Digital management method and system based on big data analysis

Abstract

The invention discloses a digital management method and system based on big data analysis, and belongs to the technical field of text processing. Text content in a document is divided into sentences, sentence-by-sentence analysis is achieved, characters in each sentence of text are combined to obtain three character combinations, and three types of phrase sets are formed respectively; each type of phrase set reflects different character combinations, different character combinations have different levels of meanings, calculating the significance of each phrase, constructing three types of significance sequences, extracting different levels of meanings by adopting a multi-layer semantic classification model, and predicting the importance of each sentence of text in combination with the significance of the phrases. According to the method, the importance degree of the document is obtained by combining the importance degree of each sentence of text in the document, and the document is stored according to the importance degree of the document, so that classified storage of documents with different importance degrees is realized, document management is facilitated, and the efficiency and accuracy of document management are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of text processing, and in particular to a digital management method and system based on big data analysis. Background Art

[0002] In the digital age, enterprises and organizations are faced with the challenge of managing massive amounts of information, especially the rapid growth of document data, which makes it difficult for traditional management methods to cope with it. With the development of big data technology, data analysis capabilities have been significantly improved, and enterprises are in urgent need of effectively utilizing these technologies to achieve efficient processing and management of information. However, existing document management systems mostly rely on static and subjective screening criteria, which makes it easy for important information to be overlooked or misevaluated. In addition, the lack of the ability to deeply analyze text content leads to chaotic document management and low classification and storage accuracy. Summary of the invention

[0003] In view of the above-mentioned deficiencies in the prior art, the present invention provides a digital management method and system based on big data analysis to solve the problem of low classification storage accuracy in the prior art.

[0004] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a digital management method based on big data analysis, comprising the following steps: S1, dividing the text content in the document into sentences to obtain multiple sentences; S2, combining the characters in each sentence to obtain a first-category phrase set, a second-category phrase set, and a third-category phrase set; S3, calculating the prominence of each phrase in each type of phrase set, and constructing a first type prominence sequence, a second type prominence sequence, and a third type prominence sequence; S4, using a multi-layer semantic classification model to process the first type of phrase set, the second type of phrase set and the third type of phrase set, and obtaining the importance of each sentence based on the prominence applied by the first type of prominence sequence, the second type of prominence sequence and the third type of prominence sequence; S5. Weight the importance of each sentence in the document to obtain the importance of the document, and allocate storage space to the document based on the importance of the document.

[0005] Furthermore, the S2 comprises the following sub-steps: S21, slide a sliding queue with a length of 2 on each sentence of text to obtain a first-category phrase set corresponding to each sentence of text, wherein the first-category phrase set includes multiple phrases, and a phrase length is 2; S22, slide a sliding queue with a length of 3 on each sentence of text to obtain a second-category phrase subset corresponding to each sentence of text, wherein the second-category phrase subset includes multiple phrases, and the length of a phrase is 3; S23. Use a sliding queue with a length of 4 to slide on each sentence of text to obtain a third-category phrase subset corresponding to each sentence of text, wherein the third-category phrase subset includes multiple phrases, and a phrase length is 4. The sliding window advances by the length of 1 word each time, and each time it slides, a phrase is obtained.

[0006] Furthermore, S3 includes the following sub-steps: S31, extracting each phrase in each type of phrase set, and calculating the similarity between each phrase and each keyword in the keyword set; S32, calculating the key coefficient of each phrase according to the importance and maximum similarity of the keyword annotation corresponding to the maximum similarity; S33, marking the document with allocated storage space containing the phrase as a reference document; S34, counting the frequency of occurrence of the phrase in the reference document; S35, calculating the significance of the phrase according to the frequency of occurrence of the phrase in the reference document, the importance of the reference document and the key coefficient; S36. Use the prominence of each phrase as an element to construct a significance sequence, wherein, when the phrase set in S31 is a first-category phrase set, the significance sequence in S35 is a first-category significance sequence; when the phrase set in S31 is a second-category phrase set, the significance sequence in S35 is a second-category significance sequence; when the phrase set in S31 is a third-category phrase set, the significance sequence in S35 is a third-category significance sequence.

