All-media digital content publishing operation management system

Abstract

The invention discloses an all-media digital content publishing operation management system, and particularly relates to the technical field of digital content management.The all-media digital content publishing operation management system comprises the steps that a digital content text set needing to be published is obtained, corresponding image data and text data are extracted from each content text of the digital content text set, and a text complex coefficient is obtained; the content text complex coefficient is compared with a preset threshold value, the content texts in the digital content set are divided into an easy-to-recognize text set and a difficult-to-recognize text set according to a comparison result, the texts in the easy-to-recognize text set are audited through an automatic auditing technology, and the difficult-to-recognize text set is audited manually, so that the content text complex coefficient is obtained. The automatic auditing technology and the manual auditing technology are combined, the rapid publishing requirement of all-media large-scale content can be met, and the published content meets the standard.

Description

Technical Field

[0001] The present invention relates to the technical field of digital content management, and in particular to an all-media digital content publishing operation management system. Background Art

[0002] The omnimedia digital content publishing and operation management system is a comprehensive platform designed to manage, publish and operate various types of digital content to meet the needs of the omnimedia environment. The system generally covers several main functional modules, including content management module, publishing and distribution module, operation management module, copyright and security management module, etc. Through the mutual cooperation of these modules, the system can help media organizations, publishing companies, educational institutions, etc. to efficiently manage and operate their digital content and adapt to the rapid changes and diversified needs of the omnimedia era.

[0003] Among them, in the process of all-media digital content publishing and operation, in order to ensure that the published content meets the standards, the content needs to be reviewed; however, in actual review, manual review is inefficient and difficult to cope with the rapid release needs of large-scale content in all media, and automated review technology cannot replace manual review of published content, which may result in published content not meeting standards. Summary of the invention

[0004] The purpose of the present invention is to solve the problem that in actual review, manual review is inefficient and difficult to cope with the rapid release needs of large-scale content in all media, while automated review technology cannot replace manual review of published content, which may result in the published content not meeting the standards. The present invention proposes an all-media digital content publishing operation management system.

[0005] The present invention provides an all-media digital content publishing operation management system, the system includes a preliminary acquisition module, a text complexity coefficient module, a division module and an audit management module, and the modules are connected by signals, wherein:

[0006] A preliminary acquisition module: acquiring a digital content set to be published as a digital content text set; the digital content set includes a number of complete content texts;

[0007] Text complexity coefficient module: for each content text in the digital content text set, extract the corresponding image data and text data, and obtain the complexity coefficient of each content text according to the image data and text data;

[0008] Division module: compares the complexity coefficient of the content text with a preset threshold, and divides the content text in the digital content set into an easily recognizable text set and an unrecognizable text set according to the comparison result;

[0009] Review management module: The texts in the easily identifiable text set are reviewed through automated review technology, and the texts in the difficult-to-identify text set are reviewed manually.

[0010] Optionally, extracting corresponding image data and text data includes:

[0011] The image data includes an image information entropy coefficient Dmx, and the text data includes a part-of-speech distribution uniformity coefficient Gdv and a text length coefficient Mgh.

[0012] Optionally, the image information entropy coefficient includes:

[0013] Preprocessing each image contained in each content text, and converting each preprocessed image into a grayscale image;

[0014] For any image, construct the grayscale histogram H(g) of the image, where g represents the grayscale level in the image and H(g) represents the number of pixels with grayscale level g;

[0015] Calculate the occurrence probability P(g) of each gray level, the calculation formula is: Where N is the total number of pixels in the image;

[0016] Calculate the information entropy coefficient TG of the image l , the calculation expression is: In the formula, TG l represents the information entropy coefficient of the lth image in the content text, U is the total number of gray levels, and P(g)·log 2 (P(g)) is the entropy contribution of a single gray level;

[0017] Calculate the image information entropy coefficient of each content text, and the calculation formula is: Where Dmx is the image information entropy coefficient of the content text, and y is the total number of images contained in the text content.

