A method, system, computer device and medium for intelligent recommendation of network materials

Abstract

The present invention provides an intelligent recommendation method, system, computer device and storage medium for network materials. The scheme method includes: obtaining a first material file, pre-processing the first material file by labeling, and constructing a knowledge graph based on the first material file after labeling pre-processing; obtaining a user input instruction, and extracting a first keyword from the user input instruction; matching the first keyword in the knowledge graph, filtering and obtaining a second material file according to the matching result, and generating a template file according to the second material file; the scheme is based on the knowledge graph, so that the classification of materials is not limited to the attribute classification of the materials themselves, and provides users with richer optional content. The scheme can realize the creative creation output of users' design concepts, the combination of templates is diversified, and the concept association search is more universal, which can be widely used in the field of computer technology.

Description

Technical Field

[0001] The present invention belongs to the field of computer technology, and in particular to an intelligent recommendation method, system, computer equipment and storage medium for network materials. Background Art

[0002] In the prior art, various creative platforms and corresponding software, in the process of intelligently generating creative content, will reorganize and layout the materials stored in the platform, and output fixed templates. Users can choose according to their specific needs, or find creative content templates of the same category through the classification of templates; however, the templates generated by the prior art solutions are monotonous and have limited content options for different users due to the fixed materials and template layout. Summary of the invention

[0003] In view of this, in order to at least partially solve one of the above-mentioned technical problems, an embodiment of the present invention aims to provide an intelligent recommendation method for network materials with richer recommendation schemes and wider applicable scenarios, as well as a system, device and storage medium that can implement the method accordingly.

[0004] In a first aspect, the technical solution of the present application provides a method for intelligently recommending network materials, the steps of which include:

[0005] Acquire a first material file, perform labeling preprocessing on the first material file, and construct a knowledge graph based on the first material file after the labeling preprocessing;

[0006] Obtaining a user input instruction, and extracting a first keyword from the user input instruction;

[0007] A match is performed in the knowledge graph according to the first keyword, a second material file is obtained by screening according to the matching result, and a template file is generated according to the second material file.

[0008] In a feasible embodiment of the solution of the present application, the step of pre-processing the first source file into tags includes at least one of the following steps:

[0009] Acquire size information and format information of the first material file, and generate an attribute tag;

[0010] Acquire the color value of the pixel in the first material file, and generate a color label according to the color value;

[0011] Acquire the image edge of the first source file, extract image texture and shape features according to the image edge, and generate a content label according to the image texture and the shape features;

[0012] Generate a style label for the first material file through a pre-trained deep learning model.

[0013] In a feasible embodiment of the present application, the step of constructing a knowledge graph according to the first material file after labeling preprocessing includes:

[0014] Acquire a third material file according to at least one tag of the first material file;

[0015] Using at least one of the tags as a second keyword, and generating a connection line of the knowledge graph according to an association relationship between an element of the third material file and the second keyword;

[0016] The second keyword and the element are used as nodes and combined with the connecting lines to construct the knowledge graph.

[0017] In a feasible embodiment of the present application, the step of determining the association relationship between the element of the third material file and the second keyword includes:

[0018] Performing similarity matching based on the second keyword to filter the third material file to obtain the element;

[0019] The correlation value is determined according to the content proportion of the element in the third material file; the correlation value is used to characterize the correlation relationship.

[0020] In a feasible embodiment of the present application, after the step of obtaining the third source file, the method further includes the following steps:

[0021] The third material file is subjected to denoising processing, and the denoised third material file is subjected to standardization processing to obtain a plurality of candidate elements.

[0022] In a feasible embodiment of the solution of the present application, the step of obtaining a user input instruction and extracting a first keyword from the user input instruction includes:

[0023] Segmenting the input instruction to obtain a plurality of words;

[0024] The vocabulary is semantically matched with at least one tag of the first material file, and the first keyword is determined according to the matching result.

[0025] In a feasible embodiment of the solution of the present application, a plurality of historical template files are obtained, and a typesetting rule and a third keyword in the typesetting rule are extracted from the historical template files;

[0026] It is determined that the third keyword matches the first keyword, and a target typesetting rule of the template file is obtained according to the frequency of the typesetting rule.