[0007] Furthermore, the formula for calculating the key coefficient of each phrase in S32 is: , where γ is the key coefficient of the phrase, θ key is the importance of the keyword annotation corresponding to the maximum similarity, and S is the maximum similarity.

[0008] Furthermore, the formula for calculating the significance of the phrase in S35 is: , where μ is the significance of the phrase, N i is the frequency of the phrase appearing in the i-th reference document, θ file,i is the importance of the i-th reference document, I is the number of reference documents, γ is the key coefficient of the phrase, and i is a positive integer.

[0009] Further, the multi-layer semantic classification model in S4 includes: a first type of semantic extraction unit, a second type of semantic extraction unit, a third type of semantic extraction unit, a first multiplier M1, a second multiplier M2, a third multiplier M3 and a classifier; The input end of the first type of semantic extraction unit is used to input a first type of phrase set; the input end of the second type of semantic extraction unit is used to input a second type of phrase set; the input end of the third type of semantic extraction unit is used to input a third type of phrase set; the first input end of the first multiplier M1 is connected to the output end of the first type of semantic extraction unit, and the second input end thereof is used to input a first type of significance sequence; the first input end of the second multiplier M2 is connected to the output end of the second type of semantic extraction unit, and the second input end thereof is used to input a second type of significance sequence; the first input end of the third multiplier M3 is connected to the output end of the third type of semantic extraction unit, and the second input end thereof is used to input a third type of significance sequence; the input end of the classifier is respectively connected to the output end of the first multiplier M1, the output end of the second multiplier M2 and the output end of the third multiplier M3, and the output end thereof serves as the output end of the multi-layer semantic classification model.

[0010] Furthermore, the first type of semantic extraction unit, the second type of semantic extraction unit and the third type of semantic extraction unit all include: multiple LiLSTM layers, multiple phrase semantic extraction layers and a Concat layer, each LiLSTM layer is used to process a phrase, a phrase semantic extraction layer is used to process the features output by a LiLSTM layer, and the Concat layer is used to splice the semantic feature values ​​output by multiple phrase semantic extraction layers into a semantic feature vector.

[0011] Furthermore, the expression of the phrase semantic extraction layer is: , where X is the semantic feature value output by the phrase semantic extraction layer, tanh is the hyperbolic tangent activation function, and h m is the mth feature output by the LiLSTM layer, ω m h m The weight of b m h m , M is the number of features output by the LiLSTM layer, and m is a positive integer.

[0012] Furthermore, the first multiplier M1, the second multiplier M2 and the third multiplier M3 are used to perform element-wise multiplication on two inputs.

[0013] A digital management system based on big data analysis, including: a sentence division system, a combination subsystem, a sequence construction subsystem, a classification subsystem and a space allocation subsystem; The sentence segmentation system is used to segment the text content in the document into sentences to obtain multiple sentence texts; The combining subsystem is used to combine the characters in each sentence to obtain a first-category phrase set, a second-category phrase set, and a third-category phrase set; The sequence construction subsystem is used to calculate the prominence of each word group in each type of word group set, and construct a first type prominence sequence, a second type prominence sequence and a third type prominence sequence; The classification subsystem is used to process the first type of phrase set, the second type of phrase set and the third type of phrase set using a multi-layer semantic classification model, and obtain the importance of each sentence of text based on the prominence applied by the first type of prominence sequence, the second type of prominence sequence and the third type of prominence sequence; The space allocation subsystem is used to weight the importance of each sentence in the document to obtain the importance of the document, and allocate storage space to the document based on the importance of the document.

