Brief generation method and apparatus

Abstract

The application belongs to the technical field of artificial intelligence, and discloses a briefing generation method and device, the method comprising: screening reference information from all network information content according to a preset multi-mode channel feature configuration library; extracting reference features from the reference information features; expanding each reference feature to obtain an information briefing data set; processing the information briefing data set based on a preset feedback neural network to obtain a headline, a report core and a report tail, and generating a briefing according to the headline, the report core and the report tail; the application adopts a novel adaptive information collection method based on a general configuration library and a label library, configures a label of interest to a user, acquires effective content by using an effective keyword configuration feature library and a sensitive word configuration library, shortens the information transmission time of a scientific research briefing, improves the briefing work efficiency, and effectively avoids the problem of wasting a large amount of time cost caused by manually screening information and manually writing due to the related work of information collection and writing of a briefing.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method and apparatus for generating presentations. Background Technology

[0002] Information technology is a crucial foundation for modern production. Information has gradually become a strategic, tactical, and instrumental resource, and competition among enterprises is increasingly manifested as a competition of comprehensive strength centered on information technology. The processing and application of information has become an important component of modern enterprises.

[0003] With the development of information technology and changes in people's lifestyles, the rise and popularization of social media have placed higher demands on how to capture the value contained in data resources and provide fast, timely, concise, and highly valuable research briefings. These demands extend to information gathering, processing, generation, and analysis.

[0004] Current information gathering has transitioned from traditional manual collection to automated information technology-based data acquisition, with web crawler frameworks gaining increasing popularity. However, due to the unique designs of each website requiring different processing methods, and the need to update code for website redesigns, fully automated data collection and management are not yet achieved in most sectors in my country. In today's rapidly evolving information landscape, traditional information gathering methods will gradually become obsolete, necessitating the development of new approaches. Current information processing typically employs extractive text summarization and generative text summarization. Extractive text summarization uses simple algorithms to extract three sentences to form a summary. Generative text summarization combines keywords and sentence order, first iterating through keywords, then sequentially traversing sentences until the first occurrence is found and used as the summary. Summary of the Invention

[0005] The main objective of this invention is to provide a method and apparatus for generating briefings, which aims to solve the technical problem that the collection and writing of information for briefings in the prior art often requires manual information screening and writing, which easily leads to a large waste of time and resources.

[0006] To achieve the above objectives, the present invention provides a presentation generation method, the method comprising the following steps:

[0007] Multiple reference articles are selected from the entire network's information content based on a pre-defined multi-mode channel feature configuration library;

[0008] Feature extraction is performed on the multiple reference information to obtain reference features for each reference information, and the reference features include information title, information text, information author, and information publication time;

[0009] The reference features are expanded to obtain the information briefing dataset;

[0010] The information briefing dataset is processed based on a preset feedback neural network to obtain a header, core, and tail, and a briefing is generated based on the header, core, and tail.

[0011] Optionally, the step of filtering multiple reference information from the entire network information content according to a preset multi-mode channel feature configuration library includes:

[0012] The webpage content is obtained by searching the monitoring information channel feature set in the preset multi-mode channel feature configuration library;

[0013] The webpage content is extracted to obtain initial content features;

[0014] The initial content features are matched with the information feature set in the preset multi-mode channel feature configuration library according to the preset feedback neural network;

[0015] The successfully matched webpage content is learned through a preset feedback neural network to obtain new features, and the information feature set in the preset multi-mode channel feature configuration library is updated according to the new features;

[0016] Use the content of the successfully matched web pages as reference information.

[0017] Optionally, the step of extracting features from the plurality of reference information to obtain reference features for each reference information includes the information title, information text, information author, and information publication time, including:

[0018] The title weight of each reference news item is calculated based on the news title features in the preset multi-mode channel feature configuration library, and features are extracted based on the title weights.

[0019] Calculate the paragraph link density and text density of the main text of each reference information, and extract features based on the paragraph link density and text density to obtain the main text of the information;

[0020] The time is extracted from the Uniform Resource Locator of each reference information by regular expression to obtain the initial time feature. The initial time is then formatted to obtain the information publication time.

[0021] The author of the information is obtained by matching the author characteristics of the publisher in the preset multi-mode channel feature configuration library with the reference information.

[0022] Optionally, after extracting features from the plurality of reference information to obtain reference features for each reference information, the method further includes:

[0023] The information body and title of each reference information are matched with the monitoring tag feature set in the preset multi-mode channel feature configuration library, and the relevance of the reference information is determined based on the matching result.

[0024] The body text and title of each reference information are matched with the sensitive word tag feature set in the preset multi-mode channel feature configuration library, and the reference information is determined as invalid based on the matching result.

[0025] Text keywords are extracted from the main text and title of the reference information to obtain text keywords, and the quality requirements of the reference information are determined based on the text keywords.

[0026] By removing irrelevant, invalid, and substandard reference information from the reference information, multiple reference information is obtained.

[0027] Optionally, the step of extending the reference features to obtain the information briefing dataset includes:

[0028] Use the title of each reference article as an article identifier, and convert the article title, article text, article author, and article time into an article ID;

[0029] The information title configuration feature, information body configuration feature, information publication time configuration feature, and information author configuration that successfully match each reference information with the information feature set in the preset multi-mode channel feature configuration library are used as the feature target values;

[0030] An information directory is generated based on the information identifier, the feature target value, and the information ID. Multiple information directories are used to construct an information briefing dataset.

[0031] Optionally, the step of generating an information directory based on the information identifier, the feature target value, and the information ID, and constructing an information briefing dataset based on multiple information directories, includes:

[0032] A news catalog is generated based on the news identifier, the feature target value, and the news ID;

[0033] Multiple keywords are determined based on the article text and article title of the reference information, and article tags are generated based on the keywords;

[0034] Based on the text and title of the reference information, adjacent word groups are constructed, and keyword phrases are obtained by combining the adjacent word groups.

[0035] The information directory is updated based on the information tags and keyword phrases to obtain a reference directory, and an information briefing dataset is constructed based on the reference information directory.

[0036] Optionally, updating the information directory based on the information tags and keyword phrases to obtain a reference directory, and constructing an information briefing dataset based on the reference information directory, further includes:

[0037] The information directory is updated based on the information tags and keyword phrases to obtain a reference directory;

[0038] An article summary is generated based on the main text of the referenced information;

[0039] The reference information directory is updated based on the article summary, and the updated reference information directory forms the information summary dataset.