[0027] In a second aspect, the technical solution of the present invention further provides an intelligent recommendation system for network materials, the system comprising:

[0028] A graph construction unit, configured to obtain a first material file, perform labeling preprocessing on the first material file, and construct a knowledge graph based on the first material file after the labeling preprocessing;

[0029] An instruction acquisition unit, used to acquire a user input instruction and extract a first keyword from the user input instruction;

[0030] A material screening unit, configured to match the first keyword in the knowledge graph, and screen and obtain a second material file according to the matching result;

[0031] A template generating unit generates a template file according to the second material file.

[0032] In a third aspect, the technical solution of the present invention further provides a computer device for intelligently recommending network materials, which includes:

[0033] at least one processor;

[0034] at least one memory for storing at least one program;

[0035] When at least one program is executed by at least one processor, the at least one processor runs the method described in the first aspect.

[0036] In a fourth aspect, the technical solution of the present invention further provides a storage medium, in which a program executable by a processor is stored. When the program executable by the processor is executed by the processor, it is used to run the intelligent recommendation method for network materials described in the first aspect.

[0037] The advantages and beneficial effects of the present invention will be partially given in the following description, and the other parts can be understood through the specific embodiments of the present invention:

[0038] The technical solution of the present application first pre-processes the materials by labeling, establishes a knowledge graph related to the material labels, and the user inputs the creative concept, finds the concept-related materials through the knowledge graph, uses the materials to generate a variety of templates, and recommends the templates to the user; the solution is based on the knowledge graph, so that the classification of materials is not limited to the attribute classification of the materials themselves, and provides users with richer optional content. The solution can realize the creative creation output of the user's design concept, the combination of templates is diversified, and the concept association search is more universal. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0040] Figure 1 A flowchart of a method for intelligently recommending network materials provided by an embodiment of the present invention;

[0041] Figure 2 A flowchart of the steps of another method for intelligently recommending network materials provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0042] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and are not to be construed as limitations of the present invention. For the step numbers in the following embodiments, they are only provided for the convenience of explanation, and the order between the steps is not limited in any way. The execution order of each step in the embodiment can be adaptively adjusted according to the understanding of those skilled in the art.

[0043] The material templates generated by the existing technology have the problem of being monotonous due to the fixed layout of the materials and templates. The technical solution of this application, by establishing a knowledge graph, allows users to input creative concepts in the form of instructions, obtain related materials, and then generate corresponding templates through algorithms and layout, so that the user's creative concepts can achieve the effect of intelligent output and intelligent recommendation.

[0044] First, as Figure 1 As shown, the technical solution of the present application provides an intelligent recommendation method for network materials, and its main steps may include S100-S300:

[0045] S100: Obtain a first material file, perform labeling preprocessing on the first material file, and construct a knowledge graph based on the first material file after the labeling preprocessing;

[0046] Among them, the first material file refers to an exemplary material file used to generate various types of tags, and its source can be the corresponding material obtained from the existing mature material library; in the embodiment, the file formats of various types of material files (including but not limited to the first material file, the second material file and the third material file, etc.) can be diverse, including but not limited to images, audio, video and text content materials. The process of label preprocessing is mainly to generate corresponding character format labels for the first material file. The processing process includes but is not limited to extracting the color value of the image material as a label of the color attribute, extracting the specific content in the image material as a label of the content attribute, and extracting keywords or semantic features in the text material to obtain corresponding content attribute labels, etc. The knowledge graph is the knowledge base of the semantic network. In the embodiment of the present application, the knowledge graph can be used to characterize the association relationship between labels.

[0047] Specifically, taking picture materials as an example, the material content is stored in an existing structured manner, that is, the first material file constructs a knowledge graph. First, the embodiment performs preprocessing according to the various format attributes when the material is stored. The preprocessing is to extract some public attributes of the material, including size, format, content, tone, and style. The labeling process is to add material attribute values ​​to different materials, such as jpg, png, svg, psd and other labels for format attributes; 720×1280, 4k and other labels for size attributes; forest, mountain, car, portrait and other labels for content attributes; then the generated label content is used as the node of the knowledge graph, and the connection line between the nodes (the edge of the knowledge graph) is formed according to the association relationship between the nodes, that is, the association relationship between the label content, and finally a complete knowledge graph is formed.

