File digital intelligent management system

Abstract

The invention relates to the technical field of digital archive management, and discloses an archive digitization intelligent management system, which comprises a document digitization module used for converting a paper document into a digital document through an OCR (Optical Character Recognition) technology; the semantic label extraction module is used for extracting semantic labels of the digital documents through a semantic analysis model; the document storage module is used for constructing a knowledge graph according to the semantic tag of the digital document and storing the knowledge graph into a graph database; the document retrieval module is used for inputting retrieval entries of a user into the graph database and outputting digital documents conforming to the retrieval entries; according to the method, the semantic label of the digital document is automatically extracted through the semantic analysis model, the knowledge graph is automatically constructed according to the semantic label, and the retrieval entry of the user is converted into word vector representation, so that the accuracy of digital document retrieval can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of digital archive management, and more specifically, to a digitalized intelligent archive management system. Background Art

[0002] Traditional archive management mainly relies on the physical storage and manual operation of paper documents. Staff need to manually add tags or indexes for subsequent reference. However, traditional archive management has the following obvious defects: 1. Paper documents not only require a lot of space for storage, but are also vulnerable to natural disasters and man-made damage, which puts paper documents at risk of loss and damage; 2. Traditional manual retrieval lacks semantic understanding of user needs, resulting in manual retrieval results that do not match user needs.

[0003] Therefore, there is an urgent need for a digital archive management system to solve the above problems. Summary of the invention

[0004] The present invention provides a digital intelligent archive management system to solve the technical problems in the above-mentioned background technology.

[0005] The present invention provides a digitalized intelligent archive management system, comprising:

[0006] A document digitization module, which is used to convert paper documents into digital documents through OCR technology;

[0007] Each digital document is assigned a unique number;

[0008] A semantic tag extraction module, which is used to extract semantic tags of digital documents through a semantic analysis model;

[0009] Semantic tags include: document name, field, and publication information;

[0010] A document storage module, which is used to construct a knowledge graph based on the semantic tags of digital documents and store the knowledge graph in a graph database;

[0011] The knowledge graph includes: entities, attributes of entities, and edges between entities;

[0012] Establish a mapping relationship between the entity and the unique number of the digital document;

[0013] The attributes of the entity are generated based on the labels of the digital document;

[0014] Build edges between entities belonging to the same domain;

[0015] The document retrieval module is used to input the user's search terms into the graph database and output digital documents that match the search terms.

[0016] Furthermore, a unique number for the digital document is generated by a snowflake algorithm.

[0017] Furthermore, the field is a custom parameter.

[0018] Furthermore, the publication information includes: author name, publisher name and publication date.

[0019] Further, the semantic analysis model includes: a word vector embedding layer, a first hidden layer, a second hidden layer, a third hidden layer and a classifier;

[0020] The word vector embedding layer inputs a digital document and outputs a feature sequence;

[0021] Each sequence unit of the feature sequence corresponds to the word vector of a word in the digital document;

[0022] The word vector embedding layer is built based on the word vector model;

[0023] The first hidden layer inputs the feature sequence and outputs the first feature;

[0024] The second hidden layer inputs the first feature and outputs the second feature;

[0025] The third hidden layer inputs the second feature and outputs the third feature;

[0026] The third feature is input into the classifier, and the classification space of the classifier represents the semantic label of the digital document.

[0027] Furthermore, the calculation formula of the semantic analysis model includes:

[0028] Feature 3 =max(0,Feature 2 W 1 )W 2 +b 1 ;

[0029]

[0030] Feature 1 =softmax(V T *tanh[W 4 Q+W 5 K]) × V;

[0031] Q=X×W Q ;

[0032] K=X×W K ;

[0033] V=X×W V ;

[0034] Where X represents the feature sequence of the first hidden layer input, Feature 1 Represents the first feature of the first hidden layer output, Feature 2 Represents the second feature of the second hidden layer output, Feature 3 represents the third feature of the third hidden layer output, Q, K and V represent the first intermediate feature, the second intermediate feature and the third intermediate feature respectively, and W Q , W K and W V denote the first intermediate weight parameter, the second intermediate weight parameter and the third intermediate weight parameter, respectively, 1 , W 2 , W 3 , W 4 and W 5 denote the first weight parameter, the second weight parameter, the third weight parameter, the fourth weight parameter and the fifth weight parameter, respectively, and b 1 and b 2 Represent the first bias parameter and the second bias parameter respectively, T represents the transposition operation, μ and σ represent the mean and standard deviation of the first feature respectively, softmax represents the softmax activation function, and tanh represents the hyperbolic tangent activation function.

