A news information aggregation method

Abstract

The application discloses a news information aggregation method, which comprises an aggregation recommendation system; the method comprises the following steps: S1, data crawling is performed on news and comments of a specified website platform through the aggregation recommendation system, data preprocessing is performed, and information data of a user to be pushed is acquired; S2, the news is classified, news data features are extracted, classification labels of the news are generated, and click data of the classification labels is acquired; S3, the acquired news data is split, and different types of keywords are acquired. The collected news information data can be subjected to feature analysis, and matching is performed on the labels and user data to be recommended, multiple different types of news information are aggregated and pushed, and more news information that is interesting to users and is consistent with the social activity circle of users is found by combining comment data, related publisher data and user intention data, and multiple types of information are aggregated.

Description

Technical Field

[0001] This invention relates to the field of news aggregation technology, and in particular to a method for news information aggregation. Background Technology

[0002] News is an important medium for conveying information, reflecting society, and guiding public opinion. With the development of technologies such as the Internet and big data, the methods and forms of news dissemination have also changed, giving rise to data journalism as a new form of news.

[0003] News aggregation refers to the collection of news and messages from various media and websites on the Internet (such as Toutiao, People's Daily Online, Xinhua News Agency, etc.), which are then filtered and presented to end users in a certain way. Existing examples of news aggregation include major news apps such as Toutiao, Yidian Zixun, Tencent News, and NetEase News. They mainly recommend news to users through channel or column subscriptions. The news disseminated by each platform is different, the data sources are not rich enough, and the news recommended to users is often too biased towards the content of the subscribed columns, which are relatively fixed news information.

[0004] A search revealed that Chinese patent CN111881277A discloses a multi-dimensional, highly customizable news aggregation method. It helps users define their own following behavior by extracting keyword text as search fields. However, it is similar to the column subscription method, and the types of news followed tend to be relatively fixed after long-term use. Chinese patent CN105022827A discloses a method for dynamic aggregation of web news based on a specific domain theme. This method collects web pages from multiple general search engines to obtain diverse news information and improve user experience.

[0005] However, for news aggregation, how to classify the aggregated information, how to combine users' own behavioral data with content they are interested in, and how to further aggregate and recommend content based on this, so as to improve the diversity of aggregated news types, adapt to users' personalization, and expand users' interest content, are urgent problems to be solved. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing technologies by proposing a news information aggregation method.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A news information aggregation method, the method including an aggregation recommendation system; The method includes the following steps: S1. Use an aggregation recommendation system to crawl news and comments from designated website platforms, preprocess the data, and obtain information data of users who need to be pushed to them. S2. Classify the news, extract the features of the news data, generate the category tags for the news, and obtain the click data of the category tags. S3. Segment the acquired news data to obtain different types of keywords; S4. Obtain user operation behavior data and browsing data within a time period, match multiple recommended news articles, generate multiple recommendation lists based on probability, and push the news data of the most matching recommendation list to the current user's terminal. S5. Optimize the display of recommended news by aggregating and displaying news from the news list.

[0008] Furthermore, in step S3, the keyword types are divided into text, numbers, and English. A coherent string of text, numbers, or English is treated as a complete sentence and semantically matched. If the semantics are fluent, it is used as a category label. At the same time, different types of keywords are combined and analyzed. If they can form a coherent sentence, the coherent sentence is used as a category label. New news data is then matched again based on the new semantics.

[0009] Furthermore, aggregated recommendation systems include: The data acquisition module is used to crawl news and comments from a specified website platform via network communication and to preprocess the data. The data classification and processing module is used to break down and classify the collected news data into audio data, video data, and text data. Among them, the audio data is converted into audio-text data through speech conversion. The key feature extraction module is used to extract features from the collected and classified news data, generate data classification labels based on the extracted features, and collect user operation data. The dwell feedback optimization module is used to collect the current account's operational behavior data, including commenting, viewing comments, liking, forwarding, saving, viewing authors, and author browsing time, and optimize the recommended news list data based on the operational behavior; The social optimization module is used to obtain a list of recommended users based on their current location and the location of their behavioral trajectory, and to make interactive recommendations. The imperfect add-on module is used to define tags that appear more than a set number of times as user popularity tags, and to recommend them in the second priority position based on these user popularity tags; The recommendation list module is used to generate at least one news recommendation list, display the news recommendation list based on the user's access data within a time period, and store and cache the remaining recommendation lists. When the user's browsing data changes, the recommendation lists are rearranged according to the category tags in the remaining recommendation lists and then recommended. The aggregation and display module is used to layout and generate news recommendation interfaces, aggregate and display news, and collect user interface operation data.