[0048] S200, obtaining a user input instruction, and extracting a first keyword from the user input instruction;

[0049] Specifically, the first keyword refers to a keyword extracted from the user input instruction through natural language processing after obtaining the user input instruction; wherein the user input instruction can be a text content describing the creative concept; or it can be a combination of keywords directly input by the user.

[0050] Exemplarily, taking the user input instruction in the form of a combination of keywords as an example, first of all, the user is not clear about what content or style of materials exist in the template library currently used, and can only enter keywords according to actual needs. When the template library does not have the content and classification of the material or template of this keyword, there is a situation where there is no search result after the search is completed. For another example, when the user enters the keyword "atmosphere" to search, the program first searches for the material and template of the word "atmosphere". After the search is completed, there are the following situations: First, if there is no template corresponding to, search whether there is a related word of the keyword "atmosphere" in the knowledge graph. If it does not exist, use "atmosphere" as a keyword to construct the corresponding knowledge graph, establish a knowledge graph for "atmosphere", and use the related words as search keywords to search templates or materials until the corresponding materials and templates are found. Second, if there is no template corresponding to, but there are related words of the keyword "atmosphere" in the knowledge graph, then directly use the related words as search keywords to search templates or materials until the corresponding materials and templates are found. Third, if there is a template, directly display this part of the material and template, and at the same time, search for some related materials and templates for display, and record the proportion of users using recommended materials and templates.

[0051] S300, matching the first keyword in the knowledge graph, screening to obtain a second material file according to the matching result, and generating a template file according to the second material file;

[0052] Among them, the second material file refers to a material file that is determined to meet the target requirements of the user input instruction after keyword matching; in the process of keyword matching, this embodiment can calculate the semantic similarity between keywords for the keywords that need to be matched; for example, in the embodiment, the keywords that need to be matched can be converted into the form of word vectors, and then the similarity of the two word vectors is calculated using cosine similarity. When the similarity of the two word vectors meets the preset similarity threshold, the keyword matching is determined to be successful. It can be understood that when the user input instruction contains multiple keywords, the similarity matching of multiple keywords can be performed simultaneously. After the match is successful, in the knowledge graph constructed by the embodiment, a node search is performed based on the successfully matched keywords to determine all nodes covered by the input keywords, and based on the direct association relationship between the nodes, the corresponding material content is formed into a target template file according to the preset typesetting rules.

[0053] In some optional embodiments, for source files of different formats and different contents, the process of labeling the first source file may include but is not limited to steps S110-S140:

[0054] S110, obtaining size information and format information of a first material file, and generating an attribute tag;

[0055] S120, obtaining color values ​​of pixels in the first material file, and generating color labels according to the color values;

[0056] S130, obtaining an image edge of the first source file, extracting image texture and shape features according to the image edge, and generating a content label according to the image texture and shape features;

[0057] S140, generating a style label of the first material file through a pre-trained deep learning model;

[0058] Exemplarily, during the process of uploading the material, the embodiment first obtains the basic attributes of the image material, such as size and image format, and then saves such attribute tags. The embodiment can also obtain the RGBA or HEX color value of each pixel in the image through image recognition technology, and count the proportion of the color value to the total pixels to determine the color and hue and other attributes of the image. The embodiment can remove the influence of noise and determine the image edge by image recognition on the premise of obtaining the color of the pixel; after determining the image edge, the texture and shape features of the image can be extracted, and the image content (such as trees, mountains, cars, portraits, etc.) can be finally determined and labeled with content. Finally, the embodiment can also generate style attribute tags for the image, which is to first perform style judgment on the preprocessed image through machine learning methods. The machine analyzes the hue, layout, content, etc. of the image material based on the judgment results, improves the accuracy of the machine's style judgment, and finally completes the labeling preprocessing of the material.