[0014] The beneficial effects of the present invention are as follows: the present invention divides the text content in the document into sentences to implement sentence analysis, combines the characters in each sentence to obtain three character combinations, and respectively constitutes three types of phrase sets, each type of phrase set embodies different character combinations, and different character combinations have different levels of meanings, calculates the significance of each phrase, constructs three types of significance sequences, adopts a multi-layer semantic classification model to extract different levels of meanings, and combines the significance of the phrase to predict the importance of each sentence of text, combines the importance of each sentence of text in the document to obtain the importance of the document, stores the document according to its importance, implements classified storage of documents of different importance, facilitates document management, and greatly improves the efficiency and accuracy of document management, and the present invention combines the semantics of multi-level phrases to improve the accuracy of each sentence text analysis, has the ability to deeply analyze texts, combines the significance of phrases to prevent the loss of important information, and improves the classification storage accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flowchart of a digital management method based on big data analysis; Figure 2 It is a schematic diagram of the structure of the multi-layer semantic classification model; Figure 3 It is a schematic diagram of the structures of the first type of semantic extraction unit, the second type of semantic extraction unit and the third type of semantic extraction unit. DETAILED DESCRIPTION

[0016] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0017] Embodiment 1, as Figure 1As shown, a digital management method based on big data analysis includes the following steps: S1, dividing the text content in the document into sentences to obtain multiple sentences; S2, combining the characters in each sentence to obtain a first-category phrase set, a second-category phrase set, and a third-category phrase set; S3, calculating the prominence of each phrase in each type of phrase set, and constructing a first type prominence sequence, a second type prominence sequence, and a third type prominence sequence; S4, using a multi-layer semantic classification model to process the first type of phrase set, the second type of phrase set and the third type of phrase set, and obtaining the importance of each sentence based on the prominence applied by the first type of prominence sequence, the second type of prominence sequence and the third type of prominence sequence; S5. Weight the importance of each sentence in the document to obtain the importance of the document, and allocate storage space to the document based on the importance of the document.

[0018] In this embodiment, S2 includes the following sub-steps: S21, slide a sliding queue with a length of 2 on each sentence of text to obtain a first-category phrase set corresponding to each sentence of text, wherein the first-category phrase set includes multiple phrases, and a phrase length is 2; S22, slide a sliding queue with a length of 3 on each sentence of text to obtain a second-category phrase subset corresponding to each sentence of text, wherein the second-category phrase subset includes multiple phrases, and the length of a phrase is 3; S23. Use a sliding queue with a length of 4 to slide on each sentence of text to obtain a third-category phrase subset corresponding to each sentence of text, wherein the third-category phrase subset includes multiple phrases, and a phrase length is 4. The sliding window advances by the length of 1 word each time, and each time it slides, a phrase is obtained.

[0019] In this embodiment, step S2 is illustrated by taking "artificial intelligence is changing the world" as an example. The first category of phrase sets includes: artificial, artificial intelligence, intelligent, can, is, changing, changing, changing the world, and the world; the second category of phrase sets includes: artificial intelligence, artificial intelligence, intelligent, can, is changing, changing, changing the world, and changing the world; artificial intelligence, artificial intelligence, intelligent, can, is changing, is changing, changing the world, and changing the world; The present invention can capture semantic information of different granularities in the text through sliding windows of different lengths, which is conducive to a more comprehensive analysis of the text content.

[0020] The combination of three lengths of 2, 3 and 4 set in the present invention can fully cover most Chinese language structures, thereby retaining the original information to the greatest extent.

[0021] In this embodiment, the specific sentence division in step S1 can be performed based on the following punctuation marks and rules: basic sentence division punctuation marks: period (.), question mark (?), exclamation mark (!), ellipsis (...... or ...); special case handling: the content within quotation marks ("") is usually regarded as a complete sentence unit, and the semicolon (;) can also be used as the basis for sentence division because it usually separates sentences that are relatively independent in meaning. If the content after the colon (:) is long, it can also be considered as an independent sentence.

[0022] In this embodiment, S3 includes the following sub-steps: S31, extracting each phrase in each type of phrase set, and calculating the similarity between each phrase and each keyword in the keyword set; S32, calculating the key coefficient of each phrase according to the importance and maximum similarity of the keyword annotation corresponding to the maximum similarity; S33, marking the document with allocated storage space containing the phrase as a reference document; S34, counting the frequency of occurrence of the phrase in the reference document; S35, calculating the significance of the phrase according to the frequency of occurrence of the phrase in the reference document, the importance of the reference document and the key coefficient; S36. Use the prominence of each phrase as an element to construct a significance sequence, wherein, when the phrase set in S31 is a first-category phrase set, the significance sequence in S35 is a first-category significance sequence; when the phrase set in S31 is a second-category phrase set, the significance sequence in S35 is a second-category significance sequence; when the phrase set in S31 is a third-category phrase set, the significance sequence in S35 is a third-category significance sequence.