[0010] Furthermore, the data acquisition module is also used to obtain the publishing information of publishers in the news list recommended to the current user, analyze the news data published by the publishers, and classify and record the publishers as category tags. This enables the system to filter and recommend content to users based on the publisher tags they follow when making aggregated recommendations.

[0011] Furthermore, the key feature extraction module, data processing and classification module, social optimization module, and data collection module are interconnected to classify and aggregate the collected news data in conjunction with user data, generating a recommendation list suitable for user habits.

[0012] Furthermore, the imperfect add-on module, dwell feedback optimization module, social optimization module, and recommendation list module are interconnected to optimize the weight of news aggregation for users, and to analyze and optimize based on the current user's operation behavior data and user attention data in the news information.

[0013] Furthermore, the recommendation list module, data processing and classification module, aggregation display module, and key feature extraction module are interconnected to generate a user recommendation list, arrange the aggregation display, and generate an aggregation menu that can record user operation behavior data.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention can perform feature analysis on the collected news information data, and establish classification tags based on the news information data including audio and video. By matching the tags with user data that needs to be recommended, multiple different types of news information can be aggregated and pushed. It can generate multiple recommendation lists based on the user's interest tags, determine other content types that the user is more interested in among multiple pieces of information in aggregated news by using the user's active selection behavior data, mine the user's potential interest target data, and generate personalized information aggregation recommendation data; By combining aggregation methods with social attributes, and by integrating comment data, relevant publisher data, and user intention data, we can find more news information that users are interested in and that is in agreement within their social circles, and aggregate multiple types of information. Attached Figure Description

[0015] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0016] Figure 1 This is a flowchart illustrating the news information aggregation method proposed in this invention. Figure 2 This is a flowchart illustrating the aggregation recommendation system in an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0018] Reference Figure 1 News information aggregation methods, including aggregation recommendation systems; The method includes the following steps: S1. Use an aggregation recommendation system to crawl news and comments from designated website platforms, preprocess the data, and obtain information data of users who need to be pushed to them. S2. Classify the news, extract the features of the news data, generate the category tags for the news, and obtain the click data of the category tags. For each news item, the classification method is to categorize it by sentence paragraphs, sentence keywords, noun search, and content tags, and then perform content matching based on the classification to find the intersection; S3. Segment the acquired news data to obtain different types of keywords; It should be noted that the keyword types are divided into text, numbers, and English. A coherent string of text, numbers, or English is treated as a complete sentence and semantically matched. If the semantics are coherent, it is used as a category tag. At the same time, different types of keywords are combined and analyzed. If they can form a coherent sentence, the coherent sentence is used as a category tag. New news data is then matched again based on the new semantics.

[0019] S4. Obtain user operation behavior data and browsing data within a time period, match multiple recommended news articles, generate multiple recommendation lists based on probability, and push the news data of the most matching recommendation list to the current user's terminal. S5. Optimize the display of recommended news by aggregating and displaying news from the news list.

[0020] like Figure 2 As shown, in a preferred embodiment of this application, the aggregated recommendation system includes: The data acquisition module is used to crawl news and comments from a specified website platform via network communication and to preprocess the data. It should be noted that this process also collects the user's recorded operational behavior data and the determined appropriate recommendation data, which is used for personalized recommendation information for the user.

