Electric power industry technology frontiers and personalized recommendation of articles of application thereof

Abstract

The application provides a power industry science and technology front and application article personalized recommendation method and system, a storage medium and an electronic device, and relates to the field of personalized recommendation.In the application, the related field coefficient is used to screen sentences related to the personalized knowledge graph in the article to be recommended, and a new text after information filtering is generated; the title of the article to be recommended and the new text are used as the input of the BERT model, the title representation and the sentence representation are obtained respectively, the attention mechanism at the sentence level is introduced, the attention weight of the title representation to each sentence representation is obtained, the weighted sentence representation is obtained, and finally the classifier is input.The application uses the constructed personalized knowledge graph to screen sentences, preliminarily filters out irrelevant information, introduces the sentence-level attention model, filters the preliminarily screened sentences again, greatly reduces the interference of information irrelevant to the theme, and improves the performance of the classification model, so that the personalized recommendation result with high accuracy is quickly generated.

Description

Technical Field

[0001] This invention relates to the field of personalized recommendation, specifically to a personalized recommendation method, system, storage medium, and electronic device for articles on cutting-edge technologies and their applications in the power industry. Background Technology

[0002] In today's information-saturated era, online forums have become oceans of information, with a vast number of articles and complex content. This phenomenon has brought about many problems, the most prominent of which is that users find it difficult to quickly find the information they are interested in when faced with such a massive amount of data.

[0003] For researchers and other users, forums offer certain advantages, such as rapid article updates, allowing them to stay abreast of the latest research developments in their field. However, this also presents new challenges. Due to the sheer volume of forum articles, it's difficult to quickly and comprehensively find specialized content relevant to their research area. In academic research, time and energy are precious resources, requiring significant time to sift through and evaluate information, which impacts research efficiency. Therefore, recommending articles from public accounts based on individual interests and areas of expertise is invaluable. This personalized recommendation method helps users filter out a large amount of irrelevant information. For researchers and other users, personalized recommendations enable them to acquire specialized knowledge relevant to their research more efficiently, improving the accuracy and effectiveness of knowledge acquisition, thereby better promoting the progress of academic research. Summary of the Invention

[0004] (a) Technical problems to be solved

[0005] To address the shortcomings of existing technologies, this invention provides a personalized recommendation method, system, storage medium, and electronic device for articles on cutting-edge technologies and their applications in the power industry, solving the technical problem that users find it difficult to obtain information of interest in a timely manner.

[0006] (II) Technical Solution

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A personalized recommendation method for articles on cutting-edge technologies and their applications in the power industry includes:

[0009] Constructing a knowledge graph for the power industry;

[0010] Based on the aforementioned power industry knowledge graph, a personalized knowledge graph is constructed by integrating user preference information;

[0011] By using relevant domain coefficients, sentences related to the personalized knowledge graph in the articles to be recommended are selected, and new text after information filtering is generated;

[0012] The title of the article to be recommended and the new text are used as input to the BERT model to obtain title representation and several sentence representations respectively. A sentence-level attention mechanism is introduced to obtain the attention weight of the title representation to each sentence representation, so as to obtain the weighted sentence representation.

[0013] The weighted sentence representation and title representation are concatenated and used as input to the classifier to obtain the category probability of whether the user is interested in the recommended article, so as to generate personalized recommendation results for the user.

[0014] Preferably, the construction of the power industry knowledge graph includes:

[0015] Collect and preprocess articles on cutting-edge technologies and their applications in the power industry;

[0016] Based on the preprocessed articles, a graph ontology is constructed, including:

[0017] Keywords from articles are extracted using the term frequency-inverse text frequency algorithm and the TextRank algorithm. The articles are then clustered using the k-Means clustering algorithm. Preliminary domain concepts are obtained by manually observing the keywords and clustering results. These domain concepts are then modified and domain relationships and constraints are defined by combining them with a general graph ontology, which is used to construct several graph ontologs.

