CN112560849A - Neural network algorithm-based grammar segmentation method and system - Google Patents
Neural network algorithm-based grammar segmentation method and system Download PDFInfo
- Publication number
- CN112560849A CN112560849A CN202110092477.7A CN202110092477A CN112560849A CN 112560849 A CN112560849 A CN 112560849A CN 202110092477 A CN202110092477 A CN 202110092477A CN 112560849 A CN112560849 A CN 112560849A
- Authority
- CN
- China
- Prior art keywords
- target
- test
- question
- image
- test question
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/24—Aligning, centring, orientation detection or correction of the image
- G06V10/243—Aligning, centring, orientation detection or correction of the image by compensating for image skew or non-uniform image deformations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Character Input (AREA)
- Image Analysis (AREA)
Abstract
The application discloses a method and a system for segmenting a grammar based on a neural network algorithm, wherein the method comprises the following steps: acquiring a target test paper image, wherein the target test paper image is a scanned image of a comprehensive text test paper or a comprehensive test paper; inputting the target test paper image into a target neural network model for identification to obtain an identification result of each test question in the target test paper image; classifying each test question based on the identification result to obtain a plurality of target test question sets; according to the mapping relation between the test question set and the hierarchical examination subjects, the test questions in the target test question set are stored in the memory block of the target hierarchical examination subjects corresponding to the target test question set, so that each test question obtained after the traditional literature test paper is divided can be stored in the memory block of the corresponding hierarchical examination subjects, and test question resources are provided for automatically generating the test paper under the 'new high-level examination' model.
Description
Technical Field
The application relates to the technical field of computers, in particular to a method and a system for dividing a grammar based on a neural network algorithm.
Background
From 2020, most areas of China gradually adopt a 'new college entrance examination' mode, according to a new college entrance examination reform scheme, the traditional text and physical departments are not divided again, and the technical branch is omitted. Some provinces adopt a '3 + 3' new high-level examination subject selection mode, except for Chinese, mathematics and foreign languages, students can select three subjects from 6 subjects of level examinations of ideology, politics, history, geography, physics, chemistry and biology, so that the students are endowed with sufficient free option and can independently determine subject combinations. Some provinces adopt a '3 +1+ 2' new high-level examination subject selection mode, and students need to select one subject from physical and historical grade examination subjects and select two subjects from the rest 4 grade examination subjects. With the change of the examination mode, the examination form, the examination subject type, the examination subject amount and the like are changed.
Since the reform scheme is just started to be executed, the existing test paper is a literal arts test paper (an integrated book) and a physical arts test paper (a integrated book) in the traditional mode, and the test paper under the model of 'new college entrance examination' is few. However, the generation of a test paper under the model of "new college entrance examination" cannot directly adopt the previous test paper, and only a problem assigning teacher can search related test questions from resources such as review data, a problem set, a test paper of past year and the like according to the current examination requirements, and then screen the test questions meeting the examination requirements according to personal experience to generate a test paper.
Therefore, how to divide the traditional test paper of the literature department into test paper under the model of "new college entrance examination" becomes a problem to be solved urgently.
Disclosure of Invention
The embodiment of the application provides a neural network algorithm-based grammar segmentation method and system, which can segment traditional grammar test paper, so that each test question can be stored in a memory block of a corresponding hierarchical test subject, and test question resources are provided for automatically generating test paper under a new college entrance examination model.
In a first aspect, an embodiment of the present application provides a method for segmenting a grammar based on a neural network algorithm, where the method includes:
acquiring a target test paper image, wherein the target test paper image is a scanned image of a comprehensive text test paper or a comprehensive test paper;
inputting the target test paper image into a target neural network model for identification to obtain an identification result of each test question in the target test paper image;
classifying the test questions of each channel based on the identification result to obtain a plurality of target test question sets;
and storing the test questions in the target test question set in a memory block of the target hierarchical examination subject corresponding to the target test question set according to the mapping relation between the test question set and the hierarchical examination subjects.
In a second aspect, an embodiment of the present application provides a grammar segmentation system based on a neural network algorithm, the system including:
the device comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring a target test paper image, and the target test paper image is a scanned image of a comprehensive text test paper or a comprehensive test paper;
the identification unit is used for inputting the target test paper image into a target neural network model for identification to obtain an identification result of each test question in the target test paper image;
the classification unit is used for classifying the test questions of each channel based on the identification result to obtain a plurality of target test question sets;
and the storage unit is used for storing the test questions in the target test question set in the memory blocks of the target hierarchical test subjects corresponding to the target test question set according to the mapping relation between the test question set and the hierarchical test subjects.