[0059] In some optional embodiments, the step of constructing a knowledge graph based on the first material file after labeling preprocessing may further include steps S150-S170:

[0060] S150, acquiring a third material file according to at least one tag of the first material file;

[0061] S160, using at least one tag as a second keyword, and generating a connection line of the knowledge graph according to the association relationship between the elements of the third material file and the second keyword;

[0062] S170, taking the second keyword and the element as nodes and combining them with connecting lines to construct a knowledge graph;

[0063] Among them, the third material in the embodiment can be data that is public and captured on the Internet. These data files usually exist in the form of web pages, so they are unstructured data content. Specifically, the embodiment uses the tags generated in steps S110-S140 as keywords to search in public network data to obtain corresponding material files, and performs necessary preprocessing on the searched material files to remove noise and invalid data, and unifies the format content and disassembles the content to obtain material elements. The keywords used in the search process are used as a node or entity in the knowledge graph, and the disassembled material elements are used as other nodes or entities. According to the attributes or associations between nodes or entities, the edges of the knowledge graph, that is, the connecting lines of the knowledge graph, are formed, and finally all nodes or entities that can be connected are connected to obtain a complete knowledge graph.

[0064] In some optional embodiments, the process of step S160 according to the association relationship between the elements of the third material file and the second keyword may include steps S161-S162:

[0065] S161, performing similarity matching according to the second keyword and filtering out elements from the third material file;

[0066] S162, determining a correlation value according to a content proportion of the element in the third material file;

[0067] Among them, the association value is used to characterize the association relationship. Exemplarily, taking text content materials as an example, the embodiment can convert the keywords that need to be matched into the form of word vectors, and then use cosine similarity (CosineSimilarity) to calculate the similarity of the two word vectors. When the similarity of the two word vectors meets the preset similarity threshold, the keyword match is deemed to be successful; then, the number of occurrences of words with the same part of speech in all articles is counted; based on the results after the statistics are completed, the embodiment sets a variable x, for example, x can be derived from the number of words in the article. For every 1,000 words, the proportion of a certain word that appears is 10%. We believe that this word is associated with the keyword, and x is 0.1. After the knowledge graph is established, the variable x can review the association between words and associated words, increase or decrease the value of x, and improve the association. After finding the words and associated words, the proportion of the words in the 1,000 words of the article content is used as the association value y. Use associated words, association values, and words to form a data triple (associated words, association, words) of the knowledge graph. Finally, repeat this work within a certain period of time, compare the latest correlation value y with the old y value, and on the basis that the y value is higher than the latest x value, add the new and old values ​​of y and take the average as the new y value. Continuously update the y value to improve the correlation and accuracy of keywords and related words, thereby establishing a complete knowledge graph of the knowledge field.

[0068] In addition, the keywords in the embodiment can also filter out the most commonly used keywords through the search record data generated by the user when searching, and add these keywords to the search keywords for the crawler to obtain data, making the embodiment more intelligent.

[0069] In some optional embodiments, after the process of obtaining the third source file in step S150 of the method, the method may further include step S150A:

[0070] S150A, performing denoising processing on the third material file, and performing standardization processing on the denoised third material file to obtain a number of candidate elements;

[0071] For example, taking text content as an example, the embodiment uses the tag content and the tag as a keyword in the search engine to search for the latest 100 articles. After obtaining the article content, the article content is segmented using the Jieba word segmentation. First, remove the function words that are meaningless to the material, such as conjunctions, prepositions, auxiliary words, and modal particles. Then determine the part of speech of the keyword and find words with the same or similar part of speech in the article content.

[0072] In some optional embodiments, the step S200 of obtaining a user input instruction and extracting a first keyword from the user input instruction may include steps S210-S220:

[0073] S210, segmenting the input instruction to obtain a number of words;

[0074] S220, semantically matching the vocabulary with at least one tag of the first material file, and determining a first keyword according to the matching result;

[0075] Specifically, when the user input instruction in the embodiment is a text content that describes a creative concept, the embodiment can use Jieba word segmentation to segment the user input instruction content after obtaining the user input instruction; first remove meaningless function words, such as conjunctions, prepositions, auxiliary words and modal particles; then determine the part of speech of the keyword, and search for words with the same or similar part of speech as the label obtained in steps S110-S140 in the user input instruction content.