[0018] Based on the ontology structure of the graph ontology, entity data is obtained, and relationships between entities are obtained through manual screening using semi-supervised and remote supervision methods to construct the power industry knowledge graph.

[0019] Preferably, the step of constructing a personalized knowledge graph based on the power industry knowledge graph and integrating user preference information includes:

[0020] The user's browsing history articles are categorized by different themes, and key information for each browsing history article is extracted; wherein, the key information includes at least key concepts, technologies and trends;

[0021] Iterate through all historical articles for each topic, and search for the corresponding node or set of nodes in the power industry knowledge graph. If there is no completely corresponding node, create a new node.

[0022] Associate key information from each historically viewed article with the corresponding node, and establish new relationship connections when multiple nodes are involved.

[0023] Preferably, after the personalized knowledge graph is constructed, it is periodically optimized and updated based on new historical browsing articles.

[0024] Preferably, the step of using relevant domain coefficients to filter out sentences in the article to be recommended that are related to the personalized knowledge graph, and generating new text after information filtering, includes:

[0025] The article to be recommended is preprocessed to obtain a vocabulary list, and the frequency of each word in the vocabulary list in the article to be recommended is counted.

[0026] For each word in the vocabulary list, find the corresponding node in the personalized knowledge graph and obtain the betweenness centrality of that node;

[0027] For each sentence in the article to be recommended, the product of the median centrality of all words in each sentence and the word frequency is added together to obtain the relevant domain coefficient;

[0028] All sentences are sorted according to their domain relevance coefficients. The top n sentences with the highest scores are selected, concatenated, and then a new text with filtered information is generated.

[0029] Preferably, the introduction of a sentence-level attention mechanism obtains the attention weights of the title representation on each sentence representation, thereby obtaining a weighted sentence representation; expressed as:

[0030]

[0031] v seq =∑a i v i i∈{1, 2, ..., n}

[0032] Among them, a i Here, is the attention weight, and the subscript i is the sentence representation index; Q is the attention query vector, taking the title representation v. title ;K i For a vector matrix that needs to be assigned different attention weights, take the sentence representation v from the new text. i The superscript T indicates transpose; d is the scaling factor; v seq This is the weighted representation of the sentence.

[0033] A personalized recommendation system for articles on cutting-edge technologies and their applications in the power industry includes:

[0034] The first building module is used to construct a knowledge graph for the power industry.

[0035] The second construction module is used to construct a personalized knowledge graph based on the power industry knowledge graph and by integrating user preference information.

[0036] The filtering module is used to filter out sentences in the article to be recommended that are related to the personalized knowledge graph using relevant domain coefficients, and generate new text after information filtering;

[0037] The fusion module is used to take the title of the article to be recommended and the new text as input to the BERT model, obtain title representation and several sentence representations respectively, and introduce a sentence-level attention mechanism to obtain the attention weight of the title representation to each sentence representation, so as to obtain the weighted sentence representation.

[0038] The classification module is used to concatenate the weighted sentence representation and the title representation as input to the classifier to obtain the category probability of whether the user is interested in the recommended article, so as to generate personalized recommendation results for the user.

[0039] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, enables a personalized recommendation method for articles on cutting-edge technologies and their applications in the power industry, as described above.

[0040] An electronic device, comprising:

[0041] One or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs including a personalized recommendation method for performing articles on cutting-edge technologies in the power industry and their applications as described above.

[0042] (III) Beneficial Effects

[0043] This invention provides a personalized recommendation method, system, storage medium, and electronic device for articles on cutting-edge technologies and their applications in the power industry. Compared with existing technologies, it has the following advantages:

[0044] In this invention, relevant domain coefficients are used to filter sentences in the recommended article that are related to the personalized knowledge graph, generating new text after information filtering. The title of the recommended article and the new text are used as input to the BERT model to obtain title representation and several sentence representations, respectively. A sentence-level attention mechanism is introduced to obtain the attention weights of the title representation on each sentence representation, resulting in a weighted sentence representation, which is finally input into the classifier. This invention utilizes the constructed personalized knowledge graph for sentence filtering, initially filtering out irrelevant information. The introduction of a sentence-level attention model further filters the initially filtered sentences, greatly reducing interference from information unrelated to the topic, thereby improving the performance of the classification model and quickly generating highly accurate personalized recommendation results. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 A block diagram illustrating a personalized recommendation method for articles on cutting-edge technologies and their applications in the power industry, provided as an embodiment of the present invention.