[0023] In this embodiment, the keyword set may be an industry standard terminology library or a keyword table defined by domain experts. The importance of keywords is obtained through manual annotation.

[0024] In this embodiment, the similarity in S31 may be cosine similarity.

[0025] In the present invention, each phrase is compared with each keyword in the keyword set to find out the importance of the keyword annotation corresponding to the maximum similarity. In combination with similar situations, the key coefficient of each phrase is calculated to evaluate the importance of the phrase. At the same time, a document containing the phrase is selected from the storage space as a reference document. According to the frequency of occurrence of the phrase in the reference document, the importance of the reference document and the key coefficient, the significance of the phrase is calculated, so that the attention of the keyword phrase is effectively enhanced in the multi-layer semantic classification model, and the accuracy of evaluating the importance of each sentence of text is improved.

[0026] In this embodiment, the formula for calculating the key coefficient of each phrase in S32 is: , where γ is the key coefficient of the phrase, θ key is the importance of the keyword annotation corresponding to the maximum similarity, and S is the maximum similarity.

[0027] When the maximum similarity is less than 0.5, the present invention assigns a key coefficient of 1, and does not change the attention of the corresponding phrase in the multi-layer semantic classification model. When the maximum similarity is greater than or equal to 0.5, the greater the maximum similarity and the higher the importance of the keyword annotation corresponding to the maximum similarity, the greater the key coefficient, and the higher the attention received by the corresponding phrase in the multi-layer semantic classification model.

[0028] In this embodiment, the formula for calculating the significance of a phrase in S35 is: , where μ is the significance of the phrase, N i is the frequency of the phrase appearing in the i-th reference document, θ file,i is the importance of the i-th reference document, I is the number of reference documents, γ is the key coefficient of the phrase, and i is a positive integer.

[0029] When the key coefficient is 1, it means that the word is not a keyword. Therefore, it is not key in itself and may be a commonly used conjunction. Its significance is assigned 1. When the key coefficient is greater than 1, the higher the importance of the reference document and the higher the frequency of the phrase in the reference document, the greater the significance. In the multi-layer semantic classification model, the corresponding phrase will receive more attention.

[0030] like Figure 2 As shown, the multi-layer semantic classification model in S4 includes: a first type of semantic extraction unit, a second type of semantic extraction unit, a third type of semantic extraction unit, a first multiplier M1, a second multiplier M2, a third multiplier M3 and a classifier; The input end of the first type of semantic extraction unit is used to input a first type of phrase set; the input end of the second type of semantic extraction unit is used to input a second type of phrase set; the input end of the third type of semantic extraction unit is used to input a third type of phrase set; the first input end of the first multiplier M1 is connected to the output end of the first type of semantic extraction unit, and the second input end thereof is used to input a first type of significance sequence; the first input end of the second multiplier M2 is connected to the output end of the second type of semantic extraction unit, and the second input end thereof is used to input a second type of significance sequence; the first input end of the third multiplier M3 is connected to the output end of the third type of semantic extraction unit, and the second input end thereof is used to input a third type of significance sequence; the input end of the classifier is respectively connected to the output end of the first multiplier M1, the output end of the second multiplier M2 and the output end of the third multiplier M3, and the output end thereof serves as the output end of the multi-layer semantic classification model.

[0031] The present invention constructs each sentence of text into three types of phrase sets, realizes multi-angle and multi-level analysis of the text, improves the comprehensiveness and accuracy of semantic understanding, introduces a saliency sequence to weight the semantic extraction results (through M1, M2, M3 operations), so that more important or more relevant semantic information is highlighted, and by inputting the three-way processing results into the same classifier, realizes comprehensive integration and analysis of text information, and improves the accuracy of classification.

[0032] In this embodiment, the classifier may adopt a BP neural network or a fully connected layer.