In a third aspect, an embodiment of the present application provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program for electronic data exchange, where the computer program enables a computer to perform some or all of the steps described in any one of the methods of the first aspect of the embodiments of the present application.
According to the technical scheme provided by the application, the target test paper image is obtained and is a scanned image of a comprehensive text test paper or a comprehensive test paper; inputting the target test paper image into a target neural network model for identification to obtain an identification result of each test question in the target test paper image; classifying each test question based on the identification result to obtain a plurality of target test question sets; according to the mapping relation between the test question set and the hierarchical examination subjects, the test questions in the target test question set are stored in the memory block of the target hierarchical examination subjects corresponding to the target test question set, so that each test question obtained after the traditional literature test paper is divided can be stored in the memory block of the corresponding hierarchical examination subjects, and test question resources are provided for automatically generating the test paper under the 'new high-level examination' model.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a flow chart diagram of a method for segmenting a grammar based on a neural network algorithm according to an embodiment of the present application;
FIG. 2a is a schematic diagram of a target test paper image segmentation provided in an embodiment of the present application;
FIG. 2b is a schematic diagram of a first area in a target test paper image according to an embodiment of the present application;
FIG. 2c is a schematic diagram of an image segmentation model structure provided in an embodiment of the present application;
FIG. 3 is a schematic diagram of a graph partitioning system based on a neural network algorithm according to an embodiment of the present application.
Detailed Description
In order to make the technical solutions of the present application better understood, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
The terms "first," "second," and the like in the description and claims of the present application and in the above-described drawings are used for distinguishing between different objects and not for describing a particular order. Furthermore, the terms "include" and "have," as well as any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not limited to only those steps or elements listed, but may include other steps or elements not listed or inherent to such process, method, article, or apparatus in one possible example.
Reference herein to "an embodiment" means that a particular feature, result, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is explicitly and implicitly understood by one skilled in the art that the embodiments described herein can be combined with other embodiments.
Referring to fig. 1, fig. 1 is a schematic flowchart illustrating a neural network algorithm based grammar splitting method according to an embodiment of the present application, and as shown in fig. 1, the neural network algorithm based grammar splitting method includes the following operations.
S110, obtaining a target test paper image, wherein the target test paper image is a scanned image of a comprehensive text test paper or a comprehensive test paper.
The target test paper image may be a calendar year text comprehensive test paper or a reason comprehensive test paper acquired by a scanner or electronic equipment such as a camera under a traditional college entrance examination model, or may be a stored picture or picture-formatted document of the calendar year text comprehensive test paper or the reason comprehensive test paper, which is not limited in this embodiment of the present application.
S120, inputting the target test paper image into a target neural network model for identification to obtain an identification result of each test question in the target test paper image.
In a specific implementation, before the target test paper image is identified, the target test paper image may be preprocessed to reduce useless information in the target test paper image, so as to facilitate subsequent processing. Preprocessing a target test paper image, performing binarization, noise reduction, character segmentation, normalization and other processing operations, converting a gray level image of the target test paper image into a binary image after binarization, and setting the gray level value of a pixel point on the image to be 0 or 255; denoising can be performed on the target test paper image according to the characteristics of noise; the character segmentation is to segment the characters in the target test paper image into single characters for recognition. Tilt correction is often performed if the lines of text are tilted. The normalization is to adjust the single text image to the same size, so that all the text images apply a uniform algorithm under the same specification.
Optionally, the inputting the target test paper image into a target neural network model for recognition to obtain a recognition result of each test question in the target test paper image includes:
segmenting the target test paper image from a first direction to obtain a first intermediate image and a second intermediate image, wherein the first intermediate image comprises first layout information of the target test paper image, and the second intermediate image comprises second layout information of the target test paper image; identifying first areas of the first intermediate image and the second intermediate image to obtain starting and stopping positions of each test question; according to the starting and stopping positions of each test question, the first intermediate image and the second intermediate image are respectively segmented from a second direction to obtain a plurality of target test question images; and inputting the multiple target test question images into the target neural network for text recognition to obtain a recognition result of the target test paper images.
The traditional college entrance examination paper adopts paper with the size of A3, each piece of paper comprises first layout information and second layout information, the first layout information is located on the left side of the paper, and the second layout information is located on the right side of the paper in parallel. As shown in fig. 2a, the target test paper image is divided from the first direction to obtain a first intermediate image including the first layout information and a second intermediate image including the second layout information, so that an invalid portion of the image, for example, a blank image is cut off to obtain an intermediate image including only the target text line.
The first direction may be a longitudinal axis direction of the target test paper image, that is, in the middle of the target test paper image, the target test paper image is divided into equal length from the longitudinal axis direction, so as to obtain a first intermediate image and a second intermediate image with the same size.