[0047] Figure 2 A flowchart illustrating the construction of a knowledge graph for the power industry, provided as an embodiment of the present invention;

[0048] Figure 3 This is a schematic diagram of a sentence-level attention mechanism provided in an embodiment of the present invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are described clearly and completely. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] This application provides a personalized recommendation method, system, storage medium, and electronic device for articles on cutting-edge technologies and their applications in the power industry, solving the technical problem that users find it difficult to obtain information of interest in a timely manner.

[0051] The technical solution in this application is to solve the above-mentioned technical problems, and the general idea is as follows:

[0052] The purpose of this invention is to provide users with personalized recommendations of cutting-edge technologies and their applications in the power industry. Based on users' interests and needs, it aims to provide personalized, accurate, and novel recommendations of articles related to cutting-edge technologies and their applications, helping them to better understand the technological dynamics of the power industry and improve their professional knowledge. At the same time, it will also provide valuable market information for enterprises, helping them to better seize market opportunities.

[0053] The purpose of this invention is to construct a personalized recommendation system for cutting-edge technologies and their applications in the power industry. This system aims to provide users with personalized, accurate, and novel recommendations of articles related to cutting-edge technologies and their applications based on their interests and needs.

[0054] Specifically, the objectives of the embodiments of the present invention include the following aspects:

[0055] 1. Addressing the Information Overload Problem: The amount of information regarding cutting-edge technologies and applications in the power industry is vast and constantly growing, often making it difficult for users to sift through the massive amounts of data to find content that matches their interests and needs. Therefore, this invention aims to help users quickly find high-quality articles related to their interests through a personalized recommendation system, thus alleviating the distress caused by information overload.

[0056] 2. Enhanced User Experience: Personalized recommendation systems can tailor content recommendations based on user preferences and behavior, providing a more personalized and targeted reading experience. By analyzing user reading history, likes, and comments, the system can better understand user needs, thereby providing recommendations that better match their interests and improving user satisfaction and loyalty.

[0057] 3. Promoting Technological Development in the Power Industry: The development of cutting-edge technologies and their applications in the power industry is crucial for the sustainable development of the energy sector. Through personalized recommendation systems, users can stay informed about the latest technological advancements, innovative applications, and research findings, promoting technical exchange and cooperation within the industry and driving technological innovation and development in the power sector.

[0058] 4. Supporting Decision Making: Power industry decision-makers need to understand the latest technological frontiers and application scenarios when formulating strategies and plans. Personalized recommendation systems can provide decision-makers with targeted and practical articles related to cutting-edge technologies and applications, helping them obtain key information and make informed decisions.

[0059] In summary, the embodiments of this invention aim to construct a personalized recommendation system for cutting-edge technologies and their applications in the power industry, in order to solve the problem of information overload, improve user experience, promote the technological development of the power industry, and support decision-making. Through accurate, diverse, and novel recommendation results, it provides users with valuable content related to cutting-edge technologies and applications, meeting user needs and promoting the development and innovation of the power industry.

[0060] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0061] Example 1:

[0062] like Figure 1 As shown, this embodiment of the invention provides a personalized recommendation method for articles on cutting-edge technologies and their applications in the power industry, including:

[0063] S1. Construct a knowledge graph for the power industry;

[0064] S2. Based on the power industry knowledge graph, integrate user preference information to construct a personalized knowledge graph;

[0065] S3. Using relevant domain coefficients, sentences related to the personalized knowledge graph in the articles to be recommended are selected, and new text after information filtering is generated;

[0066] S4. The title of the article to be recommended and the new text are used as input to the BERT model to obtain the title representation and several sentence representations respectively. A sentence-level attention mechanism is introduced to obtain the attention weight of the title representation to each sentence representation in order to obtain the weighted sentence representation.