[0033] like Figure 3 As shown, the first type of semantic extraction unit, the second type of semantic extraction unit and the third type of semantic extraction unit all include: multiple LiLSTM layers, multiple phrase semantic extraction layers and a Concat layer, each LiLSTM layer is used to process a phrase, a phrase semantic extraction layer is used to process the features output by a LiLSTM layer, and the Concat layer is used to splice the semantic feature values ​​output by multiple phrase semantic extraction layers into a semantic feature vector.

[0034] In the first type of semantic extraction unit, the number of LiLSTM layers and phrase semantic extraction layers is equal to the number of phrases in the first type of phrase set; in the second type of semantic extraction unit, the number of LiLSTM layers and phrase semantic extraction layers is equal to the number of phrases in the second type of phrase set; in the third type of semantic extraction unit, the number of LiLSTM layers and phrase semantic extraction layers is equal to the number of phrases in the third type of phrase set.

[0035] A cell in the LiLSTM layer is used to encode an encoding value of the encoding vector of the input phrase.

[0036] In this embodiment, in the first category semantic extraction unit, the second category semantic extraction unit and the third category semantic extraction unit, the LiLSTM layer and the phrase semantic extraction layer are set according to the maximum number. For example, in a document, the longest sentence therein contains 20 words, the first category phrase set contains 19 phrases, the second category phrase set contains 18 phrases, and the third category phrase set contains 17 phrases. In this case, 19 LiLSTM layers and phrase semantic extraction layers are set in the first category semantic extraction unit, 18 LiLSTM layers and phrase semantic extraction layers are set in the second category semantic extraction unit, and 17 LiLSTM layers and phrase semantic extraction layers are set in the third category semantic extraction unit. For other sentences with less than 20 words, 0 padding is used.

[0037] In order to facilitate the processing of sentence texts of different lengths, when inputting the first-category semantic extraction unit, the second-category semantic extraction unit and the third-category semantic extraction unit, 0 can be used to make the length of each sentence consistent.

[0038] In this embodiment, the expression of the phrase semantic extraction layer is: , where X is the semantic feature value output by the phrase semantic extraction layer, tanh is the hyperbolic tangent activation function, and h m is the mth feature output by the LiLSTM layer, ω m h m The weight of b m h m , M is the number of features output by the LiLSTM layer, and m is a positive integer.

[0039] In this embodiment, one word group corresponds to one semantic feature value to form a semantic feature vector, and the semantic feature vector is multiplied element by element with the saliency sequence to achieve attention to the word group.

[0040] In this embodiment, the first multiplier M1, the second multiplier M2 and the third multiplier M3 are used to perform element-wise multiplication on two inputs.

[0041] In this embodiment, the formula for weighting the importance of each sentence in the document is: , where H is the importance of the document, h k is the importance of the kth sentence in the document, K is the number of sentences in the document, and k is a positive integer.

[0042] Embodiment 2, a digital management system based on big data analysis, comprising: a sentence division system, a combination subsystem, a sequence construction subsystem, a classification subsystem and a space allocation subsystem; The sentence segmentation system is used to segment the text content in the document into sentences to obtain multiple sentence texts; The combining subsystem is used to combine the characters in each sentence to obtain a first-category phrase set, a second-category phrase set, and a third-category phrase set; The sequence construction subsystem is used to calculate the prominence of each word group in each type of word group set, and construct a first type prominence sequence, a second type prominence sequence and a third type prominence sequence; The classification subsystem is used to process the first type of phrase set, the second type of phrase set and the third type of phrase set using a multi-layer semantic classification model, and obtain the importance of each sentence of text based on the prominence applied by the first type of prominence sequence, the second type of prominence sequence and the third type of prominence sequence; The space allocation subsystem is used to weight the importance of each sentence in the document to obtain the importance of the document, and allocate storage space to the document based on the importance of the document.

[0043] The specific implementation process of Example 2 is the same as that of Example 1.