[0040] Optionally, the step of processing the information briefing dataset based on a preset feedback neural network to obtain a header, core, and footer, and generating a briefing based on the header, core, and footer, includes:

[0041] Based on a preset feedback neural network, a header is generated by configuring feature tags for the information in the information briefing dataset and configuring scientific research information units.

[0042] A report is generated based on the corresponding scientific research information configuration in the information briefing dataset, the article briefings, keywords, and information catalog in the information briefing dataset;

[0043] Generate a report tail based on the corresponding configured scientific research information occurrence range in the information briefing dataset;

[0044] A briefing is generated based on the header, core, and footer.

[0045] Optionally, the step of generating a report based on the corresponding scientific research information configuration in the information briefing dataset, the article briefings, keywords, and information directory in the information briefing dataset includes:

[0046] Feature extraction is performed based on the corresponding scientific research information configuration in the information briefing dataset to obtain an effective directory;

[0047] Multiple briefing captions are generated based on the reference information in the information briefing dataset, and the multiple briefings are combined to obtain the target caption.

[0048] By performing word frequency statistics on the keywords in the information briefing dataset, multiple keywords are clustered to obtain word groups, and the core keyword information of the briefing is obtained based on the word groups.

[0049] Generate article titles based on keywords and phrases in the information briefing dataset, and generate article introductions based on article briefings in the information briefing dataset;

[0050] The article source, author information, and author information are obtained from the aforementioned news briefing dataset;

[0051] A report is generated based on the effective directory, the target annotation, the core keyword information of the briefing, the article introduction, the article source, the article author information, and the author information.

[0052] Furthermore, to achieve the above objectives, the present invention also proposes a presentation generation apparatus, the presentation generation apparatus comprising:

[0053] The information acquisition module is used to filter multiple reference information from the entire network of information content based on a preset multi-mode channel feature configuration library;

[0054] The information acquisition module is also used to extract features from the plurality of reference information to obtain reference features for each reference information, wherein the reference features include information title, information text, information author, and information publication time;

[0055] The briefing generation module is used to perform feature expansion on the reference features to obtain an information briefing dataset.

[0056] The briefing generation module is also used to process the information briefing dataset based on a preset feedback neural network to obtain a header, a core, and a footer, and to generate a briefing based on the header, core, and footer.

[0057] This invention provides a novel adaptive information collection method based on a general configuration library and tag library. By configuring tags that users are interested in, and using an effective keyword configuration feature library and a sensitive word configuration library, it obtains effective web pages / content, shortens the information transmission time of scientific research briefings, improves the efficiency of briefing work, and effectively avoids the problem of a large amount of time wasted due to the need for manual information screening and writing of briefing information. Attached Figure Description

[0058] Figure 1 This is a schematic diagram of the structure of the presentation generation device for the hardware operating environment involved in the embodiments of the present invention;

[0059] Figure 2 This is a flowchart illustrating the first embodiment of the present invention's briefing generation method;

[0060] Figure 3 This is a schematic diagram illustrating the complete presentation generation steps of an embodiment of the presentation generation method of the present invention;

[0061] Figure 4 This is a flowchart illustrating the second embodiment of the present invention's briefing generation method;

[0062] Figure 5This is a flowchart illustrating the third embodiment of the present invention's briefing generation method;

[0063] Figure 6 This is a structural block diagram of the first embodiment of the present invention's briefing generation device.

[0064] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0065] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0066] Reference Figure 1 , Figure 1 This is a schematic diagram of the hardware operating environment for generating a briefing device involved in an embodiment of the present invention.

[0067] like Figure 1 As shown, the presentation generation device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wireless-Fidelity (Wi-Fi) interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk drive. The memory 1005 may also optionally be a storage device independent of the aforementioned processor 1001.

[0068] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the presentation generation device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0069] like Figure 1 As shown, the memory 1005, which serves as a storage medium, may include an operating system, a network communication module, a user interface module, and a presentation generation program.

[0070] exist Figure 1In the presentation generation device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the presentation generation device of the present invention can be set in the presentation generation device, and the presentation generation device calls the presentation generation program stored in the memory 1005 through the processor 1001 and executes the presentation generation method provided in the embodiment of the present invention.

[0071] This invention provides a method for generating a briefing, referring to... Figure 2 , Figure 2 This is a flowchart illustrating a first embodiment of a briefing generation method according to the present invention.

[0072] In this embodiment, the briefing generation method includes the following steps:

[0073] Step S10: Select multiple reference information from the entire network information content according to the preset multi-mode channel feature configuration library.

[0074] It is understandable that the preset multi-mode channel feature configuration library can be understood as a database that includes multiple feature sets, which may include monitoring information channel feature sets, monitoring tag feature sets, sensitive word tag feature sets, and information feature sets.

[0075] It should be noted that the preset multi-mode channel feature configuration library first builds a feature library. The feature library can be initialized by collecting dynamics on technological innovation from internal groups (research institutes, Guangdong Innovation Institute, Zhejiang Innovation Institute, Guangdong Company, Zhejiang Company, relevant professional institutions, etc.) and external industry companies and institutions (such as China Telecom / China Unicom Research Institute, Internet companies, etc.) through public channels.

[0076] Then configure the multi-mode channel feature configuration library: configure the monitoring information channel feature set R1 involved in the task: configure the news information platform address and information platform name, such as: Jiangsu Development and Reform Commission, CCID Consulting, etc.; configure the WeChat official account new media address and information platform name, such as: WeChat official account, Weibo, etc.

[0077] Configure monitoring tag feature set R2: Configure brand tags, such as: Guangdong Innovation Institute, Zhejiang Mobile Company, China Telecom Research Institute, etc.; Configure product tags, such as: blockchain, wireless base station, industrial internet, etc.

[0078] Configure the sensitive word tag feature set R3: Configure suspected sensitive word tags, such as: summer vacation, internship, recruitment, etc.; confirm the algorithm to identify sensitive word tags. Based on the tags extracted from invalid articles identified by the algorithm with a word frequency greater than 20, manual confirmation is required to add the tags as suspected sensitive word tags.

[0079] Configure the news feature sets Ri1 to Ri4: extract webpage features based on webpage analysis algorithms and candidate URL ranking algorithms. The news feature sets mainly include news title features, news text features, publication time features, and publication author features, which are configured primarily through matching and regular expressions.

[0080] This process involves feature extraction of the required information, task creation based on generated tasks, and matching these tasks with task templates. After task initialization, relevant configuration information can be set for the created tasks. These configuration tasks are then integrated into the current page's feature configuration, allowing for reconfiguration of extracted and labeled information during subsequent operations.