[0067] S5. The weighted sentence representation and title representation are concatenated and used as input to the classifier to obtain the category probability of whether the user is interested in the recommended article, so as to generate personalized recommendation results for the user.

[0068] This invention utilizes a constructed personalized knowledge graph for sentence filtering, initially removing irrelevant information. Then, a sentence-level attention model is introduced to further filter the initially filtered sentences, greatly reducing interference from information unrelated to the topic. This improves the performance of the classification model and quickly generates highly accurate personalized recommendation results.

[0069] The following will detail each step of the above solution:

[0070] In step S1, a knowledge graph of the power industry is constructed.

[0071] It should be noted that building a knowledge base for the power industry is a complex and systematic process, requiring in-depth analysis and integration of extensive data and information resources from multiple perspectives. First, it's necessary to understand the core components of a power industry knowledge base. It's not merely a data storage platform, but a framework for structuring and organizing relevant information within the power industry. The power industry encompasses multiple aspects, including power generation, transmission, distribution, and new energy technologies, each with its unique data and information needs. Therefore, building a knowledge base that meets these needs requires multi-layered data collection and processing.

[0072] See Figure 2 This step specifically includes the following steps:

[0073] S11. Collect and preprocess articles on cutting-edge technologies and their applications in the power industry.

[0074] Data collection is the first step in building a knowledge base for the power industry. Data sources in the power industry are extensive, including operational data from power companies, statistical data from government departments, research papers from academic journals, and industry reports and analyses. This data covers a wide range of topics, from the operation of power generation facilities to economic analysis of the energy market. Collecting this data requires not only extensive and systematic searching but also effective filtering and organization to ensure the accuracy and usability of the data.

[0075] The collected data is typically categorized into structured and unstructured data sources. Structured data sources primarily include operational data from power companies, government statistics, industry reports, and internal reports from power companies. Unstructured data requires the use of text mining and natural language processing techniques to extract information from unstructured data such as news articles, academic papers, and professional journals. This study collected data from 68 power industry publications, including the State Grid News and Electrical Times, on relevant forums and from power industry journals, yielding a large number of articles with professional knowledge about the power industry.

[0076] S12. Construct a graph ontology based on the preprocessed articles.

[0077] It should be noted that the ontology is the core model of the power industry knowledge base. It defines the concepts of various entities in the power industry (such as power plants, substations, transmission lines, etc.) and the relationships between them. The ontology construction process usually begins with identifying the core concepts and entities, and then defining their attributes and relationships.

[0078] For example, a power plant may have attributes such as power generation capacity, power generation technology, and fuel type, and the power plant and transmission lines can be connected through a power supply relationship.

[0079] By establishing an ontology, a clear framework can be provided for the structuring of knowledge bases and information retrieval. Ontology construction typically involves collaboration between domain experts and knowledge engineers to jointly design and optimize the ontology model to ensure its accuracy and usability. This invention constructs a graph ontology by using statistical methods combined with manual induction, drawing on high-quality general graphs, and expert guidance.

[0080] Specifically, the ontology construction process includes:

[0081] 1) Extract keywords from the articles using the term frequency-inverse text frequency algorithm and the TextRank algorithm; 2) Cluster the articles using the k-Means clustering algorithm, and obtain preliminary domain concepts by manually observing keywords and clustering results; 3) Modify the domain concepts by combining general graphs and define domain relationships and their constraints, and construct several graph ontology.

[0082] S13. Based on the ontology structure of the graph ontology, acquire entity data, and combine semi-supervised and remote supervision methods to obtain the relationships between entities through manual screening in order to construct the power industry knowledge graph.