For example, the target test paper image may be divided according to a first dividing line, the first dividing line may be a vertical axis direction or a horizontal axis direction, and the first dividing line is a dividing line having a distance from the text content greater than or equal to a first threshold value or a dividing line having a distance from the frame of the target test paper image greater than or equal to a second threshold value. The first threshold value and the second threshold value may be preset in advance according to a format of a college entrance examination paper.
Furthermore, the target test paper image comprises a plurality of test questions, each test question is arranged according to the question number, and the test questions in the target test paper image can be distinguished through the question number of each test question. Since the question number of the question is located in the first area of the first and second intermediate images, the first area may be the leftmost area of the first and second intermediate images or the area of the first and second intermediate images that is a first distance from the left border, as shown in fig. 2 b. And carrying out image recognition on the image of the first area, recognizing the question number of each test question, and recording the position of the question number of each test question, which is far away from the upper frame of the first intermediate image or the second intermediate image. And calculating the starting and stopping positions of the current test question in the target test paper image according to the position of the current test question number and the position of the next test question number, wherein for example, the starting and stopping positions of the first test question are (50, 60) when the position of the question number of the first test question is 50mm away from the position of the upper frame of the first intermediate image, and the position of the question number of the second test question is 60mm away from the position of the upper frame of the first intermediate image. And segmenting the first intermediate image and the second intermediate image according to the sequence of the question numbers and the starting and stopping positions of each test question, specifically segmenting from the direction of a transverse axis at the starting and stopping position + x of each test question to obtain a target test question image of each test question, wherein x can be set according to the line spacing of the college entrance examination paper.
Exemplarily, if the first character of the first line in the second intermediate image is the question number of the test question, the test question in the first intermediate image is considered to be complete; and if the first character of the first line in the second intermediate image is not the question number of the test question, splicing the top target test question image divided from the second intermediate image and the bottom target test question image divided from the first intermediate image up and down to obtain a complete test question image.
In a possible example, the identifying the first region of the first intermediate image and the second intermediate image obtains the start-stop position of each test question; according to the starting and stopping positions of each test question, the first intermediate image and the second intermediate image are respectively segmented from a second direction to obtain a plurality of target test question images, and the method comprises the following steps:
cutting lines of the first region of the first intermediate image and the first region of the second intermediate image to obtain n line images, wherein n is a positive integer; extracting a first characteristic vector and a second characteristic vector of the n line images, wherein the first characteristic vector is a three-dimensional matrix formed by RGB values of the first region, and the second characteristic vector is a three-dimensional matrix formed by the perimeter, the area and the longest axis of the first region; performing head-to-tail splicing on the first characteristic vector and the second characteristic vector to obtain a target splicing vector, inputting the target splicing vector into a sharing sub-model of an image segmentation model to obtain a first posterior probability in each line image, wherein the sharing sub-model comprises a plurality of neural network operations (including convolution operation, activation operation, product operation and the like), and the first posterior probability is the probability that the line image comprises text content; combining the first posterior probabilities of the line images according to the segmentation sequence of the n line images to obtain a first probability vector (dimension is 1)
n), respectively inputting the probability vectors into the first task layer and the second task layer to execute operation to obtain a second probability vector and a third probability vector, wherein the second probability vector is a probability vector of text contents including target numbers in each line image. The third probability vector is a probability vector of the target number at the target position in each line image. The operation may specifically include: obtaining weight data of the first task layer or the second task layer, and obtaining a weight matrix (dimension is m) corresponding to the weight data
n) into k column vectors, m being a positive integer, from each of which a plurality of dimensions n will be extracted
The vectors of 1 (if m is smaller than n, all zero padding can be performed on the weight matrix first) are multiplied by the probability vectors respectively to obtain a plurality of product results, and the maximum value is selected from the plurality of product results to be used as the value of the corresponding column vector, so that a second probability vector or a third probability vector is obtained. M represents the number of question marks of the test question in the first task layer, and m represents the number of abscissas divided in the first area in the second task layer. Finally, determining a target line image comprising the question number of the test question by using the line image corresponding to the second probability vector with the value larger than or equal to the first threshold value, and taking the abscissa corresponding to the third probability vector with the value larger than or equal to the second threshold value of the target line image as the dividing line of the first intermediate image or the second intermediate image; and segmenting the first intermediate image and the second intermediate image according to the segmenting lines to obtain a plurality of target test question images.
The image segmentation model performs progressive decoding according to the input target splicing feature vector, and may be a Multi-task deep learning network. As shown in fig. 2c, the image segmentation model includes a shared sub-model, which may include a plurality of layers of convolutional neural networks and cyclic neural networks, a first task layer for performing a text recognition task, and a second task layer for performing a test question number position determination task.