[0081] It should be noted that the preset multi-mode channel feature configuration library can be continuously updated. After each new information is collected, features that are not in the preset multi-mode channel feature configuration library can be added to it based on the consultation.

[0082] It should be understood that the information content across the entire network can be the information obtained by using mature web crawling technology to search or access the multi-mode channel configuration library R1 configured in step one, while the reference information is the information content obtained after filtering the information across the entire network.

[0083] It should be further explained that the preset multi-mode channel feature configuration library can be continuously updated. It can be used to search for webpage content based on the monitoring information channel feature set in the preset multi-mode channel feature configuration library; extract content from the webpage content to obtain initial content features; match the initial content features with the information feature set in the preset multi-mode channel feature configuration library using a preset feedback neural network; learn new features from the unmatched webpage content through the preset feedback neural network; update the information feature set in the preset multi-mode channel feature configuration library based on the new features; and use the successfully matched webpage content as reference information.

[0084] In practice, mature web crawling technology is used to iteratively search or access the multi-mode channel configuration library configured in step one, returning all links and absolute links contained in the search object to derive the structure and content of the webpage. Each article that matches the multi-mode channel feature configuration library is marked with its status and written to the historical information database for unified management of all information articles. This prevents the repeated extraction of the same article in subsequent article extraction stages.

[0085] The extracted content is matched with features in a multi-mode channel feature configuration library using a recursive feedback neural network model. Unmatched parts of the reference information are identified, and the recursive feedback neural network continuously learns unknown features, updating and supplementing the corresponding feature library. The webpage content is then compared and contrasted with known configuration libraries, and the website is evaluated based on a website weighting algorithm. The validity of the currently acquired data is determined, and the feature libraries for news titles, news text, publication time, and publishers are continuously improved.

[0086] It should be noted that the execution subject of this embodiment is a presentation generation device, which has functions such as data processing, data communication and program execution. The presentation generation device can be an integrated controller, a control computer or other devices, or other devices with similar functions. This embodiment does not limit the scope of the invention.

[0087] Step S20: Extract features from the multiple reference information to obtain reference features for each reference information. The reference features include information title, information text, information author, and information publication time.

[0088] It should be noted that before feature extraction from the reference information, data cleaning is performed to remove redundant fields and standardize specifications.

[0089] It should be noted that the step of extracting features from the multiple reference information to obtain reference features for each reference information includes information title, information body, information author, and information publication time. This includes: calculating the title weight of each reference information based on the information title features in the preset multi-mode channel feature configuration library, and extracting features based on the title weight.

[0090] The news title feature extraction process can be as follows: During the initial news title extraction, the corresponding Document Object Model (DOM) weights are calculated based on the configured news title features. The corresponding DOM blocks are then analyzed to extract the webpage title. If the initial extraction fails, a news title can be generated based on the extracted webpage text density and text feature algorithms. Here, the news title can be simply understood as a type of news title. The jieba algorithm (an open-source library developed by Baidu engineer Sun Junyi) can be used with an existing stop word library to divide the webpage into a potential title area and a body text area. By removing markers from the body text area, webpage noise interference is reduced, and the candidate body text area is accurately extracted. An undirected weighted graph model (a graph-based ranking algorithm for text, abbreviated as TextRank) is used to calculate the weight set for each news item. An improved similarity calculation method is used to extract the news title from the candidate body text area. Simultaneously, the article news title extraction status is marked.

[0091] Calculate the paragraph link density and text density of the main text of each reference information, and extract features based on the paragraph link density and text density to obtain the main text of the information;

[0092] In practical implementation, the main text can be marked with tags, retaining only the text tags P, div, and span, while keeping all whitespace information after tag removal, marked as line blocks. Using the line number in a line block as the x-axis, take the surrounding k lines (context is acceptable, with a threshold k>3, downward direction, k being the line block thickness), collectively called a line block. Line block i is the line block with line number i as the axis. Define the line block length: the total number of characters after removing all whitespace characters (\n, \r, \t, etc.) from a line block is called the length of the line block. Define the line block distribution function: with each line as the axis, there are LinesNum(xtext)-K line blocks, creating a distribution function with [1, LinesNum(xtext-K)] as the horizontal axis and the length of each line block as the vertical axis. Using a DOM tree structure, a clustering algorithm is used to analyze the paragraph link density and text density of the main text paragraphs, and word segmentation is performed. Jieba is used to segment the text, and then the word count is calculated. A recursive feedback neural network model is used to identify text density and link density on the same news websites to obtain the most effective main text content. Text density can be calculated as the total number of words per line / the number of lines; link density can be calculated as the number of words in the link text / the total number of words. Irrelevant content is replaced with main text filtering features on the obtained main text content. Effective main text is then obtained. The extraction status of the article's main text is then marked.

[0093] The time is extracted from the Uniform Resource Locator of each reference information by regular expression to obtain the initial time feature. The initial time is then formatted to obtain the information publication time.

[0094] In specific implementation, regular expressions are used to extract the time from the time webpage in the Information Uniform Resource Locator (URL) based on the publication time feature library.

[0095] If there are no year or month characteristics for the time, and the main formats are: X days ago, X minutes ago, X hours ago, X seconds ago, etc., use the following regular expression for cleaning:

[0096] (r'\d{1,2}%s\d{1,2}%s\d{1,2}%s','%%H%s%%M%s%%S%s')

[0097] (r'\d{1,2}%s\d{1,2}%s','%%H%s%%M%s')

[0098] For dates without year characteristics, primarily expressed as month and day, the following regular expression is used for cleaning:

[0099] (r'\d{1,2}%s\d{1,2}%s\d{1,2}%s\d{1,2}%s\d{1,2}%s','%%m%s%%d%s%%H%s%%M%s%%S%s')

[0100] (r'\d{1,2}%s\d{1,2}%s\d{1,2}%s\d{1,2}%s','%%m%s%%d%s%%H%s%%M%s')

[0101] (r'\d{1,2}%s\d{1,2}%s','%%m%s%%d%s')

[0102] Other non-standard time features are cleaned using the following regular expression:

[0103] (r'\d{4}%s\d{1,2}%s\d{1,2}%s\d{1,2}%s\d{1,2}%s\d{1,2}%s','%%Y%s%%m%s%%d%s%%H%s%%M%s%%S%s')

[0104] (r'\d{4}%s\d{1,2}%s\d{1,2}%sT\d{1,2}%s\d{1,2}%s\d{1,2}%s','%%Y%s%%m%s%%d%s%%H%s%%M%s%%S%s')

[0105] (r'\d{4}%s\d{1,2}%s\d{1,2}%s\d{1,2}%s\d{1,2}%s','%%Y%s%%m%s%%d%s%%H%s%%M%s')

[0106] (r'\d{4}%s\d{1,2}%s\d{1,2}%s','%%Y%s%%m%s%%d%s')

[0107] (r'\d{2}%s\d{1,2}%s\d{1,2}%s','%%y%s%%m%s%%d%s')

[0108] Finally, mark the article's publication time and extraction status.