[0035] Furthermore, sample labels of training samples for the semantic analysis model are set through expert evaluation.

[0036] Furthermore, the labels of digital documents are converted into vector representations as attributes of entities through the word vector model.

[0037] Furthermore, the user's search terms are converted into vector representations through the same word vector model, and the similarity between the vector representations of the search terms and the vector representations of the attributes of the entities is calculated, and the similarities are sorted in descending order to obtain the top K digital documents, where K is a custom parameter;

[0038] The calculation formula for the similarity SSIM between the vector representation of the search term and the vector representation of the entity's attributes includes:

[0039]

[0040] A=Feature A ×W A ;

[0041] B=Feature B ×W B ;

[0042] Feature A Represents the vector representation of the search term, FeatureB A represents the transformation vector of the search terms, and B represents the transformation vector of the entity attributes. The number of dimensions of A and B are both n. A and W B denote the first conversion parameter and the second conversion parameter, respectively, A i Indicates the value of the i-th dimension of the conversion vector of the search term, B i The value of the i-th dimension of the transformation vector representing the attribute of the entity.

[0043] The present invention provides a digitalized intelligent management method for archives, comprising the following steps:

[0044] Step S201, converting a paper document into a digital document using OCR technology;

[0045] Step S202, extracting semantic tags of digital documents through a semantic analysis model;

[0046] Step S203, constructing a knowledge graph according to the semantic tags of the digital document, and storing the knowledge graph in a graph database;

[0047] Step S204: input the user's search terms into the graph database, and output digital documents that match the search terms.

[0048] The beneficial effects of the present invention are as follows: the present invention automatically extracts semantic tags of digital documents through a semantic analysis model, automatically constructs a knowledge graph based on the semantic tags, and converts user search terms into word vector representations, which can improve the accuracy of digital document retrieval. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic diagram of a digitalized intelligent archive management system of the present invention;

[0050] Figure 2 It is a flow chart of a digitalized intelligent management method of archives of the present invention.

[0051] In the figure: document digitization module 101, semantic tag extraction module 102, document storage module 103, document retrieval module 104. DETAILED DESCRIPTION

[0052] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0053] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in one or more embodiments of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0054] like Figure 1-2 As shown, a digital intelligent archive management system includes:

[0055] A document digitization module 101, which is used to convert paper documents into digital documents by using OCR technology;

[0056] Each digital document is assigned a unique number;

[0057] A semantic tag extraction module 102, which is used to extract semantic tags of digital documents through a semantic analysis model;

[0058] Semantic tags include: document name, field, and publication information;

[0059] The document storage module 103 is used to construct a knowledge graph according to the semantic tags of the digital documents and store the knowledge graph in a graph database;

[0060] The knowledge graph includes: entities, attributes of entities, and edges between entities;

[0061] Establish a mapping relationship between the entity and the unique number of the digital document;

[0062] The attributes of the entity are generated based on the labels of the digital document;

[0063] Build edges between entities belonging to the same domain;

[0064] The document retrieval module 104 is used to input the user's search terms into the graph database and output digital documents that meet the search terms.

[0065] In one embodiment of the present invention, a unique number for a digital document is generated by a snowflake algorithm.

[0066] In one embodiment of the present invention, the unique number of the digital document may also be generated through UUID or binary encoding.

[0067] In one embodiment of the present invention, the field is a custom parameter. For example, the field can be set according to an industry classification list, such as mining, manufacturing, construction, finance, information transmission, software and information technology services, etc.

[0068] In one embodiment of the present invention, the publication information includes: author name, publisher name and publication date.

[0069] In one embodiment of the present invention, the semantic tag may also include: document summary, document type, and cited documents.

[0070] In one embodiment of the present invention, the semantic analysis model includes: a word vector embedding layer, a first hidden layer, a second hidden layer, a third hidden layer and a classifier;

[0071] The word vector embedding layer inputs a digital document and outputs a feature sequence;

[0072] Each sequence unit of the feature sequence corresponds to the word vector of a word in the digital document;

[0073] The word vector embedding layer is built based on the word vector model, which can be Word2Vec, FastText, or BERT.