In particular, the image segmentation model may classify each line image, the types of classification including text and non-text. And after the target splicing feature vector is input into a sharing sub-model of the text recognition model, the sharing sub-model outputs the posterior probability including the text in each line image.
Further, the posterior probability including the text in each line image is respectively used as the input of the first task layer. The first task layer calculates a second probability vector according to the posterior probability including the text in each line image by adopting a text recognition strategy, wherein the second probability vector is the posterior probability vector of which the text content in each line image is a target number, and therefore the question number of each test question in the first intermediate image and the second intermediate image is analyzed according to the value of the second probability vector.
Then, the question number and the first posterior probability of each test question are used as the input of the second task layer. And the second task layer determines a third probability vector in each line image according to the question number of each test question and the posterior probability including the text in each line image, and further determines whether the question number of each line image is at the preset horizontal coordinate position of the target test paper image or not to obtain the dividing line of the first intermediate image or the second intermediate image. The dividing line is a horizontal line of the vertical coordinate + y of the question number of each test, and y can be set according to the line spacing of the college entrance examination paper.
According to the technical scheme, the image segmentation of the first intermediate image or the second intermediate image is realized by identifying the segmentation line in the first region, the operation amount of the image segmentation is reduced, the Multi-task deep learning network is adopted to identify the test question number by combining two characteristics, the segmentation is carried out according to the coordinate of the test question number, the integrity of each test question is ensured, and the accuracy of the image segmentation is improved.
Optionally, the target neural network model comprises a text recognition model.
In the embodiment of the present application, the target neural network model may be a text recognition model for recognizing text content in the target test question image, and the text recognition model may be a convolutional neural network, a cyclic neural network, or the like.
The step of inputting the target test question image into the target neural network for text recognition to obtain a recognition result of the target test paper image comprises the following steps:
inputting the target test question image into the text recognition model to obtain a text recognition result and a text format of each test question; converting the text recognition result of each test question into a target text according to the text format of each test question; and combining the target text of each test question according to the starting and stopping positions of each test question to obtain the identification result of the target test paper image.
The character features of the target test question image are extracted through a feature extraction layer, such as a convolutional layer, in the text recognition model, and the character features can be geometric features, such as character positions, areas, end points of character lines, break points, intersection points and the like. And finally, inputting the multidimensional characteristic vector to an output layer in a character recognition model for conversion, so that a result of recognizing the characters in the character image to be recognized can be output, wherein the recognition result comprises a recognized text recognition result and a text format, the characters in the image can be recognized accurately, and the character recognition is performed through the character recognition model, so that the problem of inaccurate recognition of the similar characters can be avoided, and the character recognition accuracy is further improved.
Further, the text format may include a font size, a line spacing, a word arrangement, and the like of the text. Non-textual characters such as images, tables, formulas, etc. may be included in the test paper. By confirming the text format of the non-literal characters, the text recognition result can be converted into the target text with the same format as the high examination paper.
Optionally, the inputting the feature vector extracted from the target test question image into the text recognition model to obtain a text recognition result and a text format of each test question includes:
respectively identifying the target test question images from left to right according to a preset line spacing; if the recognition result of the first character is a non-text character, marking the first character, recording the start-stop coordinates of the first character, and confirming the target line spacing and the target area of the first character according to the start-stop position; and if the recognition result of the first character is a character symbol, confirming the target line spacing and the target area of the first character according to a preset line spacing and a preset font.
The examination paper of the college entrance examination generally adopts a fixed format, for example, the Chinese part generally adopts a Song style, if the title has a material analysis title, the material part adopts a regular style, and the English and digital parts adopt Times New Roman; the text is in Song style five, the characters in the form need to be arranged in the middle, and the character size is in Song style five. Therefore, in the embodiment of the present application, the text content within the preset line spacing may be identified, where the preset line spacing is greater than or equal to the line spacing of the test paper. Specifically, the target test question image is sequentially recognized from top to bottom and from left to right according to a preset line space, if the recognition result of the first character is recognized to be a non-character, such as a table, a graph, a formula and the like, the first character is marked, the start-stop coordinates of the first character are recorded, and the target area and the target line space occupied by the first character are calculated according to the start-stop position of the first character. If the recognition result of the first character is recognized as the text content, the format of the first character can be directly set according to a preset format. The preset format may include a preset line space and a preset font, and the preset line space and the preset size may be set according to the format of the college entrance examination paper.
It should be noted that the start-stop coordinates include coordinates of an upper left corner and a lower right corner of the first character, or coordinates of a lower left corner and an upper right corner of the first character, which is not limited in this embodiment of the present application.