[0021] The data classification and processing module is used to break down and classify the collected news data into audio data, video data, and text data. Among them, the audio data is converted into audio-text data through speech conversion. It should be noted that by using AI video analysis, the video is analyzed frame by frame. By matching relevant search results for at least one frame, the keywords describing the scene are obtained and news headlines are automatically generated. In addition, comments under each category and trending news item will have highly popular identical comments. These comments are used to determine the specific event and category to which the news belongs. Furthermore, image recognition and image matching are performed to determine the news category.

[0022] The key feature extraction module is used to extract features from the collected and classified news data, generate data classification labels based on the extracted features, and collect user operation data. It should be noted that the data classification labels are specifically generated as tags based on the text features extracted from the news data, keywords extracted as tags after converting the audio extracted from the news data into text information, and key thumbnails extracted from the audio data matched with the video data extracted from the news data for image matching and classification.

[0023] The dwell feedback optimization module is used to collect the current account's operational behavior data, including commenting, viewing comments, liking, forwarding, saving, viewing authors, and author browsing time, and optimize the recommended news list data based on the operational behavior; The social optimization module is used to obtain a list of recommended users based on their current location and the location of their behavioral trajectory, and to make interactive recommendations. The imperfect add-on module is used to define tags that appear more than a set number of times as user popularity tags. Based on these user popularity tags, the second-ranked recommendation position is selected (i.e., another user popularity tag is selected as the ranking popularity tag, where the news with the new ranking popularity tag contains the user popularity tag). It should be noted that the user popularity tag is the information category tag with the highest repetition in the news information that the user has viewed in history. When the user receives a recommendation, the user popularity tag will be used as the first search keyword, and a note will be added under the news information when recommending it to the user. In addition, because the pre-analysis of news information and the extraction of category tags are similar, when not recommending to individual users, the system will pre-analyze the crawled news information data in the background and generate category tags. When a category tag or keyword is searched, viewed by users a lot, or clicked by users a lot, the tag will be used as the real-time user popularity tag. The real-time user popularity tag will be used as the first user popularity tag to generate a real-time popularity recommendation list for users.

[0024] It should be noted that this imperfect add-on module reduces the occurrence of the most clicked and viewed tags by users, increases the number of hidden tags that users may be interested in, thereby reducing the most interesting news information, expanding the number of news aggregation types used for recommendations, and broadening the types of information that users may be interested in.

[0025] The recommendation list module is used to generate at least one news recommendation list, display the news recommendation list based on the user's access data within a time period, and store and cache the remaining recommendation lists. When the user's browsing data changes, the recommendation lists are rearranged according to the category tags in the remaining recommendation lists and then recommended. In addition, it is used to define recommendation lists, cache the obtained recommendation lists, use thumbnails, thumbnail text, and thumbnail videos as recommendation titles, and generate a backup click menu to change access to the recommendation list. When the menu is clicked too many times, it is determined that the user prefers diversified recommendations and the recommendation list is retrieved again.

[0026] The aggregation and display module is used to layout and generate news recommendation interfaces, aggregate and display news, and collect user interface operation data.

[0027] In a specific embodiment of this application: the data acquisition module is also used to obtain the publishing information of publishers in the news list recommended to the current user, analyze the news data published by the publishers, and classify and record the publishers as category tags. This enables the filtering and recommendation based on the publisher tags followed when making aggregated recommendations to users.

[0028] In a specific embodiment of this application: the key feature extraction module, the data processing and classification module, the social optimization module, and the data collection module are interconnected to classify and aggregate the collected news data in combination with user data, and generate a recommendation list suitable for user habits.

[0029] In a specific embodiment of this application: the imperfect addition module, the dwell feedback optimization module, the social optimization module, and the recommendation list module are interconnected to optimize the weight of news aggregation for users, and to analyze and optimize based on the current user's operation behavior data and the user's attention data in the news information.

[0030] Furthermore, the user operation data obtained through the dwell feedback optimization module is used to classify news that users have stayed for more than a certain period of time or that have high recommendation scores in the recommendation list, and to collect all the news published by the news publisher, and to aggregate and recommend news with the same tags as the news within a certain period of time. In addition, it also retrieves comment information from recommended news articles and aggregates and recommends news articles with similar key text comments in a synchronized manner.

[0031] In a specific embodiment of this application: the recommendation list module, the data processing and classification module, the aggregation and display module, and the key feature extraction module are interconnected to generate a user recommendation list and arrange the aggregation and display method. They generate an aggregation menu that can record user operation behavior data, such as a search box, a dwell box, user click data, user page dwell data, and user scrolling data. This type of menu links to the corresponding operation menu and only records user operation behavior. For example, when the search box is clicked, the search command is invoked and the user is redirected to the search results interface.

[0032] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for aggregating news information, characterized in that, This method includes an aggregated recommendation system; The method includes the following steps: S1. Use an aggregation recommendation system to crawl news and comments from designated website platforms, preprocess the data, and obtain information data of users who need to be pushed to them. S2. Classify the news, extract the features of the news data, generate the category tags for the news, and obtain the click data of the category tags. S3. Segment the acquired news data to obtain different types of keywords; S4. Obtain user operation behavior data and browsing data within a time period, match multiple recommended news articles, generate multiple recommendation lists based on probability, and push the news data of the most matching recommendation list to the current user's terminal. S5. Optimize the display of recommended news by aggregating and displaying news from the news list.

2. The news information aggregation method according to claim 1, characterized in that, In step S3, the keyword types are divided into text, numbers, and English. A coherent string of text, numbers, or English is treated as a complete sentence and semantically matched. If the semantics are coherent, it is used as a category label. At the same time, different types of keywords are combined and analyzed. If they can form a coherent sentence, the coherent sentence is used as a category label. New news data is then matched again based on the new semantics.

3. The news information aggregation method according to claim 1, characterized in that, Aggregated recommendation systems include: The data acquisition module is used to crawl news and comments from a specified website platform via network communication and to preprocess the data. The data classification and processing module is used to break down and classify the collected news data into audio data, video data, and text data. Among them, the audio data is converted into audio-text data through speech conversion. The key feature extraction module is used to extract features from the collected and classified news data, generate data classification labels based on the extracted features, and collect user operation data. The dwell feedback optimization module is used to collect the current account's operational behavior data, including commenting, viewing comments, liking, forwarding, saving, viewing authors, and author browsing time, and optimize the recommended news list data based on the operational behavior; The social optimization module is used to obtain a list of recommended users based on their current location and the location of their behavioral trajectory, and to make interactive recommendations. The imperfect add-on module is used to define tags that appear more than a set number of times as user popularity tags, and to recommend them in the second priority position based on these user popularity tags; The recommendation list module is used to generate at least one news recommendation list, display the news recommendation list based on the user's access data within a time period, and store and cache the remaining recommendation lists. When the user's browsing data changes, the recommendation lists are rearranged according to the category tags in the remaining recommendation lists and then recommended. The aggregation and display module is used to layout and generate news recommendation interfaces, aggregate and display news, and collect user interface operation data.

4. The news information aggregation method according to claim 3, characterized in that, The data acquisition module is also used to obtain the publishing information of publishers in the news list recommended to the current user, analyze the news data published by the publishers, and classify and record the publishers as category tags. This enables the system to filter and recommend content to users based on the publisher tags they follow when making aggregated recommendations.

5. The news information aggregation method according to claim 4, characterized in that, The key feature extraction module, data processing and classification module, social optimization module, and data collection module are interconnected to classify and aggregate the collected news data by combining user data, and generate a recommendation list suitable for user habits.

6. The news information aggregation method according to claim 5, characterized in that, The imperfect add-on module, dwell feedback optimization module, social optimization module, and recommendation list module are interconnected to optimize the weight of news aggregation for users, and to analyze and optimize based on the current user's operation behavior data and user attention data in the news information.

7. The news information aggregation method according to claim 6, characterized in that, The recommendation list module, data processing and classification module, aggregation display module, and key feature extraction module are interconnected to generate a user recommendation list, arrange the aggregation display, and generate an aggregation menu that can record user operation behavior data.

Images