accuracy index of the preset target text is denoted as η def , the information accuracy index of the preset target text is the value recorded based on the historical data evaluation of previous years; when η<η def When η>η, the system sends a good signal; when η>η def When the system sends an abnormal signal.

[0037] S1: Obtaining a resume, in response to the user's resume automatic recognition operation, obtaining a first resume text sent by the user;

[0038] S2: resume preprocessing, performing a preprocessing operation on the first resume text sent by the user, obtaining a first text corresponding to the first resume text, determining the first text as a target text, dividing the target text into sub-texts, and numbering the sub-texts in sequence as i=1, 2, ..., n;

[0039] S3: Resume text collection, which is used to collect various types of data from any sub-text in the target text segmentation, and transmit the collected data to the resume text analysis step; the collected data includes image color feature data, image recognition data, brightness recognition feature data, and recognition time consumption data;

[0040] S4: resume text analysis, used to analyze various types of data collected in the resume text collection step, and transmit the analysis results to the comprehensive resume analysis step; including an image color feature data calculation unit, an image recognition data calculation unit, a brightness recognition feature data calculation unit, and a recognition time-consuming data calculation unit;

[0041] S5: Comprehensive resume analysis, the data transmitted from the resume text analysis step is imported into the comprehensive resume analysis step, the information accuracy index of the target text is calculated, and transmitted to the discrimination step; this step includes a comprehensive resume calculation unit;

[0042] S6: a judgment step: the information accuracy index for the target text is compared with the information accuracy index of the preset target text in the actual process and the judgment result is transmitted to the feedback step;

[0043] S7: Feedback step, which is used to send the signal sent in the determination step to the corresponding management personnel in a feedback manner.

[0044] Technical effects and advantages of the present invention:

[0045] 1. The present invention obtains a first resume text sent by a user, performs a preprocessing operation on the first resume text sent by the user, obtains a first text corresponding to the first resume text, determines the first text as a target text, divides the target text into sub-texts, and collects various types of data from any sub-text in the target text division. The collected data includes image color feature data, image recognition data, brightness recognition feature data, and recognition time-consuming data. Data collection helps to improve the comprehensiveness of intelligent resume automated analysis;

[0046] 2. The present invention obtains various coefficient values ​​by analyzing various types of collected data, thereby solving the problems existing in the automatic information extraction process of intelligent resumes, which helps to provide an important basis for the subsequent system to realize the automatic information extraction and analysis of intelligent resumes; through the comprehensive resume analysis module, the various coefficient values ​​analyzed are comprehensively analyzed through mathematical models, and the information accuracy index of the target text is calculated, which helps to present a more intuitive effect and facilitate decision-making;

[0047] 3. The present invention helps to ensure the accuracy of the system by comparing the information accuracy index of the target text with the information accuracy index of the preset target text; then the abnormal signal sent by the system can be discovered in time through the feedback module, and the system will feed back the processing results to the management personnel. The present invention improves the comprehensiveness and security of the automatic information extraction of the intelligent resume. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of the system structure of the present invention.

[0049] Figure 2 It is a schematic diagram of the structure of the method of the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] As attached Figure 1 Shown in the figure is an intelligent resume OCR parsing and automatic information extraction system based on deep learning, including a resume acquisition module, a resume preprocessing module, a resume text collection module, a resume text analysis module, a comprehensive resume analysis module, a discrimination module and a feedback module.

[0052] The resume acquisition module acquires the first resume text sent by the user in response to the automatic identification operation of the browser target plug-in channel.

[0053] In this embodiment, it should be specifically explained that the resume automatic recognition operation specifically operates as follows:

[0054] A1: Plug-in design: The target plug-in is specially developed for the paid recruitment platform in the browser and is compatible with mainstream browsers. After the plug-in is installed, it will automatically run in the background to monitor the user's operations on the recruitment platform in real time. When the user browses to the target's resume page, the plug-in will automatically trigger the screenshot function and save the resume page to the database.