S130, classifying the test questions of each channel based on the identification result to obtain a plurality of target test question sets.
In the examination paper of the college entrance examination, the examination questions of the graded examination subjects of the same question type are continuous, each examination question is a question of the examination outline of the graded examination subject, and the question stem of the examination paper comprises keywords representing the graded examination subject.
Optionally, the classifying the test questions of each channel based on the recognition result to obtain a plurality of target test question sets includes:
acquiring question stem information and a plurality of keywords of each test question from the text content of each test question; determining the target graded examination subject of each examination question according to the plurality of keywords; determining a target knowledge point of each test question for the question stem information and the keywords; and comparing the target knowledge point of each test question with the test outline of the target hierarchical test subject, classifying the test questions corresponding to the target knowledge point into a target test question set corresponding to the target hierarchical test subject when the target knowledge point is within the test outline range of the target hierarchical test subject, and otherwise deleting the test questions corresponding to the target knowledge point.
Specifically, the question stem information is obtained from the text content of each test question, then a plurality of keywords of each test question are determined from the question stem information, the keywords are respectively matched with the keywords in the keyword list of each pre-stored hierarchical test subject, and the target hierarchical test subject corresponding to the test question is determined according to the matched keywords. Further, the question range of the test question is determined based on the plurality of keywords, question stem information is analyzed, the question range of the test question is further narrowed, and the target knowledge point is determined, for example, for the comprehensive test question "1, and the narration about DNA and RNA, it is correct that: DNA has hydrogen bonds, RAN does not have hydrogen bonds, B, a virus contains DNA and RNA C, the existing DNA in prokaryotic cells, and RNA D, chloroplast, mitochondria and ribosome all contain DNA, the problem range of the test question can be determined to be the knowledge point between the DAN and the RAN in organisms according to the keywords 'DAN' and 'RAN', but the knowledge point range of the DAN and the RAN is also large, the difference between the DAN and the RAN can be determined by analyzing question stem information, wherein 'correct' can know the research of the test question, and then the information in the analysis option is analyzed, and the target knowledge point of the test question can be determined to be the knowledge point of the difference between the DNA and the RNA in the cells by analyzing the words of 'hydrogen bonds', 'cells', 'chloroplasts', 'viruses', and the like.
The knowledge schema under the new high-level examination model is also adjusted, so that after the target knowledge point is obtained, the target knowledge point can be compared with the knowledge schema of the target hierarchical examination subject, and if the adjusted knowledge schema of the target knowledge point is deleted, the test question corresponding to the target knowledge point is deleted, so that the richness of the test question resources can be increased. Each hierarchical examination subject corresponds to a test set. When the target knowledge points are in the knowledge outline, classifying the test questions of the target knowledge points into a target test question set corresponding to the target graded test subject, so as to divide and store the test questions of the college entrance exam in the past year according to the graded test subjects, and providing test question resources for automatically generating test paper under a 'new college entrance exam' model.
And S140, storing the test questions in the target test question set in the memory block of the target hierarchical examination subject corresponding to the target test question set according to the mapping relation between the test question set and the hierarchical examination subjects.
In a specific implementation, the memory may be divided into a plurality of memory blocks, and a mapping relationship between the memory blocks and the hierarchical examination subject is established. After the target test question sets of all the hierarchical examination subjects are obtained through classification, the test questions in the target test question sets are stored in the memory blocks of the target hierarchical examination subjects according to the mapping relation between the test question sets and the hierarchical examination subjects, so that when the examination papers of the hierarchical examination subjects under a new college entrance examination model are generated, the test questions meeting the requirements of paper grouping can be directly searched in the memory blocks, and the generation of the test papers can be accelerated.
It can be seen that the grammatical segmentation method based on the neural network algorithm provided by the embodiment of the application obtains the target test paper image, wherein the target test paper image is a scanned image of a comprehensive text test paper or a comprehensive physical test paper; inputting the target test paper image into a target neural network model for identification to obtain an identification result of each test question in the target test paper image; classifying each test question based on the identification result to obtain a plurality of target test question sets; according to the mapping relation between the test question set and the hierarchical examination subjects, the test questions in the target test question set are stored in the memory block of the target hierarchical examination subjects corresponding to the target test question set, so that each test question obtained after the traditional literature test paper is divided can be stored in the memory block of the corresponding hierarchical examination subjects, and test question resources are provided for automatically generating the test paper under the 'new high-level examination' model.