[0076] In some optional embodiments, the process of generating a template file according to the second material file in step S300 may include steps S310-S320:

[0077] S310, obtaining several historical template files, and extracting typesetting rules and the third keyword in the typesetting rules from the historical template files;

[0078] S320, determining that the third keyword matches the first keyword, and obtaining a target typesetting rule of the template file according to the frequency of the typesetting rule;

[0079] Specifically, after determining the materials associated with the keywords, the embodiment adopts the template layout with a high probability of adopting the previously generated templates, manually checks the combination template rules, uses the existing template layout, etc., and uses the above layout to combine the materials according to the material attribute tags to generate diversified templates; at the same time, a part of the generated templates is randomly recorded to the management background for review, and the results are recorded through machine learning to find out the common points of the layout with a high layout pass rate and optimize the layout rules. The generated templates are sorted and displayed from multiple aspects such as the completeness of the template, the rationality of the layout design, the template usage rate, and the conformity of the template with the creative concept, so that users can experience the use effects such as the creation of design concepts, the diversification of templates, and intelligent recommendations.

[0080] Combined with Figure 2 , a complete description of the complete embodiment of the present application scheme is given below:

[0081] The implementation example first pre-processes the materials by labeling them, and establishes a knowledge graph related to the material labels. The user inputs the creative concept, finds the material related to the concept through the knowledge graph, and uses the material to generate a variety of templates, which are recommended to the user. Material labeling pre-processes the materials by labeling them according to the content, size, etc. of the materials; establishes a complete knowledge graph to associate the concept with the actual material; combines the materials into a template that meets the user's needs through data and algorithms, and finally outputs the template for recommendation to the user.

[0082] In a second aspect, the technical solution of the present invention further provides an intelligent recommendation system for network materials, the system comprising:

[0083] A graph construction unit, used to obtain a first material file, perform labeling preprocessing on the first material file, and construct a knowledge graph based on the first material file after the labeling preprocessing;

[0084] An instruction acquisition unit, used to acquire an instruction input by a user, and extract a first keyword from the instruction input by the user;

[0085] A material screening unit, used to match the first keyword in the knowledge graph, and screen the second material file according to the matching result;

[0086] The template generating unit generates a template file according to the second material file.

[0087] In a third aspect, the technical solution of the present invention also provides a computer device for intelligently recommending network materials, which includes: at least one processor; at least one memory for storing at least one program; when the at least one program is executed by the at least one processor, the at least one processor runs the method in the first aspect.

[0088] From the above specific implementation process, it can be concluded that the technical solution provided by the present invention has the following advantages or strengths compared with the prior art:

[0089] This technical solution is based on the knowledge graph, so that the classification of materials is not limited to the attribute classification of the materials themselves, providing users with richer optional content. The solution can realize the creative output of users' design concepts, diversify the combination of templates, and make concept-related searches more universal.

[0090] In some selectable embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the present invention is provided by way of example, for the purpose of providing a more comprehensive understanding of technology. The disclosed method is not limited to the operation and logic flow presented herein. Selectable embodiments are expected, wherein the order of various operations is changed and the sub-operation of a part for which is described as a larger operation is performed independently.

[0091] In addition, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise specified, one or more of the functions and / or features can be integrated into a single physical device and / or software module, or one or more functions and / or features can be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the present invention. More specifically, in view of the properties, functions and internal relationships of the various functional modules in the device disclosed herein, the actual implementation of the module will be understood within the conventional skills of the engineer. Therefore, those skilled in the art can implement the present invention set forth in the claims without excessive experimentation using ordinary techniques. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0092] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, which can be embodied in any computer-readable medium for use by an instruction execution system, apparatus or device (such as a computer-based system, a system including a processor or other system that can fetch instructions from an instruction execution system, apparatus or device and execute instructions), or used in combination with these instruction execution systems, apparatuses or devices.

[0093] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0094] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

