Ai model-based method for intelligent rss feed screening and automated article generation, and related device
Patent Information
- Authority / Receiving Office
- HK · HK
- Patent Type
- Patents
- Current Assignee / Owner
- LIGHTWAN CORP LTD
- Filing Date
- 2026-02-11
- Publication Date
- 2026-07-10
AI Technical Summary
Existing RSS systems cannot meet users' needs for precise, personalized, and automated content services. Traditional RSS aggregation and rule-based filtering technologies suffer from problems such as information overload, poor user experience, and insufficient semantic understanding.
It adopts an AI-based RSS feed intelligent filtering and automated article generation method. By intelligently identifying RSS feeds, dynamically grouping, structurally filtering and automatically generating articles, combined with user-defined prompts, it achieves accurate filtering and high-quality article generation.
It significantly improves the automation level and user experience of RSS content processing, lowers the threshold for manual configuration, improves the accuracy and quality of content filtering and generation, meets personalized needs, reduces labor costs, and enhances system reliability and user satisfaction.
Smart Images
Figure 00000000_0000_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of artificial intelligence technology, and in particular to a method and related apparatus for intelligent filtering and automated article generation of RSS subscription sources based on an AI model. Background Technology
[0002] With the explosive growth of internet information, RSS (Really Simple Syndication), as an efficient information subscription and content aggregation technology, has become a core means for news websites, blogs, forums, and other platforms to push updated content to users. By subscribing to RSS feeds of interest, users can obtain real-time updates from target platforms without having to visit multiple websites one by one, significantly improving information acquisition efficiency. However, as the number of RSS feeds subscribed to by users increases, traditional RSS systems have gradually exposed insurmountable technical bottlenecks, failing to meet users' needs for "precise, personalized, and automated" content services.
[0003] Existing RSS content processing solutions mainly fall into two categories, but both have significant drawbacks:
[0004] Option 1: Traditional RSS aggregation technologies (such as Feedly and Inoreader) adopt a "periodic polling + time sorting + keyword filtering" model, which can only list content in chronological order and cannot understand the semantics of the content and user intent. The filtering ability is limited to simple keyword matching. Users still need to spend a lot of time manually filtering valuable information, resulting in problems of "information overload" and "poor user experience".
[0005] Option 2: Rule-based RSS filtering technology. This technology uses regular expressions, keyword blacklists and whitelists, and metadata (content length, publication time) to build filtering rules. However, it requires users to have technical backgrounds to manually maintain the rules, and it cannot cope with changes in content format (the rules are prone to becoming invalid). It also lacks semantic understanding capabilities and is difficult to meet personalized filtering needs.
[0006] This application aims to address the technical problems of "low accuracy of RSS content filtering, insufficient automation, and lack of personalized services" in existing technologies. By deeply integrating AI models with RSS technology, it aims to upgrade from "passive aggregation" to "proactive intelligent processing". Summary of the Invention
[0007] The purpose of this application is to provide a method and related apparatus for intelligent filtering and automated article generation of RSS subscription sources based on an AI model, which can upgrade RSS from passive aggregation to active intelligent processing.
[0008] To achieve the above objectives, this application provides the following solution:
[0009] Firstly, this application provides a method for intelligent filtering and automated article generation of RSS feeds based on an AI model, including the following steps:
[0010] The system retrieves subscription task configuration information, periodically checks pending tasks and identifies delayed tasks for batch processing, while monitoring task execution status. Subscription task configuration information includes the RSS feed base URL, push cycle type, filter prompts, and generated prompts.
[0011] The system performs intelligent identification and verification on the base URL of the RSS source, obtains the latest content of the valid RSS source and preprocesses it, and converts heterogeneous RSS data into a standardized internal data structure. The intelligent identification and verification on the base URL of the RSS source includes first directly accessing the base URL of the RSS source. If the access fails, it traverses the preset common RSS path suffix library to attempt URL concatenation. The system then verifies the RSS format validity of the content returned after concatenation to determine the valid RSS source.
[0012] The time window is calculated based on the push cycle type, and the RSS content published within the time window is filtered out to remove expired RSS content.
[0013] The number of tokens for the selected RSS content is calculated, and the content is dynamically grouped according to the token processing limit of the AI screening model to ensure that a single article is not split and the total number of tokens in each group does not exceed the processing limit.
[0014] The AI filtering model is invoked using a preset JSON Schema, and the filtering prompts are combined to perform structured filtering on each group of RSS content, resulting in structured filtering results.
[0015] The filtering results of all groups are reorganized, the number of tokens is recalculated, and the groups are regrouped according to the token processing limit of the AI-generated model.
[0016] The AI-generated model is invoked, and combined with generated prompts, to automatically generate articles for each group of RSS content after regrouping.
[0017] The generated article is used to generate a matching article title using a title generation model, and the complete article data is saved to the database. At the same time, a notification is sent to the subscribing users.