[0109] The author of the information is obtained by matching the author features in the preset multi-mode channel feature configuration library with the reference information. Specifically, the author information can be extracted from the configured author feature library. If the author information is not obtained, the information platform name configured in step one is used and the author feature is marked. The status of the article information author is marked.

[0110] It should be further explained that after extracting features from the multiple reference information to obtain reference features for each reference information, the process further includes: matching the information text and information title of each reference information with the monitoring tag feature set in the preset multi-mode channel feature configuration library, and determining whether the reference information is relevant based on the matching result; matching the information text and information title of each reference information with the sensitive word tag feature set in the preset multi-mode channel feature configuration library, and determining whether the reference information is invalid based on the matching result; extracting text keywords from the information text and information title of the reference information to obtain text keywords, and determining whether the reference information meets the quality requirements based on the text keywords; and removing irrelevant reference information, invalid reference information, and reference information that does not meet the quality requirements from the reference information to obtain multiple reference information.

[0111] Understandably, when searching for reference information, there may be content that does not meet the current needs or contains sensitive words that cannot be used. In such cases, it is necessary to filter the reference information by looking at the main text and tags.

[0112] In practice, the news text / title are repeatedly matched against a vocabulary list based on the monitoring tag feature set to determine relevance. Then, the news text / title are used to extract keywords using the Term Frequency-Inverse Word Frequency (TF-IWF) algorithm based on the monitoring tag feature set to assess news quality. Finally, the news text / title are repeatedly filtered out using a sensitive word filtering algorithm based on the sensitive word tag feature set from step one. After all these operations are completed, the valid news articles are retained.

[0113] Step S30: Perform feature expansion on the reference features to obtain the information briefing dataset.

[0114] Understandably, feature expansion of each reference can be considered because the title, body, author, and time of the reference are very fixed, but the features in the title are extended, that is, the text has many synonyms or similar descriptions. Through expansion, a more comprehensive information summary dataset can be obtained.

[0115] It should be noted that the news briefing dataset includes a lot of feature information, such as keywords, titles, body text, tags, keyword phrases, etc., but not every feature information is useful and effective. Therefore, the news briefing dataset needs to be cleaned and integrated.

[0116] In practice, a news briefing dataset is created, with the complete news title as the unique identifier key. The news title, news body, news author, and news time are converted into corresponding IDs as values ​​that correspond to the news title key and are saved in the news briefing dataset. It can be understood that each key has a corresponding value.

[0117] Step S40: Process the information briefing dataset based on a preset feedback neural network to obtain a header, core, and tail, and generate a briefing based on the header, core, and tail.

[0118] Understandably, the preset feedback neural network can be a recursive feedback neural network, which can vectorize features.

[0119] It should be noted that data processing can be achieved by vectorizing the word-segmented newsletter features of the newsletter dataset using a recursive feedback neural network. The vectorized newsletter dataset Ri is used as the input parameter, and the target feature value Rj in the newsletter dataset is used as the reference object. Various features Ci are set for the recursive feedback neural network feature learning, and the target value is calculated as the output parameter Cj. Initially, the following features are used for feature target learning.

[0120] In practical implementation, feature target learning can refer to the following steps:

[0121] Hopfield feedback neural network article title feature Ci1:

[0122] Titles are generally marked as HTML DOM block tags: <h> ”、" ”、" "wait.

[0123] The title is usually located in the HTML DOMTitle position.

[0124] Titles are generally located in the HTML DOM. <h1>-< / h1> <h3>”、" ”、" ”、" <strong>" and other tags

[0125] The title text should generally be longer than 10 characters and shorter than 35 characters.

[0126] Hopfield feedback neural network article text features Ci2:

[0127] The main body of an article is typically marked with the HTML DOM block tag: " ”、" "

[0128] HTML web pages typically contain " "" Marked with tags such as "" and having more than two paragraphs.

[0129] The main body of the article generally contains many sentences in the HTML DOM and has many punctuation marks such as "." and "," (>5).

[0130] Based on the tags obtained by the known algorithm, the tag density of the article body should be between 1% and 5% = 1000 * number of tags / number of words.

[0131] The main body of the article should occupy a relatively large portion, with a text density = len(main text markup block) / len(total webpage code) greater than 30%.

[0132] The main text of the article should not contain "Previous Article / Page" or "Next Article / Page", so these should be excluded.

[0133] The main body of the article should not contain sensitive words such as "related links", "related news", or "related reports", and should not contain too many hyperlinks. Paragraphs with more than 5 links should be considered recommended related content and should be excluded.

[0134] The main text of the article should not contain copyright notices or similar statements; these need to be removed due to weight reduction.

[0135] Common locations for the main text are as follows: Normal main text is below the title; Normal main text is below the publication date block; Normal main text is above the relevant link block.

[0136] Hopfield feedback neural network publication time feature Ci3:

[0137] The publication time is generally marked as "" in the HTML DOM block. ”、" ”、" ”、" ".

[0138] Publication time characteristics are generally time format or timestamp format, and the character length conforms to the time format.

[0139] The content is suspected to contain the following keywords: source, publication, time, etc.

[0140] Hopfield feedback neural network publishes author features Ci4:

[0141] The author's HTML DOM block tag is generally: " ”、" ”、" ".

[0142] The authors of published articles typically include commonly used keywords such as "author," "source," and "infor."

[0143] It should be emphasized that the length of the text published by the author is generally greater than 2 characters and less than 10 characters.

[0144] Assume that step three obtains a total of N valid information entries, ω ij S represents the join weight from article j to article i obtained from the existing rule base. j V represents the j-th state (+1 or -1) at which the neural network acquires the article content. j This represents the net input of neuron j, when V j When (t) = 0, the state of the neuron can be considered to remain unchanged.