[0074] The first hidden layer inputs the feature sequence and outputs the first feature;

[0075] The second hidden layer inputs the first feature and outputs the second feature;

[0076] The third hidden layer inputs the second feature and outputs the third feature;

[0077] The third feature is input into the classifier, and the classification space of the classifier represents the semantic label of the digital document.

[0078] In one embodiment of the present invention, the calculation formula of the semantic analysis model includes:

[0079] Feature 3 =max(0,Feature 2 W 1 )W 2 +b 1 ;

[0080]

[0081] Feature 1 =softmax(V T*tanh[W 4 Q+W 5 K]) × V;

[0082] Q=X×W Q ;

[0083] K=X×W K ;

[0084] V=X×W V ;

[0085] Where X represents the feature sequence of the first hidden layer input, Feature 1 Represents the first feature of the first hidden layer output, Feature 2 Represents the second feature of the second hidden layer output, Feature 3 represents the third feature of the third hidden layer output, Q, K and V represent the first intermediate feature, the second intermediate feature and the third intermediate feature respectively, and W Q , W K and W V denote the first intermediate weight parameter, the second intermediate weight parameter and the third intermediate weight parameter, respectively, 1 , W 2 , W 3 , W 4 and W 5 denote the first weight parameter, the second weight parameter, the third weight parameter, the fourth weight parameter and the fifth weight parameter, respectively, and b 1 and b 2 Represent the first bias parameter and the second bias parameter respectively, T represents the transposition operation, μ and σ represent the mean and standard deviation of the first feature respectively, softmax represents the softmax activation function, and tanh represents the hyperbolic tangent activation function.

[0086] It should be noted that the weight parameters and bias parameters in the semantic analysis model are all learnable parameters. The gradient information of the weight parameters and bias parameters in the semantic analysis model is calculated by the back propagation algorithm. The weight parameters and bias parameters in the semantic analysis model are updated according to the gradient information with the help of the gradient descent method, so as to iterate the semantic analysis model until the loss is minimized.

[0087] In one embodiment of the present invention, sample labels of training samples for the semantic analysis model are set through expert evaluation.

[0088] It should be noted that the sample labels of the training samples correspond to the semantic labels of the digital documents, and the semantic analysis model after training can automatically extract the semantic labels of the digital documents.

[0089] In one embodiment of the present invention, the calculation formula of the loss function Loss of the semantic analysis model is as follows:

[0090]

[0091] Where T represents the number of training samples, y <t represents the sample label of the tth training sample, y t Represents the semantic label output from the input of the tth training sample to the semantic analysis model.

[0092] In one embodiment of the present invention, the tags of digital documents are converted into vector representations as attributes of entities through a word vector model.

[0093] In one embodiment of the present invention, the user's search terms are converted into vector representations through the same word vector model, and the similarity between the vector representations of the search terms and the vector representations of the attributes of the entity is calculated, and the similarities are sorted in descending order to obtain the top K digital documents, where K is a custom parameter, and preferably, K is set to 5.

[0094] In one embodiment of the present invention, the calculation formula of the similarity SSIM between the vector representation of the search term and the vector representation of the attribute of the entity includes:

[0095]

[0096] A=Feature A ×W A ;

[0097] B=Feature B ×W B ;

[0098] Feature A Represents the vector representation of the search term, Feature B A represents the transformation vector of the search terms, and B represents the transformation vector of the entity attributes. The number of dimensions of A and B are both n. A and W B denote the first conversion parameter and the second conversion parameter, respectively, A i Indicates the value of the i-th dimension of the conversion vector of the search term, B i The value of the i-th dimension of the transformation vector representing the attribute of the entity.

[0099] In one embodiment of the present invention, Figure 2 As shown, the present invention provides a method for intelligent management of digital archives, comprising the following steps:

[0100] Step S201, converting a paper document into a digital document using OCR technology;

[0101] Step S202, extracting semantic tags of digital documents through a semantic analysis model;

[0102] Step S203, constructing a knowledge graph according to the semantic tags of the digital document, and storing the knowledge graph in a graph database;

[0103] Step S204: input the user's search terms into the graph database, and output digital documents that match the search terms.

[0104] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are within the protection of the present embodiment.