[0018] Optionally, the default library of common RSS path suffixes includes " / rss", " / feed" and " / atom.xml", and the RSS format validity validation includes verifying whether the returned data contains the fields "title", "item" and "pubDate".
[0019] Optionally, the time window is calculated based on the push cycle type, specifically including: if the push cycle type is "1 day", the time window is 24 hours before the current time; if it is "7 days", the time window is 168 hours before the current time; if it is "30 days", the time window is 720 hours before the current time, and the time zone information of the system deployment is adapted synchronously during the calculation process.
[0020] Optionally, when calculating the number of tokens for the selected RSS content, the OpenAI official token calculation algorithm is used, and a greedy algorithm is used when performing dynamic grouping. The total number of tokens for each group is accumulated in real time. When the addition of new articles causes the total number to exceed the processing limit of the AI screening model, a new group is automatically created.
[0021] Optionally, the AI filtering model can be called using a preset JSON Schema. Specifically, the "filtered Resources" array structure of the filtering results is defined through the Schema. Each element in the array contains "rssUrl" and "data" fields. The "data" field contains an array of "items". Strict mode is enabled when calling to ensure that the output fully conforms to the Schema format.
[0022] Optionally, before obtaining the subscription task configuration information, the method further includes: receiving the user-input subscription task name, description information, AI model selection instructions, and task execution timestamp, constructing a complete subscription task configuration, and storing it in the database.
[0023] Secondly, this application provides an AI-based intelligent filtering and automated article generation system for RSS feeds, including the following functional modules:
[0024] The subscription task timed processing module is used to obtain subscription task configuration information, periodically check tasks to be executed and identify delayed tasks for batch processing, and monitor task execution status; the subscription task configuration information includes the RSS source base URL, push cycle type, filter prompt words and generated prompt words.
[0025] The source intelligent identification and data acquisition module is used to perform intelligent identification and verification on the RSS source base URL, obtain the latest content of the valid RSS source and preprocess it, and convert heterogeneous RSS data into a standardized internal data structure. The intelligent identification and verification of the RSS source base URL includes first directly accessing the RSS source base URL. If the access fails, it traverses the preset common RSS path suffix library to attempt URL concatenation. The content returned after concatenation is then checked for RSS format validity to determine the valid RSS source.
[0026] The multi-dimensional time filtering module is used to calculate a time window based on the push cycle type and filter out RSS content published within the time window to filter out expired RSS content.
[0027] The content dynamic grouping processing module is used to calculate the number of tokens for the selected RSS content, and dynamically group them according to the token processing limit of the AI screening model to ensure that a single article is not split and the total number of tokens in each group does not exceed the processing limit.
[0028] The structured AI filtering module is used to call the AI filtering model using a preset JSON Schema, and perform structured filtering on each group of RSS content in combination with the filtering prompts to obtain structured filtering results.
[0029] The reorganization and regrouping module is used to reorganize the filtering results of all groups, recalculate the number of tokens, and regroup them according to the token processing limit of the AI-generated model.
[0030] The article automation generation module is used to call the AI generation model and combine the generation prompts to automatically generate articles for each group of RSS content after regrouping.
[0031] The article storage and push module is used to call the title generation model to generate a matching article title for the generated article, save the complete article data to the database, and send notifications to subscribed users.
[0032] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the AI model-based RSS subscription source intelligent filtering and automated article generation method described above.
[0033] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the AI-based RSS subscription source intelligent filtering and automated article generation method described above.
[0034] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the AI-based RSS subscription source intelligent filtering and automated article generation method described above.
[0035] According to the specific embodiments provided in this application, the following technical effects are disclosed:
[0036] This application provides a method and related apparatus for intelligent filtering and automated article generation of RSS subscription sources based on an AI model. In this method, after obtaining subscription task configuration information, a timed scheduling and batch processing mechanism can prevent content timeliness loss due to task delays. Simultaneously, real-time status monitoring ensures that single-point task failures do not affect the overall process, significantly improving system service reliability. Furthermore, intelligent recognition eliminates the need for users to manually search for valid addresses by mastering RSS technical details; automatic concatenation attempts by traversing a preset suffix library greatly reduces the user configuration threshold, increasing the RSS source recognition success rate to over 95%. Heterogeneous data standardization processing avoids subsequent processing anomalies caused by data format differences, improving data flow efficiency. Subsequently, a time filtering mechanism accurately filters content that meets timeliness requirements, preventing expired information from consuming processing resources and interfering with users' access to valid information. A dynamic grouping strategy maximizes the processing capacity of the AI filtering model while ensuring semantic coherence through the "single article without segmentation" principle, preventing AI call failures due to input load exceeding limits. Based on this, a preset JSON... Schema enforces a unified AI output format, completely resolving the post-processing challenges caused by the chaotic output formats of traditional AI filtering. Combined with user-defined filtering prompts, it enables precise filtering based on semantic understanding. For changes in content volume after filtering, secondary grouping precisely adapts to the processing capabilities of the AI generation model, preventing incomplete or low-quality articles due to content imbalances, providing stable input for high-quality automated article generation. Finally, combined with user-defined generation prompts, it automates article generation, eliminating the need for manual content integration and formatting. This reduces the traditional hours-long task of content preparation to minutes, significantly lowering labor costs and meeting the personalized needs of different users for article format and style (e.g., academic, news, popular science). The automatic saving and notification functions automate the entire process from content generation to storage to user outreach, significantly improving user experience and addressing the pain point of poor user experience in traditional RSS systems. Attached Figure Description
[0037] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0038] Figure 1 This is a flowchart illustrating an AI-based method for intelligent filtering and automated article generation of RSS feeds, provided as an embodiment of this application.