[0145] The state of the entire network can be represented by a column vector. express:

[0146]

[0147]

[0148] The recursive feedback neural network algorithm is used to extract relevant features from the full dataset and scan the entire model to transform the resulting model into a training vector model.

[0149] Conventional feature vectors:

[0150] ij(t)=[R1(t),R2(t),…,R i (t)]

[0151] Neural network feature training vectors:

[0152] ji(t)=[C1(t),C2(t),…,C i (t)]

[0153] The features of the test set are matched using a confusion matrix to obtain state variables, and the matrix is ​​expanded to a length of ij*ji=N. 2 The sequence is input into the neural network. That is:

[0154] ω ij =ω ji ,ω ij =0

[0155] Therefore, the connection weight matrix W of the network is an N×N symmetric matrix with zeros on the diagonal.

[0156] The calculation is performed randomly in a serial manner, if the network starts from the initial state at time t=0. Initially, there exists a finite time t such that the state of the network does not change after this time, that is:

[0157]

[0158] Because of s i s j It can only be +1 or -1, while ω ij and θ i Since everything is bounded, energy is also bounded, that is:

[0159]

[0160] When ΔE = 0 Then we have:

[0161] (1) If This will result in completely consistent data.

[0162] (2) If This yields fuzzy consistent data.

[0163] (3) If This will result in completely mismatched data.

[0164] The output neuron is calculated by taking values ​​of 0, -1, and 1, which correspond to "fuzzy consistency", "complete mismatch", and "complete consistency" respectively.

[0165] The evaluation determines whether the output value generated by the recursive feedback neural network algorithm is "completely consistent", "fuzzy consistent", or "cannot match" with the features in the information briefing dataset.

[0166] Furthermore, a duplicate penalty mechanism can be set to prevent subsequent information from being obtained ineffectively. If the data is completely identical, the feature corresponding to Ri is obtained; if it is fuzzy identical, the Ri feature is partially accepted, and the data is optimized using automated parameter tuning.

[0167] Understandably, the news briefing dataset can be randomly split into training and testing sets in a 7:3 ratio after the briefings are generated. By generating the training set multiple times and training iteratively, quantitative models can be trained, including multi-modal channel configuration feature sets, tag feature sets, news title feature sets, news text feature sets, publication time feature sets, and publication author feature sets. In practical implementation, the complete logic for briefing generation can be found by referring to... Figure 3 .

[0168] Understandably, by using the information briefing dataset Ri generated according to prior feature extraction rules as input parameters, matching feature Rj as reference object, and using various features Ci learned by recursive feedback neural network features, the target value is learned as the output parameter Cj. The target values ​​Rj and Cj of Ri and Ci features are compared to see if they are consistent. If they are consistent, the Cj feature is fed back into the Rj configuration library.

[0169] This embodiment provides a novel adaptive information collection method based on a general configuration library and tag library. By configuring tags that users are interested in, and using an effective keyword configuration feature library and a sensitive word configuration library, it obtains effective web pages / content, shortens the information transmission time of scientific research briefings, improves the efficiency of briefing work, and effectively avoids the problem of a large amount of time wasted due to the need for manual information screening and writing of briefing information.

[0170] refer to Figure 4 , Figure 4 This is a flowchart illustrating a second embodiment of a briefing generation method according to the present invention.

[0171] Based on the first embodiment described above, the briefing generation method of this embodiment includes the following in step S30:

[0172] Step S31: Use the title of each reference news item as a news identifier, and convert the news title, news text, news author, and news time into news IDs.

[0173] Understandably, the information ID can be the number of the referenced information.

[0174] Step S32: Use the information title configuration features, information body configuration features, information publication time configuration features, and information author configuration that are successfully matched with the information feature set in the preset multi-mode channel feature configuration library as feature target values.

[0175] Understandably, the preset multi-mode channel feature configuration library contains a feature library. The successfully matched features are used as the feature target values, and the target feature values ​​are also used as part of the features of the reference information, which can enrich the features in the information briefing dataset.

[0176] In specific implementation, the information title configuration features, information body configuration features, information publication time configuration features, and information author configuration features obtained through page feature parsing are used as feature target values ​​and as supplementary values, that is, as newly added content. These correspond to the unique identifier key information title that already exists in the obtained information brief dataset, thus expanding the information brief dataset.

[0177] Furthermore, all data in the news briefing dataset are normalized. The features in the news briefing dataset are then converted into numerical feature matrices, invalid / redundant features are eliminated, and useful features are selected to serve as training data for the model. This allows for the training of a recursive feedback neural network, enabling automatic expansion of features from a pre-defined multi-mode channel feature configuration library.

[0178] Step S33: Generate an information directory based on the information identifier, the feature target value, and the information ID, and construct an information briefing dataset based on multiple information directories.

[0179] It should be noted that the step of generating an information directory based on the information identifier, the feature target value, and the information ID, and constructing an information briefing dataset based on multiple information directories, includes:

[0180] A news catalog is generated based on the news identifier, the feature target value, and the news ID;

[0181] Multiple keywords are determined based on the article text and article title of the reference information, and article tags are generated based on the keywords;

[0182] In the specific implementation, the text and title portions of the news briefing dataset obtained in step five are combined. The TextRank algorithm is used to determine 5-10 keywords that can describe the stable meaning of the news, and corresponding news tags are generated—that is, scattered phrases that help understand the news content. These news tags, obtained through the above method, are used as supplementary generated values, corresponding to the unique identifier key (news title) already existing in the news briefing dataset obtained in step five, thus expanding the news briefing dataset.

[0183] Based on the text and title of the reference information, adjacent word groups are constructed, and keyword phrases are obtained by combining the adjacent word groups.

[0184] In practice, the main text and titles of the news briefing dataset from step five are combined, and algorithms are used to construct adjacent word groups that can describe the news, which are then combined into keyword phrases. Unlike step six, this step focuses on extracting phrases of a certain length, distinguishing them from scattered word groups.

[0185] The keyword phrases obtained through the above method are used as supplementary generated values, corresponding to the unique identifier key information title already existing in the information briefing dataset obtained in step five, thus expanding the information briefing dataset.

[0186] The information directory is updated based on the information tags and keyword phrases to obtain a reference directory, and an information briefing dataset is constructed based on the reference information directory.

[0187] It should be further explained that the step of updating the information directory based on the information tags and the keyword phrases to obtain a reference directory, and constructing an information briefing dataset based on the reference information directory, also includes:

[0188] The information directory is updated based on the information tags and keyword phrases to obtain a reference directory; an article summary is generated based on the information text of the reference information; the reference information directory is updated based on the article summary, and an information summary dataset is constructed based on the updated reference information directory.

