Medical paper table information intelligent extraction system based on deep learning

Abstract

The invention is suitable for the technical field of natural language processing, and provides a medical paper table information intelligent extraction system based on deep learning, which comprises a medical paper platform, an acquisition module, an extraction module, a storage module, a training module, a table image acquisition module and an AI control module, according to the device, the table is detected through the table image acquisition module, table lines are detected, table information extracted by a system is checked, the integrity and accuracy of the extracted table information are calculated, the extracted table information is classified through the information content classification module, and the information content classification efficiency is improved. The AI control module is combined with a machine learning model and a deep learning model, so that the system has strong autonomous learning ability, and self behaviors are continuously corrected through big data, so that the system has intelligent decision making and intelligent operation planning ability.

Description

Technical Field

[0001] The present invention relates to the technical field of natural language processing, and more specifically, to a medical paper table information intelligent extraction system based on deep learning. Background Art

[0002] Table detection and analysis technology based on deep learning: The solution of deep learning technology to solve table detection and structural analysis is mainly implemented by deep learning target detection technology and image segmentation technology. It is generally divided into two steps. First, deep learning target detection technology is used to detect table area and classify tables. Then, image segmentation technology is used to segment the table line of the image of the table area. Then, cell position and row and column analysis are performed based on the table lines.

[0003] RPA is the abbreviation of Robotic Process Automation, which can be understood as a way for "virtual robots" to replace human labor. RPA can not only simulate humans, but also utilize and integrate existing technologies to achieve its goal of process automation.

[0004] Existing forms do not perform data classification and matching after recognition, and lack the process of calculating data classification and matching capabilities. Form recognition accuracy, completeness, and coverage are all important factors in measuring form recognition technology.

[0005] Therefore, it is very necessary to design an intelligent extraction system for medical paper table information based on deep learning and combined with advanced RPA technology. Summary of the invention

[0006] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a medical paper table information intelligent extraction system based on deep learning. The table and table lines are detected through the table image acquisition module, the table information extracted by the system is verified, and the completeness and accuracy of the extracted table information are calculated. The information content classification module classifies the extracted table information, and the AI ​​control module is combined with the machine learning model and the deep learning model. It has a strong autonomous learning ability, and continuously corrects its own behavior through big data, so as to have intelligent decision-making and intelligent operation capabilities.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A medical paper table information intelligent extraction system based on deep learning, comprising a medical paper platform, wherein the medical paper platform is used to carry various types of medical paper table information;

[0009] A collection module, wherein the collection module is used to collect and construct training data;

[0010] An extraction module connected to the acquisition module in communication, the extraction module constructs a neural network model according to the data collected by the acquisition module, and extracts information based on the neural network model;

[0011] The extraction module includes a neural network model connected to the acquisition module through communication;

[0012] A storage module in communication with the extraction module, the storage module storing the medical paper table information extracted based on the neural network model;

[0013] A training module in communication with the storage module, wherein the training module uses the historical results in the storage module as training data to train the constructed neural network model;

[0014] A table image acquisition module connected to the storage module for communication, the table image acquisition module uses deep learning image processing technology to detect the table and the table lines;

[0015] And an AI control module that is communicatively connected to the storage module. The AI ​​control module combines a machine learning model and a deep learning model, has a strong autonomous learning ability, and continuously corrects its own behavior through big data, thereby having intelligent decision-making and intelligent operations capabilities.

[0016] The present invention is further configured to include: a user terminal and a human-computer interaction module; the human-computer interaction module is used to obtain the demand information in the natural language form input by the user;

[0017] Various types of medical paper table information are entered into the medical paper platform; the medical paper platform is connected to the user terminal through a human-computer interaction module; the user enters the human-computer interaction module through the user terminal and enters the medical paper table information through the medical paper platform.

[0018] The present invention is further configured as follows: the neural network model includes a BERT model, a BI LSTM model and a CRF model, the output data of the BERT model is input into the BI LSTM model, and the data output in the BI LSTM model is used as the input of the CRF model, wherein the BERT model is used to realize the function of word embedding, and the BI LSTM model and the CRF model construct a training neural network and constrain the output results.