Claims

1. A digital and intelligent archive management system. It is characterized in that include: A document digitization module, which is used to convert paper documents into digital documents through OCR technology; Each digital document is assigned a unique number; A semantic tag extraction module, which is used to extract semantic tags of digital documents through a semantic analysis model; Semantic tags include: document name, field, and publication information; A document storage module, which is used to construct a knowledge graph based on the semantic tags of digital documents and store the knowledge graph in a graph database; The knowledge graph includes: entities, attributes of entities, and edges between entities; Establish a mapping relationship between the entity and the unique number of the digital document; The attributes of the entity are generated based on the labels of the digital document; Build edges between entities belonging to the same domain; The document retrieval module is used to input the user's search terms into the graph database and output digital documents that match the search terms.

2. According to claim 1, a digital intelligent archives management system, It is characterized in that A unique number for a digital document is generated using a snowflake algorithm.

3. According to claim 1, a digital intelligent archives management system, It is characterized in that The field is custom parameters.

4. According to claim 1, a digital intelligent archives management system, It is characterized in that Publication information includes: author name, publisher name and publication date.

5. According to claim 1, a digital intelligent archives management system, It is characterized in that The semantic analysis model includes: a word vector embedding layer, a first hidden layer, a second hidden layer, a third hidden layer and a classifier; The word vector embedding layer inputs a digital document and outputs a feature sequence; Each sequence unit of the feature sequence corresponds to the word vector of a word in the digital document; The word vector embedding layer is built based on the word vector model; The first hidden layer inputs the feature sequence and outputs the first feature; The second hidden layer inputs the first feature and outputs the second feature; The third hidden layer inputs the second feature and outputs the third feature; The third feature is input into the classifier, and the classification space of the classifier represents the semantic label of the digital document.

6. According to claim 5, a digital intelligent archives management system, It is characterized in that The calculation formula of the semantic analysis model includes: Feature 3 =max(0,Feature 2 W 1 )W 2 +b 1 ; Feature 1 =softmax(V T *tanh[W 4 Q+W 5 K])×V; Q=X×W Q ; K=X×W K ; H=X×W V ; Where X represents the feature sequence of the first hidden layer input, Feature 1 Represents the first feature of the first hidden layer output, Feature 2 Represents the second feature of the second hidden layer output, Feature 3 represents the third feature of the third hidden layer output, Q, K and V represent the first intermediate feature, the second intermediate feature and the third intermediate feature respectively, and W Q , W K and W V denote the first intermediate weight parameter, the second intermediate weight parameter and the third intermediate weight parameter, respectively, 1 , W 2 , W 3 , W 4 and W 5 denote the first weight parameter, the second weight parameter, the third weight parameter, the fourth weight parameter and the fifth weight parameter, respectively, and b 1 and b 2 Represent the first bias parameter and the second bias parameter respectively, T represents the transposition operation, μ and σ represent the mean and standard deviation of the first feature respectively, softmax represents the softmax activation function, and tanh represents the hyperbolic tangent activation function.

7. According to claim 1, a digital intelligent archives management system, It is characterized in that The sample labels of the training samples used for the semantic analysis model are set through expert evaluation.

8. According to claim 1, a digital intelligent archives management system, It is characterized in that The labels of digital documents are converted into vector representations as attributes of entities through the word vector model.

9. According to claim 8, a digital intelligent archives management system, It is characterized in that The user's search terms are converted into vector representations through the same word vector model, and the similarity between the vector representation of the search terms and the vector representation of the entity's attributes is calculated. The similarities are sorted in descending order to obtain the top K digital documents, where K is a custom parameter; The calculation formula for the similarity SSIM between the vector representation of the search term and the vector representation of the entity's attributes includes: A=Feature A ×W A ; B=Feature B ×W B ; Feature A Represents the vector representation of the search term, Feature B A represents the transformation vector of the search terms, and B represents the transformation vector of the entity attributes. The number of dimensions of A and B are both n. A and W B denote the first conversion parameter and the second conversion parameter, respectively, A i Indicates the value of the i-th dimension of the conversion vector of the search term, B i The value of the i-th dimension of the transformation vector representing the attribute of the entity.

10. A method for intelligent digital management of archives. It is characterized in that Executing a digitalized intelligent archive management system as claimed in any one of claims 1 to 9 comprises the following steps: Step S201, converting a paper document into a digital document using OCR technology; Step S202, extracting semantic tags of digital documents through a semantic analysis model; Step S203, constructing a knowledge graph according to the semantic tags of the digital document, and storing the knowledge graph in a graph database; Step S204: input the user's search terms into the graph database, and output digital documents that match the search terms.

Images