[0044] In Examples 1 and 2, the document importance range is set to 0-100 points, and the storage space is divided into the following levels: 90-100 points: Allocate high-performance SSD storage, configure 1TB space, and set up automatic backup; 70-89 points: Allocate ordinary SSD storage, configure 500GB space, and back up weekly; 50-69 points: Allocate mechanical hard disk storage, configure 200GB space, and back up monthly; 30-49 points: Allocate mechanical hard disk storage, configure 100GB space, and back up quarterly; 0-29 points: Allocate mechanical hard disk storage, configure 50GB space, and back up half a year.

[0045] Specific examples: Enterprise strategic planning documents, with an importance of 95 points, are allocated 1TB high-performance SSDs, configured with real-time backup and multiple copies of storage; department quarterly reports, with an importance of 75 points, are allocated 500GB ordinary SSDs, backed up weekly, and stored in two copies; employee work logs, with an importance of 45 points, are allocated 100GB mechanical hard drives, backed up quarterly, and stored in a single copy.

[0046] The present invention divides the text content in a document into sentences to implement sentence analysis, combines the characters in each sentence to obtain three character combinations, and respectively constitutes three types of phrase sets, each type of phrase set embodies different character combinations, and different character combinations have different levels of meanings, calculates the significance of each phrase, constructs three types of significance sequences, adopts a multi-layer semantic classification model to extract different levels of meanings, and combines the significance of the phrase to predict the importance of each sentence of text, combines the importance of each sentence of text in the document to obtain the importance of the document, stores the document according to its importance, implements classified storage of documents of different importance, facilitates document management, and greatly improves the efficiency and accuracy of document management. The present invention combines the semantics of multi-level phrases to improve the accuracy of each sentence text analysis, and has the ability to deeply analyze texts. The present invention combines the significance of phrases to prevent the loss of important information, and improves the classification storage accuracy.

[0047] The above are only preferred embodiments of the present invention and are 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.

Claims

1. A digital management method based on big data analysis, characterized in that: The following steps are involved: S1, dividing the text content in the document into sentences to obtain multiple sentences; S2, combining the characters in each sentence to obtain a first-category phrase set, a second-category phrase set, and a third-category phrase set; S3, calculating the prominence of each phrase in each type of phrase set, and constructing a first type prominence sequence, a second type prominence sequence, and a third type prominence sequence; S4, using a multi-layer semantic classification model to process the first type of phrase set, the second type of phrase set and the third type of phrase set, and obtaining the importance of each sentence based on the prominence applied by the first type of prominence sequence, the second type of prominence sequence and the third type of prominence sequence; S5. Weight the importance of each sentence in the document to obtain the importance of the document, and allocate storage space to the document based on the importance of the document.

2. The digital management method based on big data analysis according to claim 1 is characterized in that: The S2 comprises the following sub-steps: S21, slide a sliding queue with a length of 2 on each sentence of text to obtain a first-category phrase set corresponding to each sentence of text, wherein the first-category phrase set includes multiple phrases, and a phrase length is 2; S22, slide a sliding queue with a length of 3 on each sentence of text to obtain a second-category phrase subset corresponding to each sentence of text, wherein the second-category phrase subset includes multiple phrases, and the length of a phrase is 3; S23. Use a sliding queue with a length of 4 to slide on each sentence of text to obtain a third-category phrase subset corresponding to each sentence of text, wherein the third-category phrase subset includes multiple phrases, and a phrase length is 4. The sliding window advances by the length of 1 word each time, and each time it slides, a phrase is obtained.

3. The digital management method based on big data analysis according to claim 1 is characterized in that: The S3 comprises the following sub-steps: S31, extracting each phrase in each type of phrase set, and calculating the similarity between each phrase and each keyword in the keyword set; S32, calculating the key coefficient of each phrase according to the importance and maximum similarity of the keyword annotation corresponding to the maximum similarity; S33, marking the document with allocated storage space containing the phrase as a reference document; S34, counting the frequency of occurrence of the phrase in the reference document; S35, calculating the significance of the phrase according to the frequency of occurrence of the phrase in the reference document, the importance of the reference document and the key coefficient; S36. Use the prominence of each phrase as an element to construct a significance sequence, wherein, when the phrase set in S31 is a first-category phrase set, the significance sequence in S35 is a first-category significance sequence; when the phrase set in S31 is a second-category phrase set, the significance sequence in S35 is a second-category significance sequence; when the phrase set in S31 is a third-category phrase set, the significance sequence in S35 is a third-category significance sequence.