[0055] A2: Paid recruitment platforms: The system's plugin can support multiple paid recruitment platforms, including domestic recruitment websites and foreign talent recruitment platforms;

[0056] A3: Screenshot saving and recognition: After the screenshot is completed, the plugin will automatically save the screenshot file to the PC-side storage database; users can set the save path and file format according to their own needs for convenient subsequent management and viewing; the system will automatically recognize and process the saved screenshot file;

[0057] A4: Through advanced OCR technology, the system can accurately recognize the text information in the screenshot and convert it into an editable electronic text format, that is, the first resume text sent by the user.

[0058] The resume preprocessing module performs preprocessing operations on the first resume text sent by the user, obtains the first text corresponding to the first resume text, determines the first text as the target text, divides the target text into each sub-text, and sequentially numbers each sub-text as i = 1, 2,..., n.

[0059] In this embodiment, specifically, the specific steps of the preprocessing operation are as follows:

[0060] B1: Text cleaning: Remove redundant information and noise, remove interference information such as special characters, punctuation marks, and numbers; convert the case, unify the text to uppercase or lowercase for subsequent processing;

[0061] B2: Sub-text division: Determine the first text as the target text and divide it into each sub-text, and sequentially number each sub-text;

[0062] B3: Stop-word removal: Remove common words that contribute less to text analysis from each sub-text, such as "of", "is", "in", etc.

[0063] The resume text collection module is used to collect various types of data from any sub-text in the target text division and transmit the collected various types of data to the resume text analysis module; the collected data includes image color feature data, image recognition data, brightness recognition feature data, and recognition time-consuming data;

[0064] In this embodiment, it is specifically necessary to explain that the image color feature data includes the high saturation distribution rate in the image, the cold-toned character recognition accuracy, and the warm-toned character recognition error rate, which are respectively denoted as Tb, Tl, and Tn; the image recognition data includes the number of correctly recognized characters in high-resolution images, the recognition accuracy of specific fonts in clear images, and the recognition speed of noise-free images, which are respectively denoted as Sg, Sq, and Sd; the brightness recognition feature data includes the number of recognized characters in over-bright images, the increase in the time consumption for over-dark image recognition, and the accuracy rate of special character recognition, which are respectively denoted as Lt, Lz, and Lq; the recognition time consumption data includes the speed of single-page character recognition, the time consumption difference value for recognizing texts in different formats, and the special character recognition time, which are respectively denoted as Bs, Bc, and Bt.

[0065] For the high saturation distribution rate in the image, the image analysis software is used to perform color analysis on the resume image, the number of high saturation pixels is counted and the ratio is calculated, and the software is installed on the server side. After the resume image is uploaded, the analysis and processing are performed. The accuracy of cold-toned character recognition and the error rate of warm-toned character recognition are performed by using OCR software to recognize characters of resume images of different tones, and the accuracy and error rate are calculated by comparing with the real text. The OCR software is installed on the server side to receive the image and analyze it. The number of correctly recognized characters in high-resolution images and the accuracy of specific font recognition in clear images are prepared. Resume image samples with different resolutions and clarity are used for recognition, statistics and calculation by OCR software. It is also installed on the server side to process the uploaded high-resolution and clear images. The speed of noise-free image recognition uses a timer to record the recognition time of the OCR software for noise-free images, and the OCR software and timer are run on the server side. The number of characters recognized by over-bright images, the increase in time consumed by over-dark image recognition, and the accuracy of special character recognition are determined by adjusting the image brightness on the server side for OCR recognition and statistics of related data. The speed of single-page character recognition, the time difference of text recognition in different formats, and the time for special character recognition are recorded and statistically analyzed using a timer installed on the server.

[0066] The resume text analysis module is used to analyze various types of data collected by the resume text collection module and transmit the analysis results to the comprehensive resume analysis module; it includes an image color feature data calculation unit, an image recognition data calculation unit, a brightness recognition feature data calculation unit and an recognition time-consuming data calculation unit.

[0067] In this embodiment, it should be specifically explained that the image color feature data calculation unit is used to establish a mathematical model of the image color feature data, import the image color feature data collected by the resume text collection module into the mathematical model of the image color feature data, and calculate the image color coefficient value of each sub-text, which is specifically expressed as:

[0068]

[0069] TA i Indicates the image color coefficient value of the i-th subtext, Tl i represents the cold tone character recognition accuracy of the i-th subtext, Tb i represents the high saturation distribution rate in the image of the i-th subtext, Tn i represents the warm-tone character recognition error rate of the i-th subtext, and n represents the number of subtexts.