Referring to fig. 3, fig. 3 is a schematic diagram of a neural network algorithm based grammar splitting system according to an embodiment of the present application, and as shown in fig. 3, the system 300 may include: an acquisition unit 310, a recognition unit 320, a classification unit 330, and a storage unit 340, wherein,
the acquiring unit 310 is configured to acquire a target test paper image, where the target test paper image is a scanned image of a comprehensive text test paper or a comprehensive test paper;
the identification unit 320 is configured to input the target test paper image into a target neural network model for identification, so as to obtain an identification result of each test question in the target test paper image;
the classification unit 330 is configured to classify the test questions of each channel based on the identification result to obtain a plurality of target test question sets;
the storage unit 340 is configured to store the test questions in the target test question set in the memory blocks of the target hierarchical test subjects corresponding to the target test question set according to the mapping relationship between the test question set and the hierarchical test subjects.
The neural network algorithm based grammar splitting system provided by the application can be used for realizing the refinement scheme of the embodiment shown in fig. 1.
The present embodiment also provides a computer storage medium, in which computer instructions are stored, and when the computer instructions are run on an electronic device, the electronic device executes the relevant method steps to implement the neural network algorithm-based grammar segmentation method in the above embodiments.
In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described embodiments of the apparatus are merely illustrative, and for example, a module or a unit may be divided into only one logic function, and may be implemented in other ways, for example, a plurality of units or components may be combined or integrated into another apparatus, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
Units described as separate parts may or may not be physically separate, and parts displayed as units may be one physical unit or a plurality of physical units, may be located in one place, or may be distributed to a plurality of different places. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a readable storage medium. Based on such understanding, the technical solutions of the embodiments of the present application may be essentially or partially contributed to by the prior art, or all or part of the technical solutions may be embodied in the form of a software product, where the software product is stored in a storage medium and includes several instructions to enable a device (which may be a single chip, a chip, or the like) or a processor (processor) to execute all or part of the steps of the methods of the embodiments of the present application. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
The above description is only for the specific embodiments of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present application, and shall be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (7)
1. A method for segmenting a grammar based on a neural network algorithm, which is characterized by comprising the following steps:
acquiring a target test paper image, wherein the target test paper image is a scanned image of a comprehensive text test paper or a comprehensive test paper;
inputting the target test paper image into a target neural network model for identification to obtain an identification result of each test question in the target test paper image;
classifying the test questions of each channel based on the identification result to obtain a plurality of target test question sets;
and storing the test questions in the target test question set in a memory block of the target hierarchical examination subject corresponding to the target test question set according to the mapping relation between the test question set and the hierarchical examination subjects.
2. The method according to claim 1, wherein the recognition result includes text content of each test question;
classifying the test questions of each channel based on the identification result to obtain a plurality of target test question sets, including:
acquiring question stem information and a plurality of keywords of each test question from the text content of each test question;
determining the target graded examination subject of each examination question according to the plurality of keywords;
analyzing the question stem information to obtain target knowledge points of each test question;
and comparing the target knowledge point of each test question with the test outline of the target hierarchical test subject, classifying the test questions corresponding to the target knowledge point into a target test question set corresponding to the target hierarchical test subject when the target knowledge point is within the test outline range of the target hierarchical test subject, and otherwise deleting the test questions corresponding to the target knowledge point.
3. The method according to claim 1 or 2, wherein the inputting the target test paper image into a target neural network model for recognition to obtain a recognition result of each test question in the target test paper image comprises:
segmenting the target test paper image from a first direction to obtain a first intermediate image and a second intermediate image, wherein the first intermediate image comprises first layout information of the target test paper image, and the second intermediate image comprises second layout information of the target test paper image;
identifying first areas of the first intermediate image and the second intermediate image to obtain starting and stopping positions of each test question;
according to the starting and stopping positions of each test question, the first intermediate image and the second intermediate image are respectively segmented from a second direction to obtain a plurality of target test question images;
and inputting the multiple target test question images into the target neural network for text recognition to obtain a recognition result of the target test paper images.
4. The method of claim 3, wherein the target neural network model comprises a text recognition model;
the step of inputting the target test question image into the target neural network for text recognition to obtain a recognition result of the target test paper image comprises the following steps:
inputting the target test question image into the text recognition model to obtain a text recognition result and a text format of each test question;
converting the text recognition result of each test question into a target text according to the text format of each test question;
and combining the target text of each test question according to the starting and stopping positions of each test question to obtain the identification result of the target test paper image.
5. The method according to claim 4, wherein the inputting the feature vectors extracted from the target test question images into the text recognition model to obtain the text recognition result and the text format of each test question comprises:
respectively identifying the target test question images from left to right according to a preset line spacing;
if the recognition result of the first character is a non-text character, marking the first character, recording start-stop coordinates of the first character, and confirming a target line spacing and a target area of the first character according to the start-stop coordinates;
and if the recognition result of the first character is a character symbol, confirming the target line spacing and the target area of the first character according to a preset line spacing and a preset font.