[0019] The present invention is further configured as follows: the processing module includes a table image acquisition unit, a text row position detection unit, a text row content detection unit, an image direction correction unit, a table line analysis and prediction unit, a table line analysis and fusion unit, a table cell information analysis unit, a table cell information fusion unit and a formatted output unit.

[0020] The present invention is further configured as follows: the deep learning model adopts a CNN+FPN network structure for feature extraction, and then adopts an RPN network structure to generate multiple table candidate areas, performs table classification, table boundary regression and table line segmentation prediction on each table candidate area, and finally obtains all lines of all wired tables and all lines of all wireless tables on the current image.

[0021] The present invention is further configured as follows: the AI ​​control module includes a ROBOT unit and an AI+RPA unit; the ROBOT unit is connected to the AI+RPA unit via an engine; and the output end of the data preprocessing module is communicatively connected to the ROBOT unit.

[0022] The present invention is further configured as follows: the AI ​​control module also includes a control program; the AI ​​control module is electrically connected to the extraction module, and the AI ​​control module is used to execute the control program to achieve the information extraction based on the neural network model as described above.

[0023] The advantages of the present invention are:

[0024] 1. The data output by the information intelligent extraction system of the present invention conforms to the standardized format and structure, which is convenient for data interaction and integration with other systems; users can easily import the extracted data into various applications for analysis, processing and utilization, thereby improving work efficiency and data utilization.

[0025] 2. The present invention is based on deep learning table detection and analysis technology, which is implemented by using deep learning target detection technology and image segmentation technology. Deep learning target detection technology is used to perform table area detection and table classification, and then the image of the table area is segmented by image segmentation technology to perform table line segmentation, and then cell position and row and column analysis are performed based on the table lines, thereby improving the detection accuracy of information and thus improving the extraction accuracy of table information in medical papers.

[0026] 3. The present invention detects the table and the table lines through the table image acquisition module, verifies the table information extracted by the system, calculates the completeness and accuracy of the extracted table information, and classifies the extracted table information through the information content classification module. It combines the machine learning model and the deep learning model through the AI ​​control module, has a strong autonomous learning ability, and continuously corrects its own behavior through big data, thereby having intelligent decision-making and intelligent operation capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a flow chart of a system for intelligent extraction of medical paper table information based on deep learning according to the present invention. DETAILED DESCRIPTION

[0028] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application may be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0029] It should be noted that, unless otherwise specified, all technical and scientific terms used in this application have the same meanings as commonly understood by ordinary technicians in the technical field to which this application belongs.

[0030] In the present invention, unless otherwise specified, the directions used, such as "up" and "down", usually refer to the directions shown in the drawings, or to the vertical, perpendicular or gravity directions; similarly, for ease of understanding and description, "left" and "right" usually refer to the left and right shown in the drawings; "inside" and "outside" refer to the inside and outside relative to the outline of each component itself, but the above-mentioned directions are not used to limit the present invention.

[0031] Embodiment 1

[0032] See also Figure 1 , the present invention provides the following technical solutions:

[0033] A medical paper table information intelligent extraction system based on deep learning, specifically, comprising a medical paper platform, the medical paper platform is used to carry various medical paper table information;

[0034] Acquisition module,The acquisition module is used to collect and construct training data.

[0035] The extraction module is connected to the acquisition module in communication, and the extraction module constructs a neural network model according to the data collected by the acquisition module, and extracts information based on the neural network model.

[0036] The extraction module includes a neural network model that is connected to the acquisition module through communication.

[0037] The storage module is connected to the extraction module in communication, and the storage module stores the medical paper table information extracted based on the neural network model.

[0038] A training module is communicatively connected to the storage module, and the training module uses the historical results in the storage module as training data to train the constructed neural network model.

[0039] A table image acquisition module is communicatively connected to the storage module, and the table image acquisition module uses deep learning image processing technology to detect the table and the table lines.

[0040] And an AI control module that communicates with the storage module. The AI ​​control module combines machine learning models and deep learning models, has strong autonomous learning capabilities, and continuously corrects its own behavior through big data, thus having intelligent decision-making and intelligent operations capabilities.

[0041] The working principle of the first embodiment of the present invention is as follows: the table and the table lines are detected through the table image acquisition module, the table information extracted by the system is checked, the completeness and accuracy of the extracted table information are calculated, and the information content classification module classifies the extracted table information. The AI ​​control module is combined with the machine learning model and the deep learning model. It has a strong autonomous learning ability and continuously corrects its own behavior through big data, thereby having intelligent decision-making and intelligent operation capabilities.

[0042] Embodiment 2

[0043] See also Figure 1 The second embodiment makes the following improvements on the basis of the first embodiment. Specifically, it also includes a user terminal and a human-computer interaction module; the human-computer interaction module is used to obtain the demand information in the natural language form input by the user; various types of medical paper form information are entered into the medical paper platform; the medical paper platform is connected to the user terminal through the human-computer interaction module; the user enters the human-computer interaction module through the user terminal and enters the medical paper form information through the medical paper platform.

[0044] The data output by the medical paper platform conforms to a standardized format and structure, facilitating data interaction and integration with other systems; users can easily import the extracted data into various applications for analysis, processing and utilization, improving work efficiency and data utilization.

[0045] The neural network models include the BERT model, the BI LSTM model and the CRF model. The output data of the BERT model is input into the BI LSTM model, and the data in the BI LSTM model is used as the input of the CRF model. The BERT model is used to realize the word embedding function. The BI LSTM model and the CRF model build a training neural network and constrain the output results.

[0046] The BERT model is an architecture that can be used for many downstream tasks, such as answering questions, classification, NER, etc. The working principle of the BERT model is the same as that of most Imagenet deep learning models. First, the BERT model is trained on a large corpus (Masked LM task), and then the model is fine-tuned for our own task by adding some extra layers at the end. The model can be classification, question answering, or NER, etc.

[0047] Bi LSTM (Bidirectional Long Short-Term Memory) is a special type of recurrent neural network (RNN) that can process sequence data and maintain long-term memory. Unlike traditional RNN models, Bi LSTM considers both past and future information, allowing the model to better capture contextual relationships in sequence data.

[0048] Conditional Random Field (CRF) is a basic model in natural language processing and is widely used in scenarios such as word segmentation, entity recognition, and part-of-speech tagging. With the popularization of deep learning, models such as BI LSTM+CRF, BERT+CRF, and TRANSFORMER+CRF have gradually appeared, and their effects have been significantly improved in these tagging scenarios.

[0049] The processing module includes a table image acquisition unit, a text line position detection unit, a text line content detection unit, an image direction correction unit, a table line analysis prediction unit, a table line analysis fusion unit, a table cell information analysis unit, a table cell information fusion unit and a formatted output unit.

[0050] Each unit in the above processing module is an existing conventional technology. For specific content, please refer to a table information extraction method and system with publication number CN114419647A, which will not be described in detail again.

[0051] The deep learning model uses the CNN+FPN network structure for feature extraction, and then uses the RPN network structure to generate multiple table candidate areas. It performs table classification, table boundary regression, and table line segmentation prediction on each table candidate area, and finally obtains all lines of all wired tables and all lines of all wireless tables on the current image.

[0052] The AI ​​control module includes a ROBOT unit and an AI+RPA unit; the ROBOT unit and the AI+RPA unit are connected via an engine; the output end of the data preprocessing module is communicated with the ROBOT unit.

[0053] The AI ​​control module combines machine learning models and deep learning models, and has strong autonomous learning capabilities. Its OCR, NLP, speech recognition and other technologies enable the RPA unit to have cognitive capabilities, and can continuously correct its own behavior through big data, thereby having intelligent decision-making and intelligent operations capabilities.

[0054] The relationship between AI and RPA is like the relationship between "brain command" and "hands and feet operation".

[0055] Its features are as follows:

[0056] 1.RPA tends to execute commands repeatedly, while AI tends to issue commands

[0057] 2. RPA robots can automate simple tasks and provide big data for AI

[0058] 3. AI can imitate and improve processes based on the data provided by RPA

[0059] 4.RPA is process-centric, while AI is data-centric

[0060] AI combines machine learning and deep learning, and has strong autonomous learning capabilities. Its OCR, NLP, speech recognition and other technologies enable RPA to have cognitive abilities, and can continuously correct its own behavior through big data, thus having intelligent decision-making and intelligent operations capabilities.

[0061] The AI ​​control module also includes a control program; the AI ​​control module is electrically connected to the extraction module; the AI ​​control module is used to execute the control program to achieve the information extraction based on the neural network model as described above.

[0062] The working principle of the second embodiment of the present invention is as follows: the table detection and analysis technology based on deep learning is realized by using deep learning target detection technology and image segmentation technology. The deep learning target detection technology is used to detect the table area and classify the table. Then the image of the table area is segmented by the image segmentation technology. Then the cell position and row and column analysis are performed according to the table lines, which improves the detection accuracy of the information and thus improves the extraction accuracy of the table information of the medical paper. The table and the table lines are detected by the table image acquisition module, the table information extracted by the system is tested, the completeness and accuracy of the extracted table information are calculated, and the information content classification module is used to classify the extracted table information. The AI ​​control module is combined with the machine learning model and the deep learning model. It has a strong autonomous learning ability and can continuously correct its behavior through big data, thus having intelligent decision-making and intelligent operation capabilities.

[0063] Obviously, the above-described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0064] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0065] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein.

[0066] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0067] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A medical paper table information intelligent extraction system based on deep learning, characterized by: Including a medical paper platform, the medical paper platform is used to carry various medical paper table information; A collection module, wherein the collection module is used to collect and construct training data; An extraction module connected to the acquisition module in communication, the extraction module constructs a neural network model according to the data collected by the acquisition module, and extracts information based on the neural network model; The extraction module includes a neural network model connected to the acquisition module through communication; A storage module in communication with the extraction module, the storage module storing the medical paper table information extracted based on the neural network model; A training module in communication with the storage module, wherein the training module uses the historical results in the storage module as training data to train the constructed neural network model; A table image acquisition module connected to the storage module for communication, the table image acquisition module uses deep learning image processing technology to detect the table and the table lines; And an AI control module that is communicatively connected to the storage module. The AI ​​control module combines a machine learning model and a deep learning model, has a strong autonomous learning ability, and continuously corrects its own behavior through big data, thereby having intelligent decision-making and intelligent operations capabilities.

2. According to claim 1, a medical paper table information intelligent extraction system based on deep learning is characterized by: It also includes a user terminal and a human-computer interaction module; the human-computer interaction module is used to obtain the demand information in the natural language form input by the user; Entering various medical paper form information into the medical paper platform; The medical paper platform is connected to the user terminal via a human-computer interaction module; The user enters the human-computer interaction module through the user terminal and enters the medical paper form information through the medical paper platform.

3. According to claim 2, a medical paper table information intelligent extraction system based on deep learning is characterized by: The neural network model includes a BERT model, a BILSTM model and a CRF model. The output data of the BERT model is input into the BILSTM model, and the data in the BILSTM model is output as the input of the CRF model. The BERT model is used to realize the function of word embedding. The BILSTM model and the CRF model construct a training neural network and constrain the output results.

4. According to claim 3, a medical paper table information intelligent extraction system based on deep learning is characterized by: The processing module includes a table image acquisition unit, a text line position detection unit, a text line content detection unit, an image direction correction unit, a table line analysis and prediction unit, a table line analysis and fusion unit, a table cell information analysis unit, a table cell information fusion unit and a formatted output unit.

5. According to claim 4, a medical paper table information intelligent extraction system based on deep learning is characterized by: The deep learning model adopts the CNN+FPN network structure for feature extraction, and then adopts the RPN network structure to generate multiple table candidate areas, performs table classification, table boundary regression and table line segmentation prediction on each table candidate area, and finally obtains all lines of all wired tables and all lines of all wireless tables on the current image.

6. According to claim 5, a medical paper table information intelligent extraction system based on deep learning is characterized by: The AI ​​control module includes a ROBOT unit and an AI+RPA unit; the ROBOT unit is connected to the AI+RPA unit via an engine.

7. The medical paper table information intelligent extraction system based on deep learning according to claim 6 is characterized by: The AI ​​control module also includes a control program; the AI ​​control module is electrically connected to the extraction module; the AI ​​control module is used to execute the control program to achieve the information extraction based on the neural network model as described above.

Images