[0070] In this embodiment, it should be specifically explained that the image recognition data calculation unit is used to establish a mathematical model of image recognition data, import the image recognition data collected by the resume text collection module into the mathematical model of image recognition data, and calculate the image recognition coefficient value of each sub-text, which is specifically expressed as:

[0071]

[0072] S i Represents the image recognition coefficient value of the i-th subtext, Sg i-1 represents the number of correctly recognized characters in the high-resolution image of the i-1th subtext, Sg i represents the number of correctly recognized characters in the high-resolution image of the i-th subtext, Sq i represents the recognition accuracy of a specific font in the clear image of the i-th subtext, Sd i represents the noise-free image recognition speed of the i-th subtext, and n represents the number of subtexts.

[0073] In this embodiment, it should be specifically explained that the brightness recognition feature data calculation unit is used to establish a mathematical model of brightness recognition feature data, import the brightness recognition feature data collected by the resume text collection module into the mathematical model of brightness recognition feature data, and calculate the brightness recognition coefficient value of each sub-text, which is specifically expressed as:

[0074]

[0075] LA i Indicates the brightness recognition coefficient value of the i-th subtext, Lt i represents the number of over-brightness image recognition characters of the i-th subtext, Lz i represents the increase in time consumption for dark image recognition of the i-th subtext, Lq i represents the special character recognition accuracy of the i-th subtext, and n represents the number of subtexts.

[0076] In this embodiment, it should be specifically explained that the recognition time-consuming data calculation unit is used to establish a mathematical model for recognition time-consuming data, import the recognition time-consuming data collected by the resume text collection module into the mathematical model for recognition time-consuming data, and calculate the recognition time-consuming coefficient value of each sub-text, which is specifically expressed as:

[0077]

[0078] BA i Indicates the recognition time coefficient value of the i-th subtext, Bs i represents the speed of single-page character recognition of the i-th subtext, Bs max Indicates the maximum speed of single-page character recognition of target text, Bc i Indicates the time difference of text recognition in different formats of the i-th subtext, Bt i represents the special character recognition time of the i-th subtext, and n represents the number of subtexts.

[0079] Comprehensive resume analysis module: The data transmitted by the resume text analysis module is imported into the comprehensive resume analysis module, the information accuracy index of the target text is calculated, and transmitted to the discrimination module; this module includes a comprehensive resume calculation unit.

[0080] In this embodiment, it should be specifically explained that the comprehensive resume calculation unit is used to establish a mathematical model of the target text, import the coefficient values ​​calculated by the resume text analysis module into the mathematical model of the target area, and calculate the information accuracy index of the target text, which is specifically expressed as:

[0081]

[0082] [α i =TA i ×λ1+SA i ×λ2+LA i ×λ3+BA i ×λ4],

[0083] Where η represents the information accuracy index of the target text, α i Represents the information accuracy index of the i-th subtext, TA i Represents the image color coefficient value of the i-th subtext, SA i Represents the image recognition coefficient value of the i-th subtext, LA i Indicates the brightness recognition coefficient value of the i-th subtext, BA irepresents the recognition time coefficient value of the i-th sub-text, λ1, λ2, λ3, λ4 represent the weight coefficients of the information accuracy index, λ1 ranges from 0.15 to 0.25, λ2 ranges from 0.20 to 0.24, λ3 ranges from 0.15 to 0.26, λ4 ranges from 0.14 to 0.23, λ5 ranges from 0.14 to 0.22, where β represents other influencing factors of the information accuracy index of the target text, where i = 1, 2, …, n.

[0084] Discrimination module: compares the information accuracy index of the target text with the information accuracy index of the preset target text in the actual process and transmits the judgment result to the feedback module.