4. The digital management method based on big data analysis according to claim 3 is characterized in that: The formula for calculating the key coefficient of each phrase in S32 is: , where γ is the key coefficient of the phrase, θ key is the importance of the keyword annotation corresponding to the maximum similarity, and S is the maximum similarity.

5. The digital management method based on big data analysis according to claim 3 is characterized in that: The formula for calculating the significance of a phrase in S35 is: , where μ is the significance of the phrase, N i is the frequency of the phrase appearing in the i-th reference document, θ file,i is the importance of the i-th reference document, I is the number of reference documents, γ is the key coefficient of the phrase, and i is a positive integer.

6. The digital management method based on big data analysis according to claim 1 is characterized in that: The multi-layer semantic classification model in S4 includes: a first type of semantic extraction unit, a second type of semantic extraction unit, a third type of semantic extraction unit, a first multiplier M1, a second multiplier M2, a third multiplier M3 and a classifier; The input end of the first type of semantic extraction unit is used to input a first type of phrase set; the input end of the second type of semantic extraction unit is used to input a second type of phrase set; the input end of the third type of semantic extraction unit is used to input a third type of phrase set; the first input end of the first multiplier M1 is connected to the output end of the first type of semantic extraction unit, and the second input end thereof is used to input a first type of significance sequence; the first input end of the second multiplier M2 is connected to the output end of the second type of semantic extraction unit, and the second input end thereof is used to input a second type of significance sequence; the first input end of the third multiplier M3 is connected to the output end of the third type of semantic extraction unit, and the second input end thereof is used to input a third type of significance sequence; the input end of the classifier is respectively connected to the output end of the first multiplier M1, the output end of the second multiplier M2 and the output end of the third multiplier M3, and the output end thereof serves as the output end of the multi-layer semantic classification model.

7. The digital management method based on big data analysis according to claim 6 is characterized in that: The first type of semantic extraction unit, the second type of semantic extraction unit and the third type of semantic extraction unit all include: multiple LiLSTM layers, multiple phrase semantic extraction layers and a Concat layer, each LiLSTM layer is used to process a phrase, a phrase semantic extraction layer is used to process the features output by a LiLSTM layer, and the Concat layer is used to splice the semantic feature values ​​output by multiple phrase semantic extraction layers into a semantic feature vector.

8. The digital management method based on big data analysis according to claim 7 is characterized in that: The expression of the phrase semantic extraction layer is: , where X is the semantic feature value output by the phrase semantic extraction layer, tanh is the hyperbolic tangent activation function, and h m is the mth feature output by the LiLSTM layer, ω m h m The weight of b m h m , M is the number of features output by the LiLSTM layer, and m is a positive integer.

9. The digital management method based on big data analysis according to claim 7 is characterized in that: The first multiplier M1 , the second multiplier M2 and the third multiplier M3 are used to perform element-wise multiplication on two inputs.

10. A digital management system based on big data analysis, implemented based on the digital management method based on big data analysis according to any one of claims 1 to 9, characterized in that: It includes: sentence division system, combination subsystem, sequence construction subsystem, classification subsystem and space allocation subsystem; The sentence segmentation system is used to segment the text content in the document into sentences to obtain multiple sentence texts; The combining subsystem is used to combine the characters in each sentence to obtain a first-category phrase set, a second-category phrase set, and a third-category phrase set; The sequence construction subsystem is used to calculate the prominence of each word group in each type of word group set, and construct a first type prominence sequence, a second type prominence sequence and a third type prominence sequence; The classification subsystem is used to process the first type of phrase set, the second type of phrase set and the third type of phrase set using a multi-layer semantic classification model, and obtain the importance of each sentence of text based on the prominence applied by the first type of prominence sequence, the second type of prominence sequence and the third type of prominence sequence; The space allocation subsystem is used to weight the importance of each sentence in the document to obtain the importance of the document, and allocate storage space to the document based on the importance of the document.

Images