[0018] Optionally, the part-of-speech distribution uniformity coefficient includes:

[0019] For each content text, all words are tagged with part-of-speech and the frequency of each part-of-speech is counted. i is the frequency of occurrence of part of speech i, f 1 、f 2 .....f n is the frequency list of all parts of speech; n is the total number of types of parts of speech;

[0020] Calculate the relative frequency of each part of speech using the following formula:

[0021]

[0022] In the formula, p i is the relative frequency of part of speech i, is the total frequency of all parts of speech;

[0023] Sort the relative frequencies of all parts of speech from small to large, and the relative frequency after sorting is p (1) , p (2) …… (n) , where p (1) ≤p (2) ≤……p (n) ;

[0024] Calculate the cumulative frequency of the relative frequency of the parts of speech after sorting. The calculation formula is:

[0025]

[0026] Among them, S i represents the cumulative frequency of the i-th part of speech;

[0027] Calculate the uniform coefficient of part-of-speech distribution, the calculation formula is:

[0028]

[0029] Where Gdv represents the uniformity coefficient of part-of-speech distribution.

[0030] Optionally, the text length coefficient includes:

[0031] For each content text, the total number of words in the text is obtained, and the total number of words contained in the special vocabulary is extracted from each content text to obtain the special vocabulary density coefficient. The calculation expression is: Wad = ln (t / c + 1), where Wad is the special vocabulary density coefficient, t is the total number of words contained in the special vocabulary, and c is the total number of words in the content text;

[0032] For each content text, the total number of words in the text is obtained, and the total number of sentences is extracted from each content text to obtain the average vocabulary size of the sentence. The calculation formula is: Edr = rf / c, where Edr is the average vocabulary size of the sentence and rf is the total number of sentences;

[0033] Calculate the text length coefficient, the calculation formula is: Mgh = a1 × Wad + a2 × Edr; where Mgh is the text length coefficient, and a1 and a2 represent the preset weight coefficients of the special vocabulary density coefficient and the average sentence vocabulary size.

[0034] Optionally, obtaining a complexity coefficient of each content text according to the image data and the text data includes:

[0035] The text complexity coefficient is obtained based on the image information entropy coefficient, the part-of-speech distribution uniformity coefficient and the text length coefficient. The calculation formula is:

[0036]

[0037] Where QED is the text complexity coefficient, Dmx, Gdv and Mgh are the image information entropy coefficient, part-of-speech distribution uniformity coefficient and text length coefficient respectively, α, β and γ are the preset proportional coefficients of the image information entropy coefficient, part-of-speech distribution uniformity coefficient and text length coefficient respectively, and α, β and γ are all greater than 0.

[0038] Optionally, comparing the content text complexity coefficient with a preset threshold, and dividing the content text in the digital content set into an easily recognizable text set and an unrecognizable text set according to the comparison result includes:

[0039] Compare the complexity coefficient of each content text with a preset threshold value, and if the complexity coefficient of the content text is less than the preset threshold value, divide the content text into an easily recognizable text set;

[0040] If the complexity coefficient of the content text is not less than a preset threshold, the content text is divided into a difficult-to-recognize text set.

[0041] Beneficial effects of the present invention:

[0042] The present invention obtains a digital content text set that needs to be published, and then for each content text in the digital content text set, extracts corresponding image data and text data to obtain a text complexity coefficient, and compares the content text complexity coefficient with a preset threshold value, and divides the content text in the digital content set into an easily identifiable text set and a difficult-to-identify text set according to the comparison result, and audits the texts in the easily identifiable text set through automated audit technology, and audits the difficult-to-identify text set manually, so that the automated audit technology and manual audit are combined, which can meet the needs of rapid publishing of large-scale content in all media and the published content meets the standards. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The present invention will be further described below in conjunction with the accompanying drawings.

[0044] Figure 1 A framework diagram of an all-media digital content publishing operation management system is provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative work shall fall within the scope of protection of the present invention.