[0083] Here, the constructed ontology is then checked and evaluated, and after modification and improvement, the final ontology for the science popularization domain is obtained. Based on the defined ontology structure for the science popularization domain, a large amount of well-structured and high-quality entity data is obtained from publicly available knowledge graphs on the Internet. Relationships between entities are obtained through manual screening, combining semi-supervised and remote supervision methods. Finally, the constructed knowledge graph is checked and evaluated under the guidance of experts in the power industry.

[0084] In step S2, based on the power industry knowledge graph, user preference information is integrated to construct a personalized knowledge graph; including:

[0085] S21. Categorize users' historical browsing articles by different themes, such as new energy power generation technology (including sub-themes such as solar, wind, and hydropower), smart grid technology, and power market dynamics. Further, it can record the article source and publication time. It also extracts key information from each historically browsed article, and can further mark its importance and relevance to the power industry.

[0086] S22. Traverse all historical articles for each article topic, and search for the corresponding node or set of nodes in the power industry knowledge graph. If there is no completely corresponding node, create a new node.

[0087] S23. Associate the key information of each historically viewed article with the corresponding node, and establish new relationship connections when multiple nodes are involved.

[0088] Specifically, after the personalized knowledge graph is constructed, it may also include:

[0089] S24. Based on the new historical browsing articles, regularly optimize and update the personalized knowledge graph.

[0090] It should be noted that value information filtering is the part of this model that extracts and filters the value of text information. It consists of two steps: First, it uses a knowledge graph to calculate sentence relevance coefficients and filters and selects sentences in each text based on these coefficients. Then, it uses BERT as the encoder to extract features from all sentences in the title and text, and uses an attention mechanism to assign different weights to each sentence to measure their importance to the article.

[0091] For details, please refer to steps S3 to S4:

[0092] In step S3, relevant domain coefficients are used to filter out sentences in the article to be recommended that are related to the personalized knowledge graph, and new text after information filtering is generated.

[0093] S31. Calculate the betweenness centrality:

[0094] The Brandes algorithm is used to calculate the betweenness centrality of each node in a knowledge graph. The basic idea of ​​the Brandes algorithm is to determine the betweenness centrality of a node by calculating the number of times each shortest path passes through it. The specific calculation formula is as follows:

[0095]

[0096] Among them, C B (v) denotes the betweenness centrality of node v, where V is the set of all nodes in the knowledge graph, and s and t are any two nodes in V other than v. σ st σ represents the total number of shortest paths from node s to node t. st (v) represents the number of nodes v that are passed through in the shortest path from node s to node t.

[0097] S32. Calculation of Domain Relevance Coefficient:

[0098] First, the articles to be recommended are preprocessed, including word segmentation, removal of stop words, etc., to obtain a vocabulary list, and the frequency of each word in the vocabulary list in the articles to be recommended is counted, i.e., word frequency.

[0099] Secondly, for each word in the vocabulary list, find the corresponding node in the personalized knowledge graph and obtain the betweenness centrality of that node.

[0100] Next, for each sentence in the article to be recommended, the products of the betweenness centrality and word frequency of all words in each sentence are summed to obtain the relevant domain coefficient p. The mathematical expression is:

[0101]

[0102] Here, tf represents word frequency. Through the above steps, the coefficient of the power industry-related field based on betweenness centrality and word frequency can be calculated. This coefficient can reflect the degree of relevance between the text and the technological frontiers and applications of the power industry.

[0103] Finally, all sentences are sorted according to their domain relevance coefficients, and the top n sentences with the highest scores are selected, concatenated, and a new text with filtered information is generated.

[0104] In step S4, the title of the article to be recommended and the new text are used as input to the BERT model to obtain title representation and several sentence representations respectively. A sentence-level attention mechanism is introduced to obtain the attention weight of the title representation on each sentence representation to obtain the weighted sentence representation.