6. A neural network algorithm based grammar segmentation system, the system comprising:
the device comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring a target test paper image, and the target test paper image is a scanned image of a comprehensive text test paper or a comprehensive test paper;
the identification unit is used for inputting the target test paper image into a target neural network model for identification to obtain an identification result of each test question in the target test paper image;
the classification unit is used for classifying the test questions of each channel based on the identification result to obtain a plurality of target test question sets;
and the storage unit is used for storing the test questions in the target test question set in the memory blocks of the target hierarchical test subjects corresponding to the target test question set according to the mapping relation between the test question set and the hierarchical test subjects.
7. A computer-readable storage medium, characterized in that a computer program for electronic data exchange is stored, wherein the computer program causes a computer to perform the method according to any one of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110092477.7A CN112560849B (en) | 2021-01-24 | 2021-01-24 | Neural network algorithm-based grammar segmentation method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110092477.7A CN112560849B (en) | 2021-01-24 | 2021-01-24 | Neural network algorithm-based grammar segmentation method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112560849A true CN112560849A (en) | 2021-03-26 |
| CN112560849B CN112560849B (en) | 2021-08-20 |
Family
ID=75035765
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110092477.7A Active CN112560849B (en) | 2021-01-24 | 2021-01-24 | Neural network algorithm-based grammar segmentation method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112560849B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113705706A (en) * | 2021-09-01 | 2021-11-26 | 北京云蝶智学科技有限公司 | Data classification method based on image recognition |
Citations (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1932795A (en) * | 2006-10-10 | 2007-03-21 | 青岛中科恒信信息技术有限公司 | Examination paper intelligent setting questions and organizing system |
| CN101667225A (en) * | 2009-09-25 | 2010-03-10 | 江苏省电力试验研究院有限公司 | Multidimensional question bank management system |
| CN103150938A (en) * | 2013-03-14 | 2013-06-12 | 南京信息工程大学 | Automatic question assignment method |
| CN106570109A (en) * | 2016-11-01 | 2017-04-19 | 深圳市前海点通数据有限公司 | Method for automatically generating knowledge points of question bank through text analysis |
| CN108932508A (en) * | 2018-08-13 | 2018-12-04 | 杭州大拿科技股份有限公司 | A kind of topic intelligent recognition, the method and system corrected |
| CN108959664A (en) * | 2018-09-26 | 2018-12-07 | 江苏曲速教育科技有限公司 | Distributed file system based on picture processor |
| CN109326161A (en) * | 2018-12-05 | 2019-02-12 | 杭州大拿科技股份有限公司 | A kind of paper corrects all-in-one machine |
| CN109389061A (en) * | 2018-09-26 | 2019-02-26 | 苏州友教习亦教育科技有限公司 | Paper recognition methods and system |
| CN109815948A (en) * | 2019-01-14 | 2019-05-28 | 辽宁大学 | A kind of paper partitioning algorithm under complex scene |
| CN110210413A (en) * | 2019-06-04 | 2019-09-06 | 哈尔滨工业大学 | A kind of multidisciplinary paper content detection based on deep learning and identifying system and method |
| CN110414529A (en) * | 2019-06-26 | 2019-11-05 | 深圳中兴网信科技有限公司 | Paper information extracting method, system and computer readable storage medium |
| CN110929573A (en) * | 2019-10-18 | 2020-03-27 | 平安科技（深圳）有限公司 | Examination question checking method based on image detection and related equipment |
| CN111259277A (en) * | 2020-01-10 | 2020-06-09 | 京丰大数据科技(武汉)有限公司 | Intelligent education test question library management system and method |
| CN111652141A (en) * | 2020-06-03 | 2020-09-11 | 广东小天才科技有限公司 | Question segmentation method, device, equipment and medium based on question number and text line |
| CN111737450A (en) * | 2020-08-05 | 2020-10-02 | 江西风向标教育科技有限公司 | Test paper grouping method and device, storage medium and computer equipment |
| CN111767424A (en) * | 2020-09-02 | 2020-10-13 | 北京新唐思创教育科技有限公司 | Image processing method, image processing device, electronic equipment and computer storage medium |
-
2021
- 2021-01-24 CN CN202110092477.7A patent/CN112560849B/en active Active
Patent Citations (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1932795A (en) * | 2006-10-10 | 2007-03-21 | 青岛中科恒信信息技术有限公司 | Examination paper intelligent setting questions and organizing system |
| CN101667225A (en) * | 2009-09-25 | 2010-03-10 | 江苏省电力试验研究院有限公司 | Multidimensional question bank management system |
| CN103150938A (en) * | 2013-03-14 | 2013-06-12 | 南京信息工程大学 | Automatic question assignment method |