[0085] In this embodiment, it should be specifically explained that the information accuracy index of the preset target text is denoted as η def , the information accuracy index of the preset target text is the value recorded based on the historical data evaluation of previous years; when η<η def When η>η, it means that the information accuracy index of the target text is less than the information accuracy index of the preset target text, indicating that the resume reading is in a good signal; when η>η def When , it means that the biological microscope rational index of the target area is greater than the biological microscope rational index of the preset target area, indicating that the reading resume is in an abnormal signal, and the system will send a signal to the feedback module.

[0086] The feedback module is used to send the signal sent by the discrimination module to the corresponding management personnel in a feedback manner.

[0087] In this embodiment, it is specifically necessary to explain that the feedback modes include: text display, image display and table display.

[0088] As attached Figure 2 As shown, this embodiment provides a multifunctional biological microscope interaction method based on wireless interaction, comprising the following steps:

[0089] S1: Obtaining a resume, in response to the automatic identification operation of the browser target plug-in channel, obtaining a first resume text sent by the user;

[0090] S2: resume preprocessing, performing a preprocessing operation on the first resume text sent by the user, obtaining a first text corresponding to the first resume text, determining the first text as a target text, dividing the target text into sub-texts, and numbering the sub-texts in sequence as i=1, 2, ..., n;

[0091] S3: Resume text collection, which is used to collect various types of data from any sub-text in the target text segmentation, and transmit the collected data to the resume text analysis step; the collected data includes image color feature data, image recognition data, brightness recognition feature data, and recognition time consumption data;

[0092] S4: resume text analysis, used to analyze various types of data collected in the resume text collection step, and transmit the analysis results to the comprehensive resume analysis step; including an image color feature data calculation unit, an image recognition data calculation unit, a brightness recognition feature data calculation unit, and a recognition time-consuming data calculation unit;

[0093] S5: Comprehensive resume analysis, the data transmitted from the resume text analysis step is imported into the comprehensive resume analysis step, the information accuracy index of the target text is calculated, and transmitted to the discrimination step; this step includes a comprehensive resume calculation unit;

[0094] S6: a judgment step: the information accuracy index for the target text is compared with the information accuracy index of the preset target text in the actual process and the judgment result is transmitted to the feedback step;

[0095] S7: Feedback step, which is used to send the signal sent in the determination step to the corresponding management personnel in a feedback manner.

[0096] The present invention obtains a first resume text sent by a user, performs a preprocessing operation on the first resume text sent by the user, obtains a first text corresponding to the first resume text, determines the first text as a target text, divides the target text into sub-texts, and sequentially numbers the sub-texts as i=1, 2, ..., n; collects various types of data from any sub-text in the target text division, the collected data includes image color feature data, image recognition data, brightness recognition feature data and recognition time-consuming data, and through data collection, it is helpful to improve the comprehensiveness of intelligent resume automatic analysis; then, analyzes the collected various types of data to obtain various coefficient values, so as to analyze the automatic information of the intelligent resume. The problems existing in the extraction process are helpful to provide an important basis for the subsequent system to realize the automatic information extraction and analysis of intelligent resumes; through the comprehensive resume analysis module, the various coefficient values ​​​​of the analysis are comprehensively analyzed through a mathematical model, and the information accuracy index of the target text is calculated, which helps to present a more intuitive effect and facilitate decision-making; by comparing the information accuracy index of the target text with the information accuracy index of the preset target text, it helps to ensure the accuracy of the system; the feedback module can promptly detect abnormal signals sent by the system, and once the signal is received, the system will feedback the processing results to the management personnel; the present invention improves the comprehensiveness and security of the automatic information extraction of intelligent resumes.

[0097] Secondly, in the drawings of the embodiments disclosed in the present invention, only the structures related to the embodiments disclosed in the present invention are involved, and other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of the present invention can be combined with each other;

[0098] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent resume OCR parsing and automatic information extraction system based on deep learning, characterized in that: include: Resume acquisition module: in response to the automatic identification operation of the browser target plug-in channel, acquire the first resume text sent by the user; Resume preprocessing module: preprocessing the first resume text sent by the user, obtaining the first text corresponding to the first resume text, determining the first text as the target text, dividing the target text into sub-texts, and numbering the sub-texts in sequence as i=1, 2, ..., n; Resume text collection module: used to collect various types of data from any sub-text in the target text segmentation, and transmit the collected data to the resume text analysis module; the collected data includes image color feature data, image recognition data, brightness recognition feature data and recognition time consumption data; Resume text analysis module: used to analyze various data collected by the resume text collection module and transmit the analysis results to the comprehensive resume analysis module; including image color feature data calculation unit, image recognition data calculation unit, brightness recognition feature data calculation unit and recognition time-consuming data calculation unit; The image color feature data calculation unit is used to establish a mathematical model of image color feature data, import the image color feature data collected by the resume text collection module into the mathematical model of image color feature data, and calculate the image color coefficient value of each sub-text, which is specifically expressed as: TA i Indicates the image color coefficient value of the i-th subtext, Tl i represents the cold tone character recognition accuracy of the i-th subtext, Tb i represents the high saturation distribution rate in the image of the i-th subtext, Tn i represents the warm-tone character recognition error rate of the i-th subtext, and n represents the number of subtexts; The image recognition data calculation unit is used to establish a mathematical model of image recognition data, import the image recognition data collected by the resume text collection module into the mathematical model of image recognition data, and calculate the image recognition coefficient value of each sub-text, which is specifically expressed as: S i Represents the image recognition coefficient value of the i-th subtext, Sg i-1 represents the number of correctly recognized characters in the high-resolution image of the i-1th subtext, Sg i represents the number of correctly recognized characters in the high-resolution image of the i-th subtext, Sq i represents the recognition accuracy of a specific font in the clear image of the i-th subtext, Sd i represents the noise-free image recognition speed of the i-th subtext, and n represents the number of subtexts; The brightness recognition feature data calculation unit is used to establish a mathematical model of brightness recognition feature data, import the brightness recognition feature data collected by the resume text collection module into the mathematical model of brightness recognition feature data, and calculate the brightness recognition coefficient value of each sub-text, which is specifically expressed as: LA i Indicates the brightness recognition coefficient value of the i-th subtext, Lt i represents the number of over-brightness image recognition characters of the i-th subtext, Lz i represents the increase in time consumption for dark image recognition of the i-th subtext, Lq i represents the special character recognition accuracy of the i-th subtext, and n represents the number of subtexts; The recognition time-consuming data calculation unit is used to establish a mathematical model for recognition time-consuming data, import the recognition time-consuming data collected by the resume text collection module into the mathematical model for recognition time-consuming data, and calculate the recognition time-consuming coefficient value of each sub-text, which is specifically expressed as: BA i Indicates the recognition time coefficient value of the i-th subtext, Bs i represents the speed of single-page character recognition of the i-th subtext, Bs max Indicates the maximum speed of single-page character recognition of target text, Bc i Indicates the time difference of text recognition in different formats of the i-th subtext, Bt i represents the special character recognition time of the i-th subtext, and n represents the number of subtexts; Comprehensive resume analysis module: It is used to import the data transmitted by the resume text analysis module into the comprehensive resume analysis module, calculate the information accuracy index of the target text, and transmit it to the discrimination module; this module includes a comprehensive resume calculation unit; The comprehensive resume calculation unit is used to establish a mathematical model of the target text, import the coefficient values ​​calculated by the resume text analysis module into the mathematical model of the target area, and calculate the information accuracy index of the target text, which is specifically expressed as: [α i =TA i ×λ1+SA i ×λ2+LA i ×λ3+BA i ×λ4], Where η represents the information accuracy index of the target text, α i Represents the information accuracy index of the i-th subtext, TA i Represents the image color coefficient value of the i-th subtext, SA i Represents the image recognition coefficient value of the i-th subtext, LA i Indicates the brightness recognition coefficient value of the i-th subtext, BA i represents the recognition time coefficient value of the i-th subtext, λ1, λ2, λ3, λ4 represent the weight coefficients of the information accuracy index, λ1 ranges from 0.15 to 0.25, λ2 ranges from 0.20 to 0.24, λ3 ranges from 0.15 to 0.26, λ4 ranges from 0.14 to 0.23, and λ5 ranges from 0.14 to 0.22, where β represents other influencing factors of the information accuracy index of the target text, where i = 1, 2, ..., n; Discrimination module: used to compare the information accuracy index of the target text with the information accuracy index of the preset target text in the actual process and transmit the judgment result to the feedback module; The feedback module is used to send the signal sent by the discrimination module to the corresponding management personnel in a feedback manner.