[0105] It should be noted that BERT (Bidirectional Encoder Representations from Transformers) is a pre-trained language model based on the Transformer architecture. It learns rich linguistic knowledge and semantic representations through unsupervised learning on large-scale texts. In text classification, BERT's main role is feature extraction and representation learning. It can transform the input text into low-dimensional, dense vector representations containing the text's semantic information. The BERT model consists of multiple layers of bidirectional Transformer decoders, and the core of the Transformer model is the attention mechanism. Transformers use multi-head attention to capture long-distance dependencies in the text, enabling better acquisition of semantic information about the target word and its context in various semantic scenarios. The formula for calculating H is:

[0106] H = Concat(h) 1, h2,…,h i )*W 0

[0107] Among them, W 0 The weight matrix represents the redistribution of weights to the output vector; h i h represents the output vector of the i-th head. i The calculation formula is:

[0108] h i =Attention(Q) i ,K i V i )

[0109]

[0110] Among them, Q i ,K i V i These are query, key, and value matrices, d k It is the dimension of the key. Through multi-head attention mechanisms, different positions and semantic information in the text can be focused on simultaneously.

[0111] In the BERT model, [CLS] is a special character that does not possess any semantic meaning on its own. It can be used to represent the comprehensive semantic information of the input text. In this embodiment, the embedding representation of this character is used as the feature vector v of the input text. The new text and title obtained in the previous steps are input into the BERT model respectively. The vector containing the [CLS] character is taken as the representation of each sentence and the title, respectively, v. doc =[v1,v2,...,v n ]、v title .

[0112] It is important to note that not every short sentence in a long text contains semantic information equally relevant to the text's topic. To evaluate the semantic importance of different short sentences in a long text for the classification task, this model uses an attention mechanism to obtain the attention weight of the title for each short sentence. Based on the attention weight of each short sentence, the sentence vectors of each short sentence are weighted and summed to obtain the short sentence feature vectors with more significant classification features. In long natural language texts, the title at the beginning of the text often contains the topic sentence or a summary sentence; its semantic content helps in extracting and analyzing the text's content features. Therefore, the model concatenates the short sentence feature vectors with the title feature vector as the text classification feature vector.

[0113] Therefore, in obtaining the representation v of each sentence doc =[v1,v2,...,v n ] and title representation v title Afterwards, as Figure 3 As shown, this step introduces a sentence-level attention mechanism to obtain the attention weights of the title representation on each sentence representation, thus obtaining a weighted sentence representation; represented as:

[0114]

[0115] v seq =∑a i v i i∈{1, 2, ..., n}

[0116] Among them, a i Here, is the attention weight, and the subscript i is the sentence representation index; Q is the attention query vector, taking the title representation v. title ;K i For a vector matrix that needs to be assigned different attention weights, take the sentence representation v from the new text. i The superscript T indicates transpose; d is the scaling factor; v seq This is the weighted representation of the sentence.

[0117] In step S5, the weighted sentence representation and title representation are concatenated and used as input to the classifier to obtain the category probability of whether the user is interested in the recommended article, so as to generate personalized recommendation results for the user.

[0118] like Figure 3 As shown, the attention mechanism is used to obtain the attention corresponding to the category of the sentence vector. Then, the sentence vectors are aggregated using the attention mechanism to obtain v. output This vector contains text information that has been filtered and attention-based.

[0119] This step will v output Inputting the data into a classifier consisting of fully connected layers and a softmax function yields the class probabilities:

[0120] p = Softmax(Wv) output +b)

[0121] Finally, personalized recommendations are generated based on the probability of whether a user is interested in a recommended article by category.

[0122] For example: the closer the category probability is to 1, the more interested the user is, and the article to be recommended can be recommended to the user; conversely, the closer it is to 1, the less interested the user is, and the article to be recommended can not be recommended to the user.

[0123] Thus, this embodiment of the invention completes the entire process of a personalized recommendation method for articles on cutting-edge technologies and their applications in the power industry.

[0124] Example 2:

[0125] This invention provides a personalized recommendation system for articles on cutting-edge technologies and their applications in the power industry, comprising:

[0126] The first building module is used to construct a knowledge graph for the power industry.

[0127] The second construction module is used to construct a personalized knowledge graph based on the power industry knowledge graph and by integrating user preference information.

[0128] The filtering module is used to filter out sentences in the article to be recommended that are related to the personalized knowledge graph using relevant domain coefficients, and generate new text after information filtering;

[0129] The fusion module is used to take the title of the article to be recommended and the new text as input to the BERT model, obtain title representation and several sentence representations respectively, and introduce a sentence-level attention mechanism to obtain the attention weight of the title representation to each sentence representation, so as to obtain the weighted sentence representation.

[0130] The classification module is used to concatenate the weighted sentence representation and the title representation as input to the classifier to obtain the category probability of whether the user is interested in the recommended article, so as to generate personalized recommendation results for the user.

[0131] Example 3:

[0132] This invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement a personalized recommendation method for articles on cutting-edge technologies and their applications in the power industry, as described in Embodiment 1.

[0133] Example 4:

[0134] This invention provides an electronic device, comprising:

[0135] One or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs including a personalized recommendation method for performing articles on cutting-edge technologies and their applications in the power industry as described in Example 1.

[0136] It is understood that the personalized recommendation system, computer-readable storage medium, and electronic device for articles on the frontiers of power industry technology and their applications provided in this embodiment of the invention correspond to the personalized recommendation method for articles on the frontiers of power industry technology and their applications. The explanations, examples, and beneficial effects of the relevant content can be referred to the corresponding parts of the personalized recommendation method for articles on the frontiers of power industry technology and their applications, and will not be repeated here.

[0137] In summary, compared with existing technologies, it has the following beneficial effects:

[0138] 1. In this embodiment of the invention, a personalized knowledge graph is constructed to filter sentences, initially filtering out irrelevant information. A sentence-level attention model is introduced to filter the sentences after the initial screening, which greatly reduces the interference of information unrelated to the topic, thereby improving the performance of the classification model and quickly generating highly accurate personalized recommendation results.

[0139] 2. This invention incorporates knowledge graphs for sentence filtering: Addressing the issues of lengthy popular science articles making it difficult for models to focus on key information and the poor classification performance of traditional models, this invention proposes a method for identifying the value of professionally generated content in the power industry by combining sentence filtering with knowledge graphs. This filters out irrelevant information, reduces interference from topic-irrelevant content, and improves the model's classification performance.

[0140] 3. This invention introduces a sentence-level attention model: To more effectively reduce interference from topic-irrelevant information and thus improve the performance of the classification model, this paper introduces a sentence-level attention model. This model further filters the sentence information after initial screening. Specifically, different sentences in an article contribute differently to understanding the article content, and the sentence-level attention model can calculate the contribution of sentence vectors to distinguishing article categories. In this way, more attention can be paid to sentence information that is more important to the classification task, further removing less relevant information, thereby improving the model's accuracy and performance.

[0141] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0142] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A personalized recommendation method for articles on cutting-edge technologies and their applications in the power industry, characterized in that, include: Constructing a knowledge graph for the power industry; Based on the aforementioned power industry knowledge graph, a personalized knowledge graph is constructed by integrating user preference information; wherein, the preference information refers to key information from the user's historical browsing articles, and the key information includes at least key concepts, technologies and trends; By using relevant domain coefficients, sentences related to the personalized knowledge graph in the articles to be recommended are selected, and new text after information filtering is generated; The title of the article to be recommended and the new text are used as input to the BERT model to obtain title representation and several sentence representations respectively. A sentence-level attention mechanism is introduced to obtain the attention weight of the title representation to each sentence representation, so as to obtain the weighted sentence representation. The weighted sentence representation and title representation are concatenated and used as input to the classifier to obtain the category probability of whether the user is interested in the recommended article, so as to generate personalized recommendation results for the user. The process of using relevant domain coefficients to filter out sentences in the recommended article that are related to the personalized knowledge graph, and generating new text after information filtering, includes: The article to be recommended is preprocessed to obtain a vocabulary list, and the frequency of each word in the vocabulary list in the article to be recommended is counted. For each word in the vocabulary list, find the corresponding node in the personalized knowledge graph and obtain the betweenness centrality of that node; For each sentence in the article to be recommended, the product of the median centrality of all words in each sentence and the word frequency is added together to obtain the relevant domain coefficient; All sentences are sorted according to their domain relevance coefficients. The top n sentences with the highest scores are selected, concatenated, and then a new text with filtered information is generated.

2. The personalized recommendation method as described in claim 1, characterized in that, The construction of the power industry knowledge graph includes: Collect and preprocess articles on cutting-edge technologies and their applications in the power industry; Based on the preprocessed articles, a graph ontology is constructed; Based on the ontology structure of the graph ontology, entity data is obtained, and the relationships between entities are obtained through manual screening by combining semi-supervised and remote supervision methods to construct the power industry knowledge graph. The construction of the graph ontology based on the preprocessed articles includes: Keywords from articles are extracted using the term frequency-inverse text frequency algorithm and the TextRank algorithm. The articles are then clustered using the k-Means clustering algorithm. Preliminary domain concepts are obtained by manually observing the keywords and clustering results. These domain concepts are then modified by combining them with a general graph, and domain relationships and their constraints are defined to construct several graph ontology.

3. The personalized recommendation method as described in claim 1, characterized in that, The process of constructing a personalized knowledge graph based on the power industry knowledge graph and integrating user preference information includes: The user's browsing history articles are categorized by different themes, and key information for each browsing history article is extracted; Iterate through all historical articles for each topic, and search for the corresponding node or set of nodes in the power industry knowledge graph. If there is no completely corresponding node, create a new node. Associate key information from each historically viewed article with the corresponding node, and establish new relationship connections when multiple nodes are involved.

4. The personalized recommendation method as described in claim 3, characterized in that, After the personalized knowledge graph is constructed, it is regularly optimized and updated based on new historical browsing articles.

5. The personalized recommendation method as described in claim 1, characterized in that, The introduced sentence-level attention mechanism obtains the attention weights of the title representation for each sentence representation, resulting in a weighted sentence representation, expressed as follows: in, Attention weights, subscript Sentence representation index; The vector for the attention query is represented by the title. ; For a vector matrix that needs to be assigned different attention weights, take the sentence representation from the new text. superscript Indicates transpose; This is the scaling factor; This is the weighted representation of the sentence.

6. A personalized recommendation system for articles on cutting-edge technologies and their applications in the power industry, characterized in that, For performing the personalized recommendation method as described in claim 1, including: The first building module is used to construct a knowledge graph for the power industry. The second construction module is used to construct a personalized knowledge graph based on the power industry knowledge graph and by integrating user preference information. The filtering module is used to filter out sentences in the article to be recommended that are related to the personalized knowledge graph using relevant domain coefficients, and generate new text after information filtering; The fusion module is used to take the title of the article to be recommended and the new text as input to the BERT model, obtain title representation and several sentence representations respectively, and introduce a sentence-level attention mechanism to obtain the attention weight of the title representation to each sentence representation, so as to obtain the weighted sentence representation. The classification module is used to concatenate the weighted sentence representation and the title representation as input to the classifier to obtain the category probability of whether the user is interested in the recommended article, so as to generate personalized recommendation results for the user.

7. A computer-readable storage medium, characterized in that, It stores a computer program, which, when executed by the processor, can implement the personalized recommendation method for articles on cutting-edge technologies and their applications in the power industry as described in any one of claims 1 to 5.

8. An electronic device, characterized in that, include: One or more processors; Memory; And one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs including a personalized recommendation method for performing articles on the frontiers of power industry technology and their applications as described in any one of claims 1 to 5.

Images