[0189] In practice, the main body of the news summary dataset from step five is used to generate corresponding article summaries using the GPT-2 model. These article summaries are then used as supplementary values, corresponding to the unique identifier key (news title) already existing in the news summary dataset obtained in step five, to expand the news summary dataset.

[0190] This embodiment utilizes a method for generating research briefings based on a feedback neural network with fused multi-feature labels. It combines large-scale pre-training with diverse and rich knowledge, employing continuous learning techniques to constantly absorb new knowledge in vocabulary, structure, and semantics from massive amounts of text data, enabling the model to continuously evolve. In actual training, this model can autonomously learn and improve itself without altering any training data, rapidly enhancing model performance. Furthermore, by expanding the features of the information briefing dataset, it can match the content of the reference information, facilitating the training of subsequent models and enabling faster and more accurate identification of the corresponding content in the briefings.

[0191] refer to Figure 5 , Figure 5 This is a flowchart illustrating a second embodiment of a briefing generation method according to the present invention.

[0192] Based on the first embodiment described above, the briefing generation method of this embodiment includes the following in step S40:

[0193] Step S41: Based on the preset feedback neural network, generate a header by configuring feature tags for the information in the information briefing dataset and configuring the scientific research information unit.

[0194] It should be noted that the collected information texts are categorized, and users can find the information they need based on the tagging algorithm.

[0195] The categorized information is reassembled, and users can simply click the "Group" tag to see the newly assembled document, making it easy to read articles and obtain information quickly.

[0196] Add the research information documents you need to generate to the resource library, select a briefing template, and use the docx algorithm to generate a Word document research briefing with one click. It will also automatically generate news titles and content based on the collected information and title algorithm according to the user's set timer, using tags of interest.

[0197] The briefing mainly includes a header, main body, table of contents, information body, and footer. The default font is a black SimSun 12pt design.

[0198] In practice, the header generation rules are as follows:

[0199] Registration Generation: Generate a brief registration form based on the configuration of feature tags for research information.

[0200] Number generation: Generate numbers based on the starting date, week, and month.

[0201] Date generation: Generate a date based on the current information date.

[0202] Unit: Generated based on user-configured research information units.

[0203] The generated briefing header example is as follows: Zijin Academy - ICT Scientific Research Blockchain Information - Issue 1 - August 4, 2022

[0204] The header uses a red, 14-point Song typeface. A separator is used to separate the header and the body of the report.

[0205] Step S42: Generate a report based on the scientific research information configuration, article summaries, keywords, and information directory in the information summary dataset.

[0206] It should be noted that the step of generating a report based on the corresponding scientific research information configuration in the information summary dataset, the article summary, keywords, and information directory in the information summary dataset includes:

[0207] Feature extraction is performed based on the corresponding scientific research information configuration in the information briefing dataset to obtain an effective directory; multiple briefing notes are generated based on the reference information in the information briefing dataset, and the multiple briefing notes are combined to obtain a target note; word frequency statistics are performed on the keywords in the information briefing dataset, and multiple keywords are clustered to obtain word groups, and the core keyword information of the briefing is obtained based on the word groups; article titles are generated based on the keyword phrases in the information briefing dataset, and article introductions are generated based on the article briefings in the information briefing dataset; article source, article author information, and author information are obtained based on the information briefing dataset; a report core is generated based on the effective directory, the target note, the core keyword information of the briefing, the article introduction, the article source, the article author information, and the author information.

[0208] In practice, the reporting and verification generation rules are as follows:

[0209] Table of Contents Generation: Based on the feature tags configured for scientific research information, the effective information table of contents is extracted iteratively from step five. Effective table of contents generation method:

[0210] Note: Based on the trained model, all the valid information generated from the catalog will be used to generate multiple briefing notes using the article briefing generation method in step eight. Based on the trained model, multiple briefing notes will be generated and combined to generate a single note according to step eight.

[0211] Keyword clusters: By extracting keywords from news articles in the information directory and performing word frequency statistics, multiple keywords are clustered to form keyword clusters. These clusters are used to describe the core keyword information of the current briefing.

[0212] News article:

[0213] Title: Iteratively extract the effective information directory from step five, and generate the article title based on the trained model using the keyword phrases from step seven.

[0214] Introduction: Iteratively extract the effective information directory from step five, and generate the article introduction for step eight based on the trained model.

[0215] Article source: Extract the effective information directory in step five of the loop, and fill in the article source, article author information, and author information extracted in step three: page feature analysis based on the trained model.

[0216] The newspaper's closing font is a red, imitation Song typeface, 12 points. A separator is used to separate the front and back sections.

[0217] Step S43: Generate a tail based on the corresponding configuration of scientific research information occurrence range in the information briefing dataset.

[0218] Step S44: Generate a briefing based on the header, core, and footer.

[0219] It's worth noting that the newsletter push function allows users to set up information delivery departments on the page, with news items distributed daily, weekly, or monthly. The system can then schedule the delivery of the latest, most relevant news to specific users, facilitating quick and easy access to the latest information.

[0220] It should be noted that the function can quickly generate routine briefings (short reports, daily reports, weekly reports, monthly reports, etc.) and special briefings (using a large amount of special information as material and intelligence analysis methods as a means to visualize the overview, development, impact, and effects of the special field). It can also quickly generate special briefings from large-scale data, charts, animations and other multimedia data through the briefing production function.

[0221] This embodiment automatically generates information and reports by collecting dynamics of technological innovation from internal and external companies and institutions in the industry through public channels. At the same time, it can import information and data obtained from public and internal channels to realize benchmark comparative analysis in the field of technological innovation. The results can be presented in the form of tables and graphs, reducing the work of information collection and writing of reports and avoiding the waste of a lot of time caused by manual information screening and writing.

[0222] Reference Figure 6 , Figure 6 This is a structural block diagram of the first embodiment of the present invention's briefing generation device.

[0223] like Figure 6 As shown, the briefing generation device proposed in this embodiment of the invention includes:

[0224] The information acquisition module is used to filter multiple reference information from the entire network of information content based on a preset multi-mode channel feature configuration library;

[0225] The information acquisition module is also used to extract features from the plurality of reference information to obtain reference features for each reference information, wherein the reference features include information title, information text, information author, and information publication time;

[0226] The briefing generation module is used to perform feature expansion on the reference features to obtain an information briefing dataset.

[0227] The briefing generation module is also used to process the information briefing dataset based on a preset feedback neural network to obtain a header, a core, and a footer, and to generate a briefing based on the header, core, and footer.

[0228] This embodiment provides a novel adaptive information collection method based on a general configuration library and tag library. By configuring tags that users are interested in, and using an effective keyword configuration feature library and a sensitive word configuration library, it obtains effective web pages / content, shortens the information transmission time of scientific research briefings, improves the efficiency of briefing work, and effectively avoids the problem of a large amount of time wasted due to the need for manual information screening and writing of briefing information.

[0229] In one embodiment, the information acquisition module 10 is further configured to filter out multiple reference information from the information content of the entire network according to a preset multi-mode channel feature configuration library;

[0230] Feature extraction is performed on the multiple reference information to obtain reference features for each reference information, and the reference features include information title, information text, information author, and information publication time;

[0231] The reference features are expanded to obtain the information briefing dataset;

[0232] The information briefing dataset is processed based on a preset feedback neural network to obtain a header, core, and tail, and a briefing is generated based on the header, core, and tail.

[0233] In one embodiment, the information acquisition module 10 is further configured to search for web page content based on the monitoring information channel feature set in the preset multi-mode channel feature configuration library;

[0234] The webpage content is extracted to obtain initial content features;

[0235] The initial content features are matched with the information feature set in the preset multi-mode channel feature configuration library according to the preset feedback neural network;

[0236] The successfully matched webpage content is learned through a preset feedback neural network to obtain new features, and the information feature set in the preset multi-mode channel feature configuration library is updated according to the new features;

[0237] Use the content of the successfully matched web pages as reference information.

[0238] In one embodiment, the information acquisition module 10 is further configured to calculate the title weight of each reference information based on the information title features in the preset multi-mode channel feature configuration library, and perform feature extraction based on the title weight;

[0239] Calculate the paragraph link density and text density of the main text of each reference information, and extract features based on the paragraph link density and text density to obtain the main text of the information;

[0240] The time is extracted from the Uniform Resource Locator of each reference information by regular expression to obtain the initial time feature. The initial time is then formatted to obtain the information publication time.

[0241] The author of the information is obtained by matching the author characteristics of the publisher in the preset multi-mode channel feature configuration library with the reference information.

[0242] In one embodiment, the information acquisition module 10 is further configured to match the information text and information title of each reference information with the monitoring tag feature set in the preset multi-mode channel feature configuration library, and determine whether the reference information is relevant based on the matching result;

[0243] The body text and title of each reference information are matched with the sensitive word tag feature set in the preset multi-mode channel feature configuration library, and the reference information is determined as invalid based on the matching result.

[0244] Text keywords are extracted from the main text and title of the reference information to obtain text keywords, and the quality requirements of the reference information are determined based on the text keywords.

[0245] By removing irrelevant, invalid, and substandard reference information from the reference information, multiple reference information is obtained.

[0246] In one embodiment, the briefing generation module 20 is further configured to use the information title of each reference information as an information identifier, and convert the information title, information text, information author and information time into information ID;

[0247] The information title configuration feature, information body configuration feature, information publication time configuration feature, and information author configuration that successfully match each reference information with the information feature set in the preset multi-mode channel feature configuration library are used as the feature target values;

[0248] An information directory is generated based on the information identifier, the feature target value, and the information ID. Multiple information directories are used to construct an information briefing dataset.

[0249] In one embodiment, the briefing generation module 20 is further configured to generate an information catalog based on the information identifier, the feature target value, and the information ID;

[0250] Multiple keywords are determined based on the article text and article title of the reference information, and article tags are generated based on the keywords;

[0251] Based on the text and title of the reference information, adjacent word groups are constructed, and keyword phrases are obtained by combining the adjacent word groups.

[0252] The information directory is updated based on the information tags and keyword phrases to obtain a reference directory, and an information briefing dataset is constructed based on the reference information directory.

[0253] In one embodiment, the briefing generation module 20 is further configured to update the information directory based on the information tags and the keyword phrases to obtain a reference directory;

[0254] An article summary is generated based on the main text of the referenced information;

[0255] The reference information directory is updated based on the article summary, and the updated reference information directory forms the information summary dataset.

[0256] In one embodiment, the briefing generation module 20 is further configured to generate a header based on the information configuration feature tags and the configuration of scientific research information units in the information briefing dataset using a preset feedback neural network.

[0257] A report is generated based on the corresponding scientific research information configuration in the information briefing dataset, the article briefings, keywords, and information catalog in the information briefing dataset;

[0258] Generate a report tail based on the corresponding configured scientific research information occurrence range in the information briefing dataset;

[0259] A briefing is generated based on the header, core, and footer.

[0260] In one embodiment, the briefing generation module 20 is further configured to extract features based on the corresponding scientific research information configuration in the information briefing dataset to obtain an effective directory;

[0261] Multiple briefing captions are generated based on the reference information in the information briefing dataset, and the multiple briefings are combined to obtain the target caption.

[0262] By performing word frequency statistics on the keywords in the information briefing dataset, multiple keywords are clustered to obtain word groups, and the core keyword information of the briefing is obtained based on the word groups.

[0263] Generate article titles based on keywords and phrases in the information briefing dataset, and generate article introductions based on article briefings in the information briefing dataset;

[0264] The article source, author information, and author information are obtained from the aforementioned news briefing dataset;

[0265] A report is generated based on the effective directory, the target annotation, the core keyword information of the briefing, the article introduction, the article source, the article author information, and the author information.

[0266] It should be understood that the above are merely illustrative examples and do not constitute any limitation on the technical solutions of the present invention. In specific applications, those skilled in the art can make settings as needed, and the present invention does not impose any restrictions on this.

[0267] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of this invention. In practical applications, those skilled in the art can select some or all of the workflow to achieve the purpose of this embodiment according to actual needs, and no restrictions are imposed here.

[0268] Furthermore, it should be noted that in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system 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 system. Unless otherwise specified, an element defined by the phrase "comprising one…" does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0269] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0270] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory (ROM) / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0271] It should be understood that although the steps in the flowcharts of this application's embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some of the steps in the figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0272] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention. < / strong> < / h3> < / h>

Claims

1. A briefing generation method characterized by comprising: The briefing generation method includes: Multiple reference articles are selected from the entire network's information content based on a pre-defined multi-mode channel feature configuration library; The step of filtering multiple reference information from the entire network information content according to the preset multi-mode channel feature configuration library includes: The webpage content is obtained by searching the monitoring information channel feature set in the preset multi-mode channel feature configuration library; The webpage content is extracted to obtain initial content features; The initial content features are matched with the information feature set in the preset multi-mode channel feature configuration library according to the preset feedback neural network; The successfully matched webpage content is learned through a preset feedback neural network to obtain new features, and the information feature set in the preset multi-mode channel feature configuration library is updated according to the new features; Use the content of the successfully matched web pages as reference information; Feature extraction is performed on the multiple reference information to obtain reference features for each reference information, and the reference features include information title, information text, information author, and information publication time; The reference features are expanded to obtain the information briefing dataset; The information briefing dataset is processed based on a preset feedback neural network to obtain a header, core, and tail, and a briefing is generated based on the header, core, and tail. The process of processing the information briefing dataset based on a preset feedback neural network to obtain a header, core, and footer, and generating a briefing based on the header, core, and footer, includes: Based on a preset feedback neural network, a header is generated by configuring feature tags for the information in the information briefing dataset and configuring scientific research information units. A report is generated based on the corresponding scientific research information configuration in the information briefing dataset, the article briefings, keywords, and information catalog in the information briefing dataset; Generate a report tail based on the corresponding configured scientific research information occurrence range in the information briefing dataset; A briefing is generated based on the header, core, and footer.

2. The briefing generation method of claim 1, wherein, The process of extracting features from the multiple reference information items yields reference features for each item. These reference features include the information title, information text, information author, and information publication time. The title weight of each reference news item is calculated based on the news title features in the preset multi-mode channel feature configuration library, and features are extracted based on the title weights. Calculate the paragraph link density and text density of the main text of each reference information, and extract features based on the paragraph link density and text density to obtain the main text of the information; The time is extracted from the Uniform Resource Locator of each reference information by regular expression to obtain the initial time feature. The initial time is then formatted to obtain the information publication time. The author of the information is obtained by matching the author characteristics of the publisher in the preset multi-mode channel feature configuration library with the reference information.

3. The briefing generation method of claim 1, wherein, After extracting features from the plurality of reference information to obtain reference features for each reference information, the method further includes: The information body and title of each reference information are matched with the monitoring tag feature set in the preset multi-mode channel feature configuration library, and the relevance of the reference information is determined based on the matching result. The body text and title of each reference information are matched with the sensitive word tag feature set in the preset multi-mode channel feature configuration library, and the reference information is determined as invalid based on the matching result. Text keywords are extracted from the main text and title of the reference information to obtain text keywords, and the quality requirements of the reference information are determined based on the text keywords. By removing irrelevant, invalid, and substandard reference information from the reference information, multiple reference information is obtained.

4. The briefing generation method of claim 1, wherein, The feature expansion of the reference features to obtain the information briefing dataset includes: Use the title of each reference article as an article identifier, and convert the article title, article text, article author, and article time into an article ID; The information title configuration feature, information body configuration feature, information publication time configuration feature, and information author configuration that successfully match each reference information with the information feature set in the preset multi-mode channel feature configuration library are used as the feature target values; An information directory is generated based on the information identifier, the feature target value, and the information ID. Multiple information directories are used to construct an information briefing dataset.

5. The briefing generation method of claim 4, wherein, The step of generating an information directory based on the information identifier, the feature target value, and the information ID, and constructing an information briefing dataset based on multiple information directories, includes: A news catalog is generated based on the news identifier, the feature target value, and the news ID; Multiple keywords are determined based on the article text and article title of the reference information, and article tags are generated based on the keywords; Based on the text and title of the reference information, adjacent word groups are constructed, and keyword phrases are obtained by combining the adjacent word groups. The information directory is updated based on the information tags and keyword phrases to obtain a reference directory, and an information briefing dataset is constructed based on the reference information directory.

6. The briefing generation method of claim 5, wherein, The step of updating the information directory based on the information tags and keyword phrases to obtain a reference directory, and constructing an information briefing dataset based on the reference information directory, further includes: The information directory is updated based on the information tags and keyword phrases to obtain a reference directory; An article summary is generated based on the main text of the referenced information; The reference information directory is updated based on the article summary, and the updated reference information directory forms the information summary dataset.

7. The brief generation method of claim 1, wherein, The step of generating a report based on the corresponding scientific research information configuration in the information summary dataset, the article summary, keywords, and information directory in the information summary dataset includes: Feature extraction is performed based on the corresponding scientific research information configuration in the aforementioned information briefing dataset to obtain an effective directory; Multiple briefing captions are generated based on the reference information in the information briefing dataset, and the multiple briefings are combined to obtain the target caption. By performing word frequency statistics on the keywords in the information briefing dataset, multiple keywords are clustered to obtain word groups, and the core keyword information of the briefing is obtained based on the word groups. Generate article titles based on keywords and phrases in the information briefing dataset, and generate article introductions based on article briefings in the information briefing dataset; The article source, author information, and author information are obtained from the aforementioned news briefing dataset; A report is generated based on the effective directory, the target annotation, the core keyword information of the briefing, the article introduction, the article source, the article author information, and the author information.

8. A briefing generation apparatus characterized by comprising: The briefing generation device includes: The information acquisition module is used to filter multiple reference information from the entire network of information content based on a preset multi-mode channel feature configuration library; The information acquisition module is further configured to: search for web page content based on the monitoring information channel feature set in the preset multi-mode channel feature configuration library; extract content from the web page content to obtain initial content features; match the initial content features with the information feature set in the preset multi-mode channel feature configuration library based on a preset feedback neural network; learn new features from the unmatched web page content through the preset feedback neural network; update the information feature set in the preset multi-mode channel feature configuration library based on the new features; and use the successfully matched web page content as reference information. The information acquisition module is also used to extract features from the plurality of reference information to obtain reference features for each reference information, wherein the reference features include information title, information text, information author, and information publication time; The briefing generation module is used to perform feature expansion on the reference features to obtain an information briefing dataset. The briefing generation module is also used to process the information briefing dataset based on a preset feedback neural network to obtain a header, a core, and a tail, and to generate a briefing based on the header, core, and tail. The briefing generation module is further configured to generate a header based on the information configuration feature tags and configured scientific research information units corresponding to the information briefing dataset using a preset feedback neural network; generate a report core based on the scientific research information configuration, article briefings, keywords, and information catalog in the information briefing dataset; generate a report tail based on the configured scientific research information occurrence range in the information briefing dataset; and generate a briefing based on the header, report core, and report tail.

Images