[0095] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the above embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A method for intelligently recommending network materials, characterized in that: The following steps are involved: Acquire a first material file, perform labeling preprocessing on the first material file, and construct a knowledge graph based on the first material file after the labeling preprocessing; the labeling preprocessing on the first material file, The method comprises at least one of the following steps: obtaining the size information and format information of the first material file, and generating an attribute tag; Obtaining color values ​​of pixels in the first source file, and generating color labels according to the color values; obtaining image edges of the first source file, extracting image texture and shape features according to the image edges, and generating content labels according to the image texture and shape features; Generate a style label for the first material file through a pre-trained deep learning model; The step of constructing a knowledge graph based on the first material file after labeling preprocessing includes: obtaining a third material file based on at least one of the attribute label, the color label, the content label, and the style label; the third material file includes a public data file captured from the Internet, and the data file includes unstructured data content in the form of a web page; removing noise and invalid data in the third material file, unifying the format and content of the third material file and disassembling the content to obtain the elements of the third material file; taking at least one of the attribute label, the color label, the content label, and the style label as a second keyword, and generating a connecting line of the knowledge graph based on the association relationship between the element of the third material file and the second keyword; taking the second keyword and the element as nodes, and constructing the knowledge graph in combination with the connecting line; the step of obtaining the association relationship between the element of the third material file and the second keyword based on the element of the third material file includes: performing similarity matching based on the second keyword to filter the element from the third material file; determining a correlation value based on the content proportion of the element in the third material file; the correlation value is used to characterize the association relationship; Obtaining a user input instruction, and extracting a first keyword from the user input instruction; Matching is performed in the knowledge graph according to the first keyword, a second material file is obtained according to the matching result, and a template file is generated according to the second material file; the generating of the template file according to the second material file includes: obtaining a number of historical template files, extracting typesetting rules and a third keyword in the typesetting rules from the historical template files; determining that the third keyword matches the first keyword, obtaining a target typesetting rule of the template file according to the frequency of the typesetting rule, and generating a template file according to the target typesetting rule based on the content of the second material file.

2. The intelligent recommendation method for network materials according to claim 1, characterized in that: The step of obtaining a user input instruction and extracting a first keyword from the user input instruction includes: Segmenting the user input command to obtain a plurality of words; The vocabulary is semantically matched with at least one tag of the first material file, and the first keyword is determined according to the matching result.

3. An intelligent recommendation system for network materials, characterized in that: include: A graph construction unit, configured to obtain a first material file, perform labeling preprocessing on the first material file, and construct a knowledge graph based on the first material file after the labeling preprocessing; The first material file is pre-processed into a label, The method comprises at least one of the following steps: obtaining the size information and format information of the first material file, and generating an attribute tag; Obtaining color values ​​of pixels in the first source file, and generating color labels according to the color values; obtaining image edges of the first source file, extracting image texture and shape features according to the image edges, and generating content labels according to the image texture and shape features; Generate a style label for the first material file through a pre-trained deep learning model; The step of constructing a knowledge graph based on the first material file after labeling preprocessing includes: obtaining a third material file based on at least one of the attribute label, the color label, the content label, and the style label; the third material file includes a public data file captured from the Internet, and the data file includes unstructured data content in the form of a web page; removing noise and invalid data in the third material file, unifying the format and content of the third material file and disassembling the content to obtain the elements of the third material file; taking at least one of the attribute label, the color label, the content label, and the style label as a second keyword, and generating a connecting line of the knowledge graph based on the association relationship between the element of the third material file and the second keyword; taking the second keyword and the element as nodes, and constructing the knowledge graph in combination with the connecting line; the step of obtaining the association relationship between the element of the third material file and the second keyword based on the element of the third material file includes: performing similarity matching based on the second keyword to filter the element from the third material file; determining a correlation value based on the content proportion of the element in the third material file; the correlation value is used to characterize the association relationship; An instruction acquisition unit, used to acquire a user input instruction and extract a first keyword from the user input instruction; A material screening unit, configured to match the first keyword in the knowledge graph, and screen and obtain a second material file according to the matching result; A template generation unit generates a template file according to the second material file; the generation of the template file according to the second material file includes: obtaining a number of historical template files, extracting typesetting rules and a third keyword in the typesetting rules from the historical template files; determining that the third keyword matches the first keyword, obtaining a target typesetting rule for the template file according to the frequency of the typesetting rule, and generating a template file according to the target typesetting rule based on the content of the second material file.

4. A computer device for intelligently recommending network materials, characterized in that: include: at least one processor; At least one memory, used to store at least one program; when the at least one program is executed by the at least one processor, the at least one processor runs the method as described in any one of claims 1-2.

5. A storage medium storing a program executable by a processor, characterized in that: The processor-executable program is used to run the intelligent recommendation method for network materials as described in any one of claims 1-2 when executed by the processor.

Images