[0047] The embodiment of the present invention provides an all-media digital content publishing operation management system. Figure 1 , Figure 1 The framework diagram of the all-media digital content publishing operation management system provided by the embodiment of the present invention. The system includes a preliminary acquisition module, a text complexity coefficient module, a division module and an audit management module, and the modules are connected by signals, wherein:

[0048] A preliminary acquisition module: acquiring a digital content set to be published as a digital content text set; the digital content set includes a number of complete content texts;

[0049] Text complexity coefficient module: for each content text in the digital content text set, extract the corresponding image data and text data, and obtain the complexity coefficient of each content text according to the image data and text data;

[0050] Division module: compares the complexity coefficient of the content text with a preset threshold, and divides the content text in the digital content set into an easily recognizable text set and an unrecognizable text set according to the comparison result;

[0051] Review management module: The texts in the easily identifiable text set are reviewed through automated review technology, and the texts in the difficult-to-identify text set are reviewed manually.

[0052] In one embodiment, extracting corresponding image data and text data includes:

[0053] The image data includes an image information entropy coefficient Dmx, and the text data includes a part-of-speech distribution uniformity coefficient Gdv and a text length coefficient Mgh.

[0054] In one embodiment, the image information entropy coefficient includes:

[0055] Preprocessing each image contained in each content text, and converting each preprocessed image into a grayscale image;

[0056] For any image, construct the grayscale histogram H(g) of the image, where g represents the grayscale level in the image and H(g) represents the number of pixels with grayscale level g;

[0057] Calculate the occurrence probability P(g) of each gray level, the calculation formula is: Where N is the total number of pixels in the image;

[0058] Calculate the information entropy coefficient TG of the image l , the calculation expression is: In the formula, TG l represents the information entropy coefficient of the lth image in the content text, U is the total number of gray levels, and P(g)·log 2 (P(g)) is the entropy contribution of a single gray level;

[0059] Calculate the image information entropy coefficient of each content text, and the calculation formula is: Where Dmx is the image information entropy coefficient of the content text, and y is the total number of images contained in the text content.

[0060] It should be noted that preprocessing each image contained in each content text is a key step in the information entropy calculation process. The purpose of preprocessing is to lay the foundation for subsequent grayscale image conversion and information entropy coefficient calculation. The following are the specific functions of each step in the preprocessing process:

[0061] 1. Image denoising:

[0062] In practical applications, image data often contains various noises, which may come from defects in acquisition equipment, environmental interference, or image compression processing. De-noising uses technical means such as filters to eliminate random noise in images, retain the main features of the image, and make subsequent analysis more accurate. The denoised image can more accurately reflect the true content of the image without causing grayscale histogram distortion due to the influence of noise.

[0063] 2. Contrast enhancement:

[0064] The contrast of an image directly affects the distribution of the grayscale histogram. Low-contrast images will cause the grayscale values ​​to be concentrated in a narrow range, affecting the accurate calculation of information entropy. Through contrast enhancement processing, the difference between different grayscale levels in the image can be enlarged, making the details of the image more obvious. This step helps to better capture the information in the image and improve the calculation effect of information entropy, especially when processing images with rich details.

[0065] 3. Image Scaling:

[0066] Different images may have different resolutions. If they are not scaled, images with larger resolutions may contain more pixels, resulting in a higher amount of information when calculating information entropy. Therefore, scaling all images to a uniform resolution ensures the comparability of different images in information entropy calculations. The scaled images have a consistent number of pixels, avoiding inconsistent information entropy calculations caused by differences in image size.

[0067] 4. Image Standardization:

[0068] Standardization includes adjusting the brightness and color balance of the image, so that all images are calculated on the same basis for information entropy. This step ensures that different images are compared under the same lighting conditions and tonal range, thereby avoiding distortion of information entropy calculation results caused by different shooting conditions or post-processing of different images. The standardized image can more objectively reflect the complexity of the image content.

[0069] 5. Grayscale conversion:

[0070] The preprocessed image is converted to a grayscale image, which is the basis for information entropy calculation. Each pixel value of a grayscale image represents its brightness, which simplifies the information content of the image and makes the calculation of information entropy more direct and efficient. Grayscale conversion allows the originally complex color image to be presented in a single dimension (grayscale value), so that the information entropy calculation focuses on the structure and content distribution of the image without being interfered by color factors.

[0071] Through the above preprocessing steps, the information entropy coefficient calculation of the image can more accurately reflect the complexity and information richness of the image. These processing steps ensure the consistency and quality of the image data, making the subsequent automated review more reliable.

[0072] It should be noted that the image information entropy coefficient refers to the measure of the complexity and information richness of the image contained in each content text. Information entropy is a statistic used to measure the amount of information or uncertainty. For image data, the information entropy coefficient can reflect the distribution of pixel values ​​in the image and the complexity of the image content; if the image information entropy coefficient of each content text is larger, it means that the content text is more complex and requires more manual review. The reason is:

[0073] Rich image details: The higher the information entropy coefficient, the more details the image contains. Detailed images tend to contain more visual information, such as complex textures, color changes, and diverse scenes. This type of image may make it difficult for automated review systems to accurately identify and analyze it, because the system needs to process more information to understand the content of the image.

[0074] Increased diversity and uncertainty: A high information entropy coefficient reflects the diversity and uncertainty of the distribution of image pixel values. This means that there is no obvious dominant color or shape in the image, or there are many complex patterns and structures. This complexity will increase the burden on automated review algorithms, especially when the algorithms need to identify specific content or filter sensitive information.

[0075] Increased risk of misjudgment: Because complex images may contain multiple overlapping objects, complex backgrounds, or difficult-to-distinguish details, automated review systems may be more likely to make misjudgments or miss judgments. Images with high information entropy may confuse review algorithms, making it difficult to correctly classify image content, resulting in inaccurate review results.

[0076] Necessity of human intervention: When the image information entropy coefficient is high, the accuracy of automated review technology may not meet the requirements, and manual review becomes particularly important. Human reviewers can use human vision and cognitive abilities to understand the complex content in the image and make more accurate judgments to avoid errors that may be caused by automated systems.

[0077] Contextual complexity caused by content diversity: Images with high information entropy often mean that the image may involve multiple elements or situations. For example, images containing complex scenes, multiple objects, or images that require understanding of a wider context will increase the difficulty of review. In this case, the automated review system may not be able to accurately grasp the overall meaning of the content and may easily ignore contextual associations, while human reviewers can more accurately evaluate the content through overall understanding.

[0078] Therefore, when the image information entropy coefficient of each content text is larger, the content text is more complex and requires more manual review. This review method can ensure that the key information in complex images is accurately identified and processed, avoiding review errors caused by the limitations of the automated review system in processing complex images.

[0079] In one embodiment, the part-of-speech distribution uniformity coefficient includes:

[0080] For each content text, all words are tagged with part-of-speech and the frequency of each part-of-speech is counted. i is the frequency of occurrence of part of speech i, f 1 、f 2 .....f n is the frequency list of all parts of speech; n is the total number of types of parts of speech;

[0081] Calculate the relative frequency of each part of speech using the following formula:

[0082]

[0083] In the formula, p iis the relative frequency of part of speech i, is the total frequency of all parts of speech;

[0084] Sort the relative frequencies of all parts of speech from small to large, and the relative frequency after sorting is p (1) , p (2) …… (n) , where p (1) ≤p (2) ≤……p (n) ;

[0085] Calculate the cumulative frequency of the relative frequency of the parts of speech after sorting. The calculation formula is:

[0086]

[0087] Among them, S i represents the cumulative frequency of the i-th part of speech;

[0088] Calculate the uniform coefficient of part-of-speech distribution, the calculation formula is:

[0089]

[0090] Where Gdv represents the uniformity coefficient of part-of-speech distribution.

[0091] It should be noted that the part-of-speech distribution uniformity coefficient refers to the uniformity of the part-of-speech distribution of the text contained in each content text; if the part-of-speech distribution uniformity coefficient is high, the language structure of the text may be relatively complex, the difficulty of review is relatively high, and manual review is more required, because if the part-of-speech distribution uniformity coefficient is high, it means that the various parts of speech in the text (such as nouns, verbs, adjectives, etc.) are distributed more evenly throughout the text, which usually reflects that the language structure of the text is more complex and diverse; when the part-of-speech distribution uniformity coefficient is high, the language structure of the text is complex, and may include multi-level sentence structures, rich expressions, and diverse grammatical usage. This complexity increases the difficulty of automated review technology (such as NLP algorithms) to process and understand text. For example, complex sentence structures may lead to difficulties in semantic analysis, making it difficult for automated systems to accurately judge the content, emotional tendencies, or potential sensitivity of the text.

[0092] In addition, texts with a higher uniformity coefficient for part-of-speech distribution may contain more rhetoric, metaphors, or puns, which are often beyond the capabilities of current automated review technology. These factors make it difficult for automated review systems to identify potential problems or prohibited content in texts.

[0093] Therefore, when the uniformity coefficient of part-of-speech distribution of a text is high, manual review is usually required. This is because manual review can better understand complex language structures and potential semantic meanings, and thus more accurately assess whether the text content meets the publishing standards. Manual review can conduct a more in-depth analysis of complex content that is difficult for automated review to handle, ensuring the accuracy and reliability of the final review results.

[0094] It should be noted that the above calculation method has the following significant advantages when calculating the uniform coefficient of part-of-speech distribution:

[0095] 1. Accurately reflect the uniformity of part-of-speech distribution;

[0096] By calculating and ranking the relative frequency of each part of speech, and then calculating the part-of-speech distribution uniformity coefficient through the cumulative frequency and Gini coefficient formula, this method can accurately reflect the uniformity of the part-of-speech distribution in the text. Compared with simply calculating the frequency proportion of each part of speech, this method can better capture the overall characteristics of the part-of-speech distribution, thereby more accurately evaluating the complexity of the text.

[0097] 2. Considering the diversity of parts of speech

[0098] This calculation method not only focuses on the frequency of occurrence of each part of speech, but also comprehensively considers the relative proportion of different parts of speech in the text through sorting and cumulative frequency. Therefore, if there are multiple parts of speech in the text and they are evenly distributed, it will be reflected in a higher part of speech distribution uniformity coefficient. This method avoids the situation where one or two parts of speech dominate the distribution of the entire text, thereby more comprehensively measuring the language structure of the text.

[0099] 3. Adapt to texts of varying complexity;

[0100] Since this calculation method evaluates the part-of-speech distribution by ranking relative frequencies and cumulative frequencies, it can accurately calculate the part-of-speech distribution uniformity coefficient for both simple and complex texts. For complex texts, especially those with high language diversity and rich grammar, this method can better reflect the complexity of their language structure.

[0101] 4. Improve the accuracy of the automated audit system;

[0102] In text review, the part-of-speech distribution uniformity coefficient can be used as an indicator to help identify texts with complex language structures that require more attention. By calculating this coefficient, the automated review system can more accurately screen out text content that may have potential problems, assisting manual reviewers in making more accurate assessments and reducing the occurrence of misjudgments and missed judgments.

[0103] 5. Helps optimize natural language processing algorithms;

[0104] In the field of natural language processing (NLP), this calculation method provides a clear standard to measure the complexity of the text language structure. This helps optimize the NLP algorithm so that it can better understand and analyze the semantics of the text when processing complex text, thereby improving the effect of automated processing. When training the algorithm, the parameters of the model can also be adjusted based on the uniform coefficient of part-of-speech distribution to better adapt it to texts of different complexities.

[0105] In one embodiment, the text length factor includes:

[0106] For each content text, the total number of words in the text is obtained, and the total number of words contained in the special vocabulary is extracted from each content text to obtain the special vocabulary density coefficient. The calculation expression is: Wad = ln (t / c + 1), where Wad is the special vocabulary density coefficient, t is the total number of words contained in the special vocabulary, and c is the total number of words in the content text;

[0107] For each content text, the total number of words in the text is obtained, and the total number of sentences is extracted from each content text to obtain the average vocabulary size of the sentence. The calculation formula is: Edr = rf / c, where Edr is the average vocabulary size of the sentence and rf is the total number of sentences;

[0108] Calculate the text length coefficient, the calculation formula is: Mgh = a1 × Wad + a2 × Edr; where Mgh is the text length coefficient, and a1 and a2 represent the preset weight coefficients of the special vocabulary density coefficient and the average sentence vocabulary size.

[0109] It should be noted that special vocabulary refers to some words that are difficult to recognize, such as proverbs, idioms, etc., or other words, without specific limitation.

[0110] It should be noted that the text length coefficient refers to an indicator that measures the overall length of a content text, and is usually used to evaluate the complexity and readability of the text; the text length coefficient can reflect the level of detail and information density of a text. The longer the text, the richer the content and the more complex the structure. If the text length coefficient is high, the language structure of the text may be relatively complex, and the difficulty of review is relatively high, and more manual review is required, because a higher text length coefficient usually means that the text content is more complex, the amount of information is larger, and the structure is more diverse. This complexity increases the difficulty for automated review systems to accurately analyze and understand texts, especially when dealing with texts containing special vocabulary, complex sentence structures, or multi-level information. Automated review may find it difficult to fully capture these complex language phenomena, and may miss some potential problems or misjudge the content of the text. Therefore, a higher text length coefficient means that the text requires more detailed manual review to ensure the accuracy, compliance, and readability of the content. Manual review can more flexibly handle complex factors such as language nuances, contextual relationships, and cultural background, so as to understand and judge the content more accurately and ensure that the published content meets the expected standards and requirements.

[0111] In one embodiment, obtaining the complexity coefficient of each content text according to the image data and the text data includes:

[0112] The text complexity coefficient is obtained based on the image information entropy coefficient, the part-of-speech distribution uniformity coefficient and the text length coefficient. The calculation formula is:

[0113]

[0114] Where QED is the text complexity coefficient, Dmx, Gdv and Mgh are the image information entropy coefficient, part-of-speech distribution uniformity coefficient and text length coefficient respectively, α, β and γ are the preset proportional coefficients of the image information entropy coefficient, part-of-speech distribution uniformity coefficient and text length coefficient respectively, and α, β and γ are all greater than 0.

[0115] In one embodiment, comparing the content text complexity coefficient with a preset threshold, and dividing the content text in the digital content set into an easily recognizable text set and an unrecognizable text set according to the comparison result includes:

[0116] Compare the complexity coefficient of each content text with a preset threshold value, and if the complexity coefficient of the content text is less than the preset threshold value, divide the content text into an easily recognizable text set;

[0117] If the complexity coefficient of the content text is not less than a preset threshold, the content text is divided into a difficult-to-recognize text set.

[0118] Based on the all-media digital content publishing operation and management system provided by the embodiment of the present invention, by obtaining a digital content text set that needs to be published, and then for each content text in the digital content text set, extracting the corresponding image data and text data to obtain a text complexity coefficient, and comparing the content text complexity coefficient with a preset threshold, the content text in the digital content set is divided into an easily recognizable text set and a difficult-to-recognize text set according to the comparison result, and the text in the easily recognizable text set is reviewed by automated review technology, and the difficult-to-recognize text set is reviewed manually. In this way, the automated review technology and manual review are combined, which can meet the needs of rapid publication of large-scale content in all media and the published content meets the standards.

[0119] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. The all-media digital content publishing and operation management system is characterized by: The system includes a preliminary acquisition module, a text complexity coefficient module, a division module and an audit management module, and each module is connected by signals, wherein: A preliminary acquisition module: acquiring a digital content set to be published as a digital content text set; the digital content set includes a number of complete content texts; Text complexity coefficient module: for each content text in the digital content text set, extract the corresponding image data and text data, and obtain the complexity coefficient of each content text according to the image data and text data; Division module: compares the complexity coefficient of the content text with a preset threshold, and divides the content text in the digital content set into an easily recognizable text set and an unrecognizable text set according to the comparison result; Review management module: The texts in the easily identifiable text set are reviewed through automated review technology, and the texts in the difficult-to-identify text set are reviewed manually.

2. The all-media digital content publishing and operation management system according to claim 1 is characterized in that: Extracting corresponding image data and text data includes: The image data includes an image information entropy coefficient Dmx, and the text data includes a part-of-speech distribution uniformity coefficient Gdv and a text length coefficient Mgh.

3. The all-media digital content publishing and operation management system according to claim 2 is characterized in that: The image information entropy coefficient includes: Preprocessing each image contained in each content text, and converting each preprocessed image into a grayscale image; For any image, construct the grayscale histogram H(g) of the image, where g represents the grayscale level in the image and H(g) represents the number of pixels with grayscale level g; Calculate the occurrence probability P(g) of each gray level, the calculation formula is: Where N is the total number of pixels in the image; Calculate the information entropy coefficient TG of the image l , the calculation expression is: In the formula, TG l represents the information entropy coefficient of the lth image in the content text, U is the total number of gray levels, and P(g)·log2(P(g)) is the entropy contribution of a single gray level; Calculate the image information entropy coefficient of each content text, and the calculation formula is: Where Dmx is the image information entropy coefficient of the content text, and y is the total number of images contained in the text content.

4. The all-media digital content publishing and operation management system according to claim 2 is characterized in that: The part-of-speech distribution uniformity coefficient includes: For each content text, all words are tagged with part-of-speech and the frequency of each part-of-speech is counted. i is the frequency of occurrence of part of speech i, f1, f2, .....f n is the frequency list of all parts of speech; n is the total number of types of parts of speech; Calculate the relative frequency of each part of speech using the following formula: In the formula, p i is the relative frequency of part of speech i, is the total frequency of all parts of speech; Sort the relative frequencies of all parts of speech from small to large, and the relative frequency after sorting is p (1) , p (2) ......p (n) , where p (1) ≤p (2) ≤......p (n) ; Calculate the cumulative frequency of the relative frequency of the parts of speech after sorting. The calculation formula is: Among them, S i represents the cumulative frequency of the i-th part of speech; Calculate the uniform coefficient of part-of-speech distribution, the calculation formula is: Where Gdv represents the uniformity coefficient of part-of-speech distribution.

5. The all-media digital content publishing operation management system according to claim 2 is characterized in that: The text length factors include: For each content text, the total number of words in the text is obtained, and the total number of words contained in the special vocabulary is extracted from each content text to obtain the special vocabulary density coefficient. The calculation expression is: Wad = ln (t / c + 1), where Wad is the special vocabulary density coefficient, t is the total number of words contained in the special vocabulary, and c is the total number of words in the content text; For each content text, the total number of words in the text is obtained, and the total number of sentences is extracted from each content text to obtain the average vocabulary size of the sentence. The calculation formula is: Edr = rf / c, where Edr is the average vocabulary size of the sentence and rf is the total number of sentences; Calculate the text length coefficient, the calculation formula is: Mgh = a1 × Wad + a2 × Edr; where Mgh is the text length coefficient, and a1 and a2 represent the preset weight coefficients of the special vocabulary density coefficient and the average sentence vocabulary size.

6. The all-media digital content publishing and operation management system according to claim 1 is characterized in that: The complexity coefficient of each content text obtained based on image data and text data includes: The text complexity coefficient is obtained based on the image information entropy coefficient, the part-of-speech distribution uniformity coefficient and the text length coefficient. The calculation formula is: Where QED is the text complexity coefficient, Dmx, Gdv and Mgh are the image information entropy coefficient, part-of-speech distribution uniformity coefficient and text length coefficient respectively, α, β and γ are the preset proportional coefficients of the image information entropy coefficient, part-of-speech distribution uniformity coefficient and text length coefficient respectively, and α, β and γ are all greater than 0.

7. The all-media digital content publishing and operation management system according to claim 1, characterized in that: Compare the content text complexity coefficient with the preset threshold, and divide the content text in the digital content set into an easily recognizable text set and an unrecognizable text set according to the comparison result, including: Compare the complexity coefficient of each content text with a preset threshold value, and if the complexity coefficient of the content text is less than the preset threshold value, divide the content text into an easily recognizable text set; If the complexity coefficient of the content text is not less than a preset threshold, the content text is divided into a difficult-to-recognize text set.

Images