2. According to claim 1, a deep learning-based intelligent resume OCR parsing and automatic information extraction system is characterized by: The specific operation of the resume automatic recognition operation is as follows: A1: Plug-in design: The target plug-in is specially developed for the paid recruitment platform in the browser and is compatible with mainstream browsers. After the plug-in is installed, it will automatically run in the background to monitor the user's operations on the recruitment platform in real time. When the user browses to the target's resume page, the plug-in will automatically trigger the screenshot function and save the resume page to the database. A2: Paid recruitment platform: The system's plug-in can support multiple paid recruitment platforms, including domestic recruitment websites and foreign talent recruitment platforms; A3: Screenshot saving and recognition: After the screenshot is taken, the plug-in will automatically save the screenshot file to the PC storage database; users can set the save path and file format according to their needs to facilitate subsequent management and viewing; the system will automatically recognize and process the saved screenshot file; A4: Through advanced OCR technology, the system can accurately recognize the text information in the screenshot and convert it into an editable electronic text format, which is the first resume text sent by the user.

3. According to claim 1, a deep learning-based intelligent resume OCR parsing and automatic information extraction system is characterized by: The image color feature data includes the high saturation distribution rate in the image, the recognition accuracy of cold-toned characters, and the recognition error rate of warm-toned characters, which are denoted as Tb, Tl, and Tn, respectively; the image recognition data includes the number of correctly recognized characters in high-resolution images, the recognition accuracy of specific fonts in clear images, and the recognition speed of noise-free images, which are denoted as Sg, Sq, and Sd, respectively; the brightness recognition feature data includes the number of recognized characters in over-bright images, the increase in recognition time for over-dark images, and the recognition accuracy of special characters, which are denoted as Lt, Lz, and Lq, respectively; the recognition time data includes the speed of single-page character recognition, the time difference value of text recognition in different formats, and the recognition time of special characters, which are denoted as Bs, Bc, and Bt, respectively.

4. According to claim 1, a deep learning-based intelligent resume OCR parsing and automatic information extraction system is characterized by: The information accuracy index of the preset target text is denoted as η def , the information accuracy index of the preset target text is the value recorded based on the historical data evaluation of previous years; when η<η def When η>η, the system sends a good signal; when η>η def When the system sends an abnormal signal.

5. A method for intelligent resume OCR parsing and automatic information extraction based on deep learning, characterized by: The method for implementing the deep learning-based intelligent resume OCR parsing and automatic information extraction system described in any one of claims 1 to 4 comprises the following steps: S1: Obtaining a resume, in response to the automatic identification operation of the browser target plug-in channel, obtaining a first resume text sent by the user; S2: resume preprocessing, performing a preprocessing operation on the first resume text sent by the user, obtaining a first text corresponding to the first resume text, determining the first text as a target text, dividing the target text into sub-texts, and numbering the sub-texts in sequence as i=1, 2, ..., n; S3: Resume text collection, which is used to collect various types of data from any sub-text in the target text segmentation, and transmit the collected data to the resume text analysis step; the collected data includes image color feature data, image recognition data, brightness recognition feature data, and recognition time consumption data; S4: resume text analysis, used to analyze various types of data collected in the resume text collection step, and transmit the analysis results to the comprehensive resume analysis step; including an image color feature data calculation unit, an image recognition data calculation unit, a brightness recognition feature data calculation unit, and a recognition time-consuming data calculation unit; S5: Comprehensive resume analysis, the data transmitted from the resume text analysis step is imported into the comprehensive resume analysis step, the information accuracy index of the target text is calculated, and transmitted to the discrimination step; this step includes a comprehensive resume calculation unit; S6: a judgment step: the information accuracy index for the target text is compared with the information accuracy index of the preset target text in the actual process and the judgment result is transmitted to the feedback step; S7: Feedback step, which is used to send the signal sent in the determination step to the corresponding management personnel in a feedback manner.

Images