| CN106570109A (en) * | 2016-11-01 | 2017-04-19 | 深圳市前海点通数据有限公司 | Method for automatically generating knowledge points of question bank through text analysis |
| CN108932508A (en) * | 2018-08-13 | 2018-12-04 | 杭州大拿科技股份有限公司 | A kind of topic intelligent recognition, the method and system corrected |
| CN108959664A (en) * | 2018-09-26 | 2018-12-07 | 江苏曲速教育科技有限公司 | Distributed file system based on picture processor |
| CN109389061A (en) * | 2018-09-26 | 2019-02-26 | 苏州友教习亦教育科技有限公司 | Paper recognition methods and system |
| CN109326161A (en) * | 2018-12-05 | 2019-02-12 | 杭州大拿科技股份有限公司 | A kind of paper corrects all-in-one machine |
| CN109815948A (en) * | 2019-01-14 | 2019-05-28 | 辽宁大学 | A kind of paper partitioning algorithm under complex scene |
| CN110210413A (en) * | 2019-06-04 | 2019-09-06 | 哈尔滨工业大学 | A kind of multidisciplinary paper content detection based on deep learning and identifying system and method |
| CN110414529A (en) * | 2019-06-26 | 2019-11-05 | 深圳中兴网信科技有限公司 | Paper information extracting method, system and computer readable storage medium |
| CN110929573A (en) * | 2019-10-18 | 2020-03-27 | 平安科技（深圳）有限公司 | Examination question checking method based on image detection and related equipment |
| CN111259277A (en) * | 2020-01-10 | 2020-06-09 | 京丰大数据科技(武汉)有限公司 | Intelligent education test question library management system and method |
| CN111652141A (en) * | 2020-06-03 | 2020-09-11 | 广东小天才科技有限公司 | Question segmentation method, device, equipment and medium based on question number and text line |
| CN111737450A (en) * | 2020-08-05 | 2020-10-02 | 江西风向标教育科技有限公司 | Test paper grouping method and device, storage medium and computer equipment |
| CN111767424A (en) * | 2020-09-02 | 2020-10-13 | 北京新唐思创教育科技有限公司 | Image processing method, image processing device, electronic equipment and computer storage medium |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113705706A (en) * | 2021-09-01 | 2021-11-26 | 北京云蝶智学科技有限公司 | Data classification method based on image recognition |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112560849B (en) | 2021-08-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110334346B (en) | Information extraction method and device of PDF (Portable document Format) file | |
| US10846553B2 (en) | Recognizing typewritten and handwritten characters using end-to-end deep learning | |
| US11580763B2 (en) | Representative document hierarchy generation | |
| US11080910B2 (en) | Method and device for displaying explanation of reference numeral in patent drawing image using artificial intelligence technology based machine learning | |
| JP5379085B2 (en) | Method and system for classifying connected groups of foreground pixels in a scanned document image based on marking type | |
| EP2166488A2 (en) | Handwritten word spotter using synthesized typed queries | |
| US20090041361A1 (en) | Character recognition apparatus, character recognition method, and computer product | |
| JP2005242579A (en) | Document processor, document processing method and document processing program | |
| CN111507330A (en) | Exercise recognition method and device, electronic equipment and storage medium | |
| CN110991403A (en) | Document information fragmentation extraction method based on visual deep learning | |
| Van Phan et al. | A nom historical document recognition system for digital archiving | |
| CN111340020A (en) | Formula identification method, device, equipment and storage medium | |
| CN115240213A (en) | Form image recognition method and device, electronic equipment and storage medium | |
| CN114187595A (en) | Document layout recognition method and system based on fusion of visual features and semantic features | |
| Ramirez et al. | Automatic recognition of square notation symbols in western plainchant manuscripts | |
| CN113673294B (en) | Method, device, computer equipment and storage medium for extracting document key information | |
| CN112560849B (en) | Neural network algorithm-based grammar segmentation method and system | |
| CN108845999B (en) | Trademark image retrieval method based on multi-scale regional feature comparison | |
| CN113642562A (en) | Data interpretation method, device and equipment based on image recognition and storage medium | |
| WO2007070010A1 (en) | Improvements in electronic document analysis | |
| CN117076455A (en) | Intelligent identification-based policy structured storage method, medium and system | |
| CN116822634A (en) | Document visual language reasoning method based on layout perception prompt | |
| CN113486171B (en) | Image processing method and device and electronic equipment | |
| CN114579796A (en) | Machine reading understanding method and device | |
| CN116311301B (en) | Wireless form identification method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |