TWI793432B - Document management method and system for engineering project - Google Patents

Abstract

The present invention relates to a document management method for an engineering project, comprising: establishing a plurality of electronic documents including a plurality of words and uploading the plurality of electronic documents to an engineering project management platform; recognizing a plurality of keywords included in the plurality of electronic documents by an natural language analyzer included in the engineering project management platform, and storing the plurality of keywords in a keyword database; according to the plurality of keyword and association rules distributing the plurality of the electronic documents to at least one related party and classifying the plurality of the electronic documents into at least one engineering item; and providing a natural language based document enquiry component to provide for a user to find out a corresponding document through searching the plurality of keywords in the keyword database.

Description

Project document management method and system

The present invention relates to a method and system for project file management, in particular to a set of project file management method and system capable of automatically assigning files to related parties and filing them, thereby realizing file automation.

In the process of engineering projects, the derived related documents are extremely complex. Not only are there many types of documents, such as: photos, PDF, word, ..., etc., but also often involve many stakeholders, and the project management items are also extremely detailed and extensive. In conventional technology, during the process of project execution, it is usually customary to manually assign documents or to-do items to related parties, or to classify relevant materials into different project management items.

However, in the case of an extremely large amount of data, many errors often occur, such as assignment errors, classification errors, omissions, etc., and other derivative problems, such as: it is difficult to track the progress of assignments, and it takes a lot of time to inquire about relevant information. It will follow. In recent years, due to the advancement of technology, many engineering project platforms have emerged to speed up the file transfer process and establish a joint collaboration model, or apply natural language processing technology to try to automate the file processing process as much as possible.

For example, among the commonly used technologies, the Chinese Minority Technology proposed by Liu Bingjin et al. National Invention Patent No. I682286 "Document Search System Using Text Analysis Results and Natural Language Input", which discloses the use of natural language processing (natural language processing, NLP) technology for document search systems. Word analysis can enhance the search speed, and it can also receive user input on the cloud platform to search for documents with natural grammar containing keywords, so as to improve the efficiency of document search. There is also a new model of the Republic of China proposed by "Qiaba Smart Co., Ltd." Patent No. M583974 "Document Information Extraction and Archiving System", which discloses the use of text detection and text recognition technology to identify and retrieve various information recorded in documents, and store the retrieved information in its own database for use .

However, most of the above-mentioned conventional technologies are still manual for the distribution and classification of relevant documents, which is not only extremely time-consuming and error-prone, but also lacks the ability to find out the correlation between different data items in a huge amount of documents. , realize the function of jointly linking the work processes of all parties involved and automatically recording the history, making it impossible for all parties to communicate and connect, and it is also impossible to clearly grasp the progress of the project and the construction cost.

For this reason, the inventor, after painstaking attempts and research, and a spirit of perseverance, finally conceived the project "engineering project document management method and system", which can overcome the above-mentioned shortcomings. The following is a brief description of the present invention.

In view of the shortcomings of conventional technologies, this invention proposes to combine natural language processing technology with big data data mining, association rules and deep learning technologies to automate the document assignment and classification work of the platform, improve the efficiency of document processing, and enable fast search Documents and other functions; the present invention automatically identifies documents of different file types through natural language processing technology, and establishes a keyword database, and through the keyword database, two data such as contract stakeholders and project management Databases establish association rules, and then automatically assign documents to associated stakeholders and classify them into corresponding engineering project management databases. In addition, through the establishment of a keyword database to enhance the search speed, users can use the cloud interactive platform as a medium to input natural language keywords for document search, reducing the time spent in using a large number of databases.

The present invention proposes a management method for engineering project documents, which includes: creating multiple electronic documents containing plural characters and uploading them to the engineering project document management platform; using the natural language parser included in the engineering project document management platform to identify the electronic documents contained multiple keywords and store them in the keyword database; assign these electronic documents to at least one related person and classify them into at least one project management project according to the keywords and the association rules; and provide natural language document query The component is used for the user to find the corresponding electronic document by searching the keywords in the keyword database.

Preferably, the project document management method also includes one of the following: uploading these electronic documents to the document database; implementing character detection through the character detection components contained in the natural language parser To determine the location of the text; implement character recognition through the character recognition component included in the natural language parser to identify the detected text; through the named entity recognition component included in the natural language parser Implement named entity recognition to identify the keywords included in the identified text; collect the identified keywords to create the keyword database; and provide natural language document query components to provide users with natural directly search the keywords in the keyword database, and link to the inquired electronic documents accordingly, so as to find the inquired electronic documents from the electronic documents included in the document database.

Preferably, the project project file management method also includes one of the following One: The character detection element is implemented by implementing one of the morphological operation method, MSER method, NMS method, CTPN method, SegLink method, EAST method, R-CNN method, fast RCNN method, PSENet method and combinations thereof The location of the characters is detected; the character recognition element recognizes the detected characters by implementing a deep learning method, the deep learning method is a convolutional neural network, a deep convolutional neural network, One of recurrent neural network, convolutional recurrent neural network, convolutional recurrent neural network optical character recognition, attentional optical character recognition, and combinations thereof; the named entity recognition element is generated by referring to the natural language rule dataset implementing one of a rule-based method, an unsupervised learning method, a feature-based supervised learning method, the deep learning method, and combinations thereof to build the keyword database; To create the key word database.

Preferably, the project file management method also includes one of the following: establishing a deep learning data set to train the deep learning method and provide learning for the deep learning method; establishing a natural language rule data set to provide the naming The entity recognition component is used as a reference; the deep learning data set and the natural language rule data set are used as a training set to train the deep learning method; a relational person database is established, and the relational person database includes plural relational persons and each such Keywords corresponding to related parties; establish a project project database, which contains multiple project management projects and keywords corresponding to each of these project management projects; implement association rule algorithms to establish these keywords an association rule with respect to each of the related parties in the related party database and each of the engineering management items in the engineering project database; and optimizing the association rule by applying the deep learning method.

The present invention further proposes an engineering project file management system, which includes: a back-end data layer for storing natural language rule data sets, deep learning data sets, keyword databases, related person databases, engineering project databases, and document data One of the libraries; the relay logic layer, which executes the project file management platform and the natural language parser contained in the project file management platform; and the front-end display layer, which executes the project file management platform. A language file query component, wherein the user uploads multiple electronic files containing multiple characters to the engineering project file management platform through the front-end component on the user device, so that the natural language parser can identify the multiple keys contained in the electronic files words, and store them in the keyword database, so that the engineering project document management platform can assign these electronic documents to at least one related person and classify them into engineering management projects according to the keywords and the association rules, and through the The natural language document query component searches the keywords in the keyword database to find the corresponding electronic documents.

The above summary of the invention is intended to provide a simplified summary of the disclosure to enable readers to have a basic understanding of the disclosure. This summary of the invention is not intended to disclose a complete description of the invention, and is not intended to point out important/key elements or components of the embodiments of the invention. define the scope of the invention.

10: Project document management system of the present invention

100: Front-end display layer

101:Desktop Computer

103: Notebook computer

105: Tablet device

107:Smartphone

110: user device

120: user

130:File upload user interface

150: Server load balancing equipment

200: relay logic layer

250: Server load balancing equipment

300: backend data layer

500: Project file management method of the present invention

501-506: Implementation steps

Figure 1 discloses a schematic diagram of the system architecture of the engineering project document management system of the present invention;

Figure 2 shows a schematic diagram of the system operation of the engineering project document management system of the present invention;

Figure 3 shows a schematic diagram of the electronic documents used by the engineering project document management platform of the present invention;

Figure 4 discloses a schematic diagram of a rectangular text box marked on a document by the character detection device of the present invention;

FIG. 5 shows a schematic diagram of the module structure of the character detection component, the character recognition component and the named entity recognition component included in the natural language parser of the present invention;

Figure 6 discloses a schematic diagram of the platform user interface provided by the project project document management platform of the present invention to the user through the front-end display layer;

Figure 7 discloses a block diagram of the operating principle of the engineering project document management system of the present invention; and

Fig. 8 discloses a flow chart of the implementation steps of the engineering project file management method of the present invention.

The present invention can be fully understood by the following examples, so that those skilled in the art can complete it, but the implementation of the present invention can not be limited by the following examples of implementation; the drawings of the present invention are not limited No limitation on size, dimension and scale is included, and the size, dimension and scale of the present invention are not limited by the drawings of the present invention during the actual implementation.

The word "preferred" in this article is non-exclusive and should be understood as "preferably but not limited to". order, the scope of the present invention should be determined only by the appended claims and their equivalents, not by the examples illustrated in the implementation; when the term "comprising" and its variations appear in the specification and claims, it is An open-ended term without a restrictive meaning that does not exclude other features or steps.

Figure 1 discloses a schematic diagram of the system architecture of the project project document management system of the present invention; Figure 2 discloses a schematic diagram of the system operation of the project project document management system of the present invention; the project project document management system 10 of the present invention is composed of a back-end data layer 300, a relay The logic layer 200 and the front-end display layer 100 are composed of a three-tier architecture, and the back-end data layer 300 includes at least one or more database servers server, the relay logic layer 200 includes at least one or more program servers, the front-end display layer 100 includes at least one or more web servers, and the servers are connected through the Internet, and each layer includes Multiple servers will use the server load balancing equipment 150 and 250 to reasonably distribute the workload when the connection is busy, effectively use the server capacity, and speed up the response speed of each layer of servers.

The database server included in the back-end data layer 300 is mainly used to store multiple databases, including natural language rule datasets, deep learning datasets, keyword databases, relational person databases, engineering project databases, and document databases etc., the program server included in the relay logic layer 200 is mainly used to execute the back-end components of the engineering project document management platform, including a natural language parser, and the natural language parser also includes a character detection component, a character recognition component, and a named entity Identify components, etc.

The engineering project document management platform includes front-end components and back-end components. The front-end components include front-end user interface components and natural language document query components. The back-end components include natural language parsers. Execute the front-end components of the engineering project document management platform to provide the user 120 to operate the engineering project document management platform through the front-end user interface components on the user device 110, and access the keyword database, related party database or engineering project data library, the user device 110 is preferably a desktop computer 101 , a notebook computer 103 , a tablet device 105 or a smart phone 107 and the like.

The related person database stored in the back-end data layer 300 stores the keywords involved in each related person in the engineering project, so as to make the keyword database and the related person database related, and the engineering project database stores The keywords corresponding to each project management item included in the engineering project, so as to make the keyword database and the engineering project database related, The above-mentioned keyword database, relational person database and engineering project database are in structured (structured) data format or in semi-structured (semi-structured) data format, such as: CSV, record file, XML Binary text stored in , JSON, etc., or stored in an unstructured data format, such as a general computer file, in order to provide access, retrieval, training or learning for front-end components and back-end components.

In the process of engineering projects, the derived related documents are extremely complex. Not only are there many types of electronic documents, such as but not limited to: jpeg, pdf, doc, or docx, etc., but also involve a wide range of related parties and engineering management projects. In conventional techniques, various documents and materials are often assigned manually, or to-do items are assigned to project undertakers, or related materials are classified into different project management items. However, this manual assignment method often produces many errors when the amount of data is extremely complex, such as assignment errors, classification errors, omissions, etc., so the engineering project file management platform proposed by the present invention includes a natural language parser , can automatically identify the words or characters contained in electronic files, and can automatically send to related parties, automatically classify files, and automatically file files into their own engineering management projects to realize file automation.

Figure 3 reveals a schematic diagram of the electronic documents used by the engineering project document management platform of the present invention; Figure 3 reveals common electronic contract documents in the process of engineering patents, but such electronic contract documents may be uploaded in the .doc format of WORD files The engineering project file management platform of the present invention may also be uploaded to the project project file management platform of the present invention in the .pdf format of the scanned PDF file, but it is also likely to be uploaded to the project of the present invention in the .jpeg format of the image file taken by the mobile phone The project document management platform, only the electronic contract documents shown in Figure 3, may be uploaded to the platform in the format of jpeg, pdf or doc, and the file format of the electronic documents is preferably a PDF file file format, Power Point file format, Power Point file compatible format, WORD file format, WORD file compatible format, Excel file format, Excel file compatible format, JPG file format, JPEG file format, or PNG file format.

Figure 5 discloses a schematic diagram of the module architecture of the character detection component, character recognition component and named entity recognition component included in the natural language parser of the present invention; the text information contained in the above-mentioned electronic documents may be represented by characters Various types of codes, text images, or picture objects with text images are stored in various types of files. In order to identify the text contained in electronic files, the natural language parser of the present invention is configured to include character detection Components, character recognition components and named entity recognition components. The natural language parser performs character detection, character recognition and named entity recognition in sequence to perform character recognition and entity recognition on the text contained in the electronic document.

Figure 4 discloses a schematic diagram of a rectangular text box marked on a document by the character detection element of the present invention; the character detection element mainly performs scene text detection (scene text detection), and the scene character detection is intuitively understood , is to find out the specific position of all text or characters in the file from the file, and mark the found text with a rectangular text box. The character detection component is preferably selected from the application morphology operation method, MSER method, and NMS method. , CTPN method, SegLink method, EAST method, R-CNN method, fast RCNN method, PSENet method or a combination of these methods to perform scene character detection.

The character recognition component is mainly to perform scene text recognition. The scene character recognition is to recognize the text contained in the text box within the range marked by the text box, so as to extract the text. The character recognition component is relatively Jia can choose to use Optical Character Recognition (OCR) for text recognition in the early stage. The recognition process is divided into two steps: text cutting and classification For the homework, a single font is first segmented by projection cutting, and then sent to convolutional layers for classification.

In one embodiment, the present invention chooses to use deep learning related technologies for character recognition, so the step of character cutting can be omitted, and character recognition can be converted into a sequence learning problem. After the deep convolutional neural network (DCNN) and the recurrent neural network (RNN) are calculated, the output stage undergoes a certain translation, and the entire text image can be identified. For example, it is better to choose the convolutional recurrent neural network. -Optical character recognition (CRNN-OCR) or attention OCR (Attention OCR) for scene character recognition, because CRNN-OCR and attention OCR both use convolutional neural network (CNN) plus RNN in the feature learning stage The complex network structure, so it can be learned directly from sequence labels without detailed annotations, not limited by the length of sequence-like objects, and only needs a high degree of standardization in the training and testing stages. It performs better in recognition and takes up less storage space.

The named entity recognition component mainly performs the task of named entity recognition (named entity recognition), also known as proper name recognition and named entity. The purpose is to further identify entities or proper nouns with specific meanings from the recognized text. Covering things or entities that can be identified by proper nouns or names, such as person names, place names, organization names, proper nouns, time, numbers, quantities, currency, and proportional values, is a common task in natural language parsers. Use The scope is very wide; the named entity recognition element proposed by the present invention has been enhanced to further identify more entities, such as product names, models, prices, etc.; all identified entities and proper names will be stored in the back-end data layer 300 And form a keyword database.

Compared with English records, words can be separated by blanks or other symbols, so there is no problem of hyphenation processing, but Chinese sentences are full of multiple words that are coherent and have no obvious boundaries. Characters, how to decipher Chinese sentences has always been a technical problem, especially how to identify one word with multiple meanings and multiple words with one meaning, so the present invention uses a series of technologies based on natural language processing (NLP), including Text mining (Text Mining), data mining and other technologies are used to pre-process relevant data. Text mining technology can discover unknown, implicit and useful information from structured, unstructured or semi-structured text data formats. To edit, organize and analyze a large amount of Chinese information, and analyze the frequency and quantity of vocabulary and other related data, it is mostly used in trend forecasting, decision-making assistance and other fields.

In an embodiment, the present invention chooses to apply rule-based methods, unsupervised learning approaches (unsupervised learning approaches), feature-based supervised learning approaches (feature based supervised learning approaches), deep learning methods (deep learning approaches), etc. Class methods, such as but not limited to: Bidirectional Neural Network (BRNN)-Convolutional Recurrent Neural Network (CRNN) method, to perform proper name recognition or entity recognition tasks.

In a certain embodiment, the present invention chooses to apply the word segmentation and entity recognition system (CoreNLP) developed by the Central Research Institute as the core natural language processing algorithm to perform named entity recognition, and use the word segmentation rule to solve the ambiguity of Chinese sentence segmentation Problem, in one embodiment, the implementation steps of the Chinese word segmentation method include: preliminary word segmentation, tagging, unknown word detection, Chinese name extraction, European and American translation name extraction, compound word extraction, and following As for the steps of merging and sorting and re-segmentation, the segmentation method mainly consists of a thesaurus and a group of segmentation rules. The vocabulary in the thesaurus is the correct vocabulary that has been built in advance and corrected by manual inspection. These benchmark words (tokens) constitute texts (texts).

The above-mentioned process of word segmentation to obtain basic vocabulary (tokens) is also called text segmentation (text segmentation). After word segmentation and tagging are completed, part-of-speech tags need to be established (part-of-speech tagging), and a dictionary as an index to convert vocabulary and text into a sequence of numbers recognizable by the processor, so that the input deep learning technology can read, recognize, learn and Training, and establish natural language rule data sets and deep learning data sets as training sets to train natural language parsers or provide natural language parser learning. The training set is mainly implemented through the implementation of sentence-based rules, supervised Different identification methods such as traditional learning, unsupervised learning, and deep learning establish corresponding databases to assist the natural language identification unit to produce accurate identification results.

The character detection components, character recognition components, and named entity recognition components included in the above natural language parser all use the most advanced deep learning technology, so a large amount of sample data is required as a training set to provide a natural language parser that includes After training and calibration, the natural language parser can correctly perform tasks including text recognition and entity recognition, with an accuracy rate between 90% and 95%. These data training sets include but are not limited to: storage A natural language rule data set and a deep learning data set on the back-end data layer 300 database server.

After the construction of the keyword database is completed, it is necessary to establish association rules between the keyword database, the relationship database and the engineering project database, so that the relationship between the keyword database and the relationship database , and there is an association between the keyword database and the engineering project database, the related person database stored in the back-end data layer 300 contains keywords involved in each related person in the engineering project, and the engineering project database includes the engineering project After the keywords involved in each project management project in the project are analyzed and calculated by applying the association rule algorithm, the correct corresponding relationship can be established between the keyword database, the relational person database and the engineering project database, so that the platform According to the instructions, the documents are assigned to the corresponding stakeholders or classified into the relevant Corresponding engineering management items.

For example, the database of related parties will store the keyword records, vocabulary records and text records involved in the documents handled by all related parties in the past. Each record has an identifiable field, and each record will contain one or For multiple data items, the data mining of the association rule algorithm will find out frequently occurring data item sets from a large number of data item sets to judge whether the correlation strength between the data items is sufficient. , which means that the set of data items has some meaning of existence, so the algorithm will give a threshold value during the calculation process, which is better called support (support), to judge whether the frequency of occurrence of a set of data items meets the threshold basis for the value.

When the number of occurrences of a data item set is higher than the support degree, the data item set is called a high-frequency data item set. When there are two data items A and B in a certain data item set, and the frequency of discovery is higher than the support If the degree is set, the algorithm will determine that candidate data items A and B are high-frequency data item sets, and data item A and B are related, and then test the reliability of this high-frequency data item through conditional probability. When A occurs, the probability that B data item B will also occur at the same time is represented by Support(A∩B)/Support(A), and all data items whose support and reliability meet the preset standards are collected to form meaningful association laws.

In a certain embodiment, the present invention chooses to apply deep learning technology to optimize association rules. The deep learning technology described in the present invention preferably refers to, for example but not limited to: neural network (ANN), deep neural network ( DNN), Recurrent Neural Network (RNN), Convolutional Neural Network (CNN), Convolutional Recurrent Neural Network (CRNN), Generative Adversarial Network (GAN), Deep Belief Network (DBN), Fully Convolutional Neural Network Network (FCN), Multicolumn Convolutional Neural Network (MCNN), Recurrent Neural Network (RNN), Long Short-Term Memory Model (LSTM), Bidirectional Neural Network (BRNN), Deep Recurrent Neural Network (DRNN), residual network (DRN), restricted Boltzmann machine (RBM), multi-layer perception (MLP), autoencoder, attention network, ensemble learning, unsupervised classification methods, Supervised classification method, boosted tree method, gradient boosted tree method, strong gradient boosting machine method, weak gradient boosting machine method, regression tree method, random forest method, decision tree method, weak learning method, strong learning method, strong voting method, Weak voting methods, support vector machines (support vector machines) classifiers, or rule-based methods, etc.

Fig. 6 discloses a schematic diagram of the platform user interface provided to the user by the project project file management platform of the present invention through the front-end display layer; the project project file management platform of the present invention is provided to users by executing the front-end components on the front-end display layer 100 The provider provides a series of platform user interfaces. By accessing these platform user interfaces on the user device, users can operate the engineering project document management platform and upload electronic documents to the engineering project document management platform.

As shown in Figure 6, after the user starts the browser on the user device and enters the correct Uniform Resource Locator (URL) in the address bar, he can access the file upload user disclosed in Figure 6 Interface 130, the user uploads the electronic file to the engineering project file management platform according to the instructions in the file upload user interface 130, and the natural language parser will automatically execute to identify the text contained in the electronic file, and according to the finally recognized professional Name or entity, assign documents to related parties, and automatically include them in the engineering project database to which they belong.

The front-end component of the engineering project file management platform of the present invention also includes a natural language file query component, which is mainly in the user interface provided by the platform front-end component, for example: the file upload user interface 130 disclosed in Figure 6, embedded in a row of queries field to provide users with natural language input words, the natural language document query component will execute the keyword detection algorithm based on association rules according to the input words, and find matching keywords from the keyword database keywords, and according to the instructions in the keyword database, link to the corresponding electronic documents in the document database, and then access the corresponding electronic documents for users to read and view, users can quickly find out from a large number of documents required files, greatly reducing the time for file searches.

The 7th figure discloses the block diagram of the operating principle of the engineering project document management system of the present invention; the engineering project document management system 10 proposed by the present invention has its overall operating principle as disclosed in the 7th figure. First, the learning and training of the natural language parser is carried out. The paper document used as a template is converted into an electronic document by taking photos, scanning or other means, or directly built as an electronic document, uploaded to the server on the cloud platform and used as a training set, and then cooperates with the natural language rule data set and depth Learning datasets are used to train natural language parsers or provide natural language parser learning to generate association rules and establish keyword databases, relational person databases, and engineering project databases.

After the natural language parser completes the study, the user begins to upload the electronic files to the engineering project file management platform of the present invention. keywords, and the identified keywords are used to update the keyword database, and then according to the association rules, the electronic files are automatically sent to the related parties, and filed into the specific project management items of the engineering projects to which they belong. Users can also reverse the To quickly find the corresponding electronic documents in the document database by operating natural language document query components, and then perform subsequent access, retrieval or viewing, users can quickly find the required documents from a large number of documents, greatly reducing document searches time.

Figure 8 discloses a flow chart of the implementation steps of the engineering project file management method of the present invention; in summary, the engineering project file management method 500 of the present invention preferably includes the following steps: creating multiple electronic files containing plural characters and uploading them for engineering project file management Files to which the platform belongs Database, the engineering project document management platform includes a natural language parser (step 501); implement character detection through the character detection component included in the natural language parser to determine the location of the text (step 502) ; implement character recognition through the character recognition component included in the natural language parser to identify the detected characters (step 503); implement named entity recognition through the named entity recognition component included in the natural language parser To identify the keywords included in the recognized text and store them in the keyword database (step 504); assign the electronic documents to at least one related person according to the keywords and the association rules and classified into at least one project management project (step 505); and provide a natural language document query component to provide users with natural language to directly search the keywords in the keyword database, and link to the queried electronic files, so as to find the queried electronic files from the electronic files included in the file database (step 506).

The above embodiments of the present invention can be arbitrarily combined or replaced with each other, thereby deriving more implementation forms, but none of them depart from the scope of protection intended by the present invention. More embodiments of the present invention are further provided as follows:

Embodiment 1: A method for managing engineering project documents, which includes: creating multiple electronic documents containing plural characters and uploading them to the engineering project document management platform; using the natural language parser included in the engineering project document management platform to identify the electronic documents Include multiple keywords and store them in a keyword database; assign these electronic documents to at least one related person and classify them into at least one engineering management project according to the keywords and according to the association rules; and provide natural language documents The query element is used for the user to find the corresponding electronic document by searching the keywords in the keyword database.

Embodiment 2: The engineering project file management method as described in Embodiment 1, also includes One of the following: upload the electronic documents to the document database; implement character detection through the character detection components contained in the natural language parser to determine the location of the characters; analyze the natural language The character recognition component included in the natural language parser implements character recognition to identify the detected characters; through the named entity recognition component included in the natural language parser, the named entity recognition is implemented to identify the recognized characters include the keywords; gather the identified keywords to build the keyword database; and provide a natural language document query component to allow users to use natural language to directly search for the keywords in the keyword database words, and link to the inquired electronic document accordingly, so as to find the inquired electronic document from among the electronic documents contained in the document database.

Embodiment 3: The engineering project file management method as described in Embodiment 2, further comprising one of the following: the character detection element is implemented through a morphological operation method, MSER method, NMS method, CTPN method, SegLink method , EAST method, R-CNN method, fast RCNN method, PSENet method and a combination thereof to detect the location of the characters; the character recognition component recognizes the detected characters by implementing a deep learning method etc., the deep learning method is convolutional neural network, deep convolutional neural network, recurrent neural network, convolutional recurrent neural network, convolutional recurrent neural network optical character recognition, attention optical character Recognition and a combination thereof; the named entity recognition component implements a rule-based method, an unsupervised learning method, a feature-based supervised learning method, the deep learning method, and combinations thereof by referring to the natural language rule data set one of them, to establish the keyword database; and the named entity recognition component implements text segmentation operation, word segmentation marking operation, part-of-speech marking operation, entity marking operation, entity extraction, Proper name extraction, anaphora resolution, relation extraction or grammatical analysis to build the keyword database.

Embodiment 4: The engineering project document management method as described in embodiment 3 also includes one of the following: establishing a deep learning data set to train the deep learning method and provide the deep learning method for learning; establishing a natural language rule data set , to provide the named entity recognition element as a reference; use the deep learning data set and the natural language rule data set as a training set to train the deep learning method; establish a relational person database, and the relational person database includes plural relational persons and the keywords corresponding to each of these related parties; establish a project project database, which contains multiple project management projects and keywords corresponding to each of these project management projects; implement the algorithm of association rules to Establishing an association rule of the keywords with respect to each of the related persons in the related person database and each of the engineering management items in the engineering project database; and optimizing the association rule by applying the deep learning method.

Embodiment 5: The engineering project file management method as described in embodiment 4, wherein the deep learning method is a neural network, a deep neural network, a recurrent neural network, a convolutional neural network, a convolutional recurrent neural network Road, Generative Adversarial Networks, Deep Belief Networks, Fully Convolutional Neural Networks, Multi-column Convolutional Neural Networks, Recurrent Neural Networks, Long Short-Term Memory Models, Bidirectional Neural Networks, Deep Recurrent Neural Networks, Residual Networks, Restricted Boltzmann Machines, Multilayer Perception, Autoencoders, Attention Networks, Ensemble Learning, Unsupervised Classification Methods, Supervised Classification Methods, Rule-Based Methods, Boosted Tree Methods, Gradient Boosted Tree Methods, Strong gradient boosting machine methods, weak gradient boosting machine methods, regression tree methods, random forest methods, decision tree methods, weak learning methods, strong learning methods, strong voting methods, weak voting methods, or support vector machine classifiers.

Embodiment 6: The engineering project document management method as described in Embodiment 1, wherein the electronic documents are construction records, construction logs, construction photos, meeting records, supervisory inspections, independent checklists or contract documents.

Embodiment 7: An engineering project file management system, which includes: a back-end data layer for storing natural language rule data sets, deep learning data sets, keyword databases, related person databases, engineering project databases, and document data One of the libraries; the relay logic layer, which executes the project file management platform and the natural language parser contained in the project file management platform; and the front-end display layer, which executes the project file management platform. A language file query component, wherein the user uploads multiple electronic files containing multiple characters to the engineering project file management platform through the front-end component on the user device, so that the natural language parser can identify the multiple keys contained in the electronic files words, and store them in the keyword database, so that the engineering project document management platform can assign these electronic documents to at least one related person and classify them into engineering management projects according to the keywords and the association rules, and through the The natural language document query component searches the keywords in the keyword database to find the corresponding electronic documents.

Embodiment 8: The engineering project document management system as described in Embodiment 7, wherein the back-end data layer includes at least one or more database servers, and the relay logic layer includes at least one or more program servers , the front-end display layer includes at least one or more web servers, wherein the program servers distribute the workload through the first server load balancing device, and the web servers are distributed through the second server load balancing device And distribute the workload.

Embodiment 9: The engineering project document management system as described in Embodiment 7, wherein the file format of these electronic files is PDF file format, Power Point file format, Power Point file compatible format, WORD file format, WORD file compatible format , Excel file format, Excel file compatible format, JPG file format, JPEG file format or PNG file format.

Embodiment 10: The engineering project document management system as described in Embodiment 7, wherein The user device is a desktop computer, a notebook computer, a tablet device or a smart phone.

The various embodiments of the present invention can be combined or replaced arbitrarily with each other, thereby deriving more implementation forms, but none of them depart from the intended protection scope of the present invention, and the definition of the protection scope of the present invention is fully defined by the patent scope of the present invention application The recorder shall prevail.

500: Project file management method of the present invention

501-506: Implementation steps

Claims (9)

A method for managing engineering project documents is applied to automatic management of multiple electronic engineering project documents, which includes: creating multiple electronic engineering project documents containing plural characters by operating a user device set on a display layer and operating in A front-end component executed on the user device uploads the electronic engineering project documents to an engineering project document management platform and a document database; by executing the engineering project document management platform set on a logical layer includes one of A character detection component included in the natural language parser implements a character detection to determine the location of the characters, and executes a character recognition component included in the natural language parser to perform a character recognition to identify all The detected words, and executing a named entity recognition component included in the natural language parser to implement a named entity recognition to identify the keywords included in the recognized words, thereby identifying the The plural keywords contained in the electronic engineering project document are stored in a keyword database pre-established by implementing a named entity recognition method; the identified keywords are stored in the keyword database; by executing The engineering project document management platform includes a back-end component to access an association rule set in a data layer, and assign the electronic engineering project documents to at least one related party according to the keywords and according to the association rule And categorized and stored in at least one project management project; and provide a natural language document query element for a user to directly search the keywords in the keyword database by using natural language, so as to obtain from the document database Find the queried electronic files in the included electronic engineering project files. The engineering project file management method described in claim 1 further includes one of the following: the character detection component is implemented by implementing a morphological operation method, an MSER method, and an NMS method, a CTPN method, a SegLink method, an EAST method, a R-CNN method, a fast RCNN method, a PSENet method and a combination thereof to detect the location of the characters; the character recognition element Recognizing the detected characters by implementing a deep learning method, the deep learning method being a convolutional neural network, a deep convolutional neural network, a recurrent neural network, a convolutional recurrent neural network One of a method, a convolutional recurrent neural network optical character recognition, an attentional optical character recognition, and combinations thereof; the named entity recognition element implements an applied rule-based method by referring to the natural language rule dataset, one of an unsupervised learning method, a feature-based supervised learning method, the deep learning method, and a combination thereof to build the keyword database; and the named entity recognition element is generated by referring to the natural language rule dataset performing a text segmentation operation, a word segmentation marking operation, a part-of-speech marking operation, an entity marking operation, an entity extraction, a proper name extraction, an anaphora resolution operation, a relation extraction operation or a grammatical analysis operation, to build the keyword database. The engineering project document management method as described in request item 2 also includes one of the following: pre-establish a deep learning data set in the data layer to train the deep learning method and provide learning for the deep learning method; Create a natural language rule data set in the data layer to provide the named entity recognition element as a reference; use the deep learning data set and the natural language rule data set as a training set to train the deep learning method; Establish a database of related parties, the database of related parties includes multiple related parties and keywords corresponding to each of the related parties; Establish an engineering project database in the data layer in advance, the engineering project database includes multiple engineering management items and keywords corresponding to each of these engineering management items; implement an association rule algorithm to create these keywords The association rule with respect to each of the related parties in the related party database and each of the engineering management items in the engineering project database; and optimizing the association rule by applying the deep learning method. The engineering project file management method as described in claim 3, wherein the deep learning method is a type of neural network, a deep neural network, a recurrent neural network, a convolutional neural network, and a convolutional recurrent neural network Road, a generative confrontation network, a deep belief network, a fully convolutional neural network, a multi-column convolutional neural network, a recurrent neural network, a long short-term memory model, a bidirectional neural network, a Deep recurrent neural network, a residual network, a restricted Boltzmann machine, a multi-layer perception, an autoencoder, an attention network, an ensemble learning, an unsupervised classification method, a supervised classification method, a rule-based method, a boosted tree method, a gradient boosted tree method, a strong gradient boosting machine method, a weak gradient boosting machine method, a regression tree method, a random forest method, a decision tree method, a A weak learning method, a strong learning method, a strong voting method, a weak voting method or a support vector machine classifier. The engineering project document management method as described in Claim 1, wherein the electronic engineering project document is a construction record, a construction log, a construction photo, a meeting record, a supervisory inspection, an independent checklist or a contract document . An engineering project file management system is applied to the automatic management of multiple electronic engineering project files, which includes: A back-end data layer for storing one of a natural language rule data set, a deep learning data set, a keyword database, a relational person database, an engineering project database, and a document database; Following the logic layer, it implements a project document management platform and a natural language parser included in the project document management platform; and a front-end display layer, which implements a front-end component and a natural language included in the project document management platform A document query component, wherein a user uploads a plurality of electronic engineering project documents containing plural characters to the engineering project document management platform and the document database through the front-end component on a user device for the natural language parser executing a character detection component included to perform a character detection to determine the location of the characters, executing a character recognition component included to perform a character recognition to identify the detected characters, and executing a named entity recognition component to implement a named entity recognition to identify the keywords contained in the recognized text, thereby identifying the plural keywords contained in the electronic engineering project documents, and storing to the keyword database, so that the engineering project document management platform assigns the electronic engineering project documents to at least one related person according to the keywords and according to an association rule, and classifies and stores them in an engineering management project, and Provide the user with natural language to directly search the keywords in the keyword database through the natural language document query element, so as to find the queried electronic documents from the electronic engineering project documents included in the document database . The engineering project document management system as described in claim 6, wherein the back-end data layer includes at least one or more database servers, the relay logic layer includes at least one or more program servers, and the front-end display The tier includes at least one or more web servers, wherein the program servers distribute workload through a first server load balancing device, the web servers The workload is distributed through a second server load balancing device. The engineering project document management system as described in claim item 6, wherein the file format of the electronic engineering project documents is a PDF file format, a Power Point file format, a Power Point file compatible format, a WORD file format, a WORD file format file compatible format, an Excel file format, an Excel file compatible format, a JPG file format, a JPEG file format or a PNG file format. The engineering project document management system as described in Claim 6, wherein the user device is a desktop computer, a notebook computer, a tablet device or a smart phone.

Images