[0039] Figure 2This is a schematic diagram of the functional modules of an AI-based intelligent filtering and automated article generation system for RSS subscription sources, provided as an embodiment of this application.
[0040] Figure 3 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation
[0041] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0042] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, this application will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0043] This application provides a method for intelligent filtering and automated article generation of RSS subscription sources based on an AI model. In an exemplary embodiment, such as... Figure 1 As shown, it includes the following steps:
[0044] A1. Obtain subscription task configuration information, periodically check pending tasks and identify delayed tasks for batch processing, while monitoring task execution status; subscription task configuration information includes RSS source base URL, push cycle type, filter prompts and generated prompts.
[0045] In one exemplary embodiment, the @Cron decorator and the EVERY_HOUR expression are used to check the tasks to be executed every hour, identify delayed tasks (such as tasks that have not been executed in the past 3 hours), process them in batches according to the maximum concurrency of the system, and record the task execution log (start time, status, completion time).
[0046] A2. Intelligent identification and verification are performed on the base URL of the RSS feed to obtain the latest content from valid RSS feeds and preprocess it, converting heterogeneous RSS data into a standardized internal data structure. The intelligent identification and verification of the base URL of the RSS feed includes first directly accessing the base URL of the RSS feed; if access fails, iterating through a preset library of common RSS path suffixes to attempt URL concatenation, and then verifying the RSS format validity of the returned content to determine a valid RSS feed. In an exemplary embodiment of this application, the preset library of common RSS path suffixes includes " / rss", " / feed", and " / atom.xml", and the RSS format validity verification includes verifying whether the returned data contains the fields "title", "item", and "pubDate".
[0047] A3. Calculate the time window based on the push cycle type and filter out RSS content published within the time window to remove expired RSS content. In this embodiment, step A3, "calculate the time window based on the push cycle type," specifically includes: if the push cycle type is "1 day," the time window is 24 hours prior to the current time; if it is "7 days," the time window is 168 hours prior to the current time; if it is "30 days," the time window is 720 hours prior to the current time, and the calculation process is synchronized with the system's deployed time zone information.
[0048] A4. Calculate the number of tokens for the selected RSS content, and dynamically group them according to the token processing limit of the AI screening model to ensure that a single article is not split and the total number of tokens in each group does not exceed the processing limit. Specifically, in this embodiment, the OpenAI official token calculation algorithm is used when calculating the number of tokens for the selected RSS content; a greedy algorithm is used when performing dynamic grouping, accumulating the total number of tokens in each group in real time. When adding a new article causes the total number of tokens to exceed the processing limit of the AI screening model, a new group is automatically created.
[0049] A5. Using a preset JSON Schema to call the AI filtering model, and combining the filtering prompts, perform structured filtering on each group of RSS content to obtain structured filtering results. In this embodiment, "using a preset JSON Schema to call the AI filtering model" in step A5 specifically includes: defining the "filtered Resources" array structure of the filtering results through the Schema, where each element contains "rssUrl" and "data" fields, and the "data" field contains an "items" array, and the strict mode is enabled when calling to ensure that the output fully conforms to the Schema format.
[0050] A6. Reorganize the filtering results of all groups, recalculate the number of tokens, and regroup according to the token processing limit of the AI generation model.
[0051] A7. Call the AI generation model and combine it with generated prompts to automatically generate articles for each group of RSS content after regrouping.
[0052] A8. The generated article is used to generate a matching article title using the title generation model, and the complete article data is saved to the database. At the same time, a notification is sent to the subscribing users.
[0053] In another exemplary embodiment of this application, before the "obtaining subscription task configuration information step" in step A1, the method further includes: receiving the subscription task name, description information, AI model selection instructions and task execution timestamp input by the user, constructing a complete subscription task configuration and storing it in the database.
[0054] Specifically, it receives user input of the subscription task name, description information, push period type (1 day / 7 days / 30 days), AI model selection instructions (filter model, generate model, title generate model) and task execution timestamp, constructs a complete subscription task configuration, and stores it in the database.
[0055] In another exemplary embodiment of this application, step A2, "performing intelligent identification and verification of the underlying URL of the RSS source," specifically includes:
[0056] Read the base URL of the RSS feed in the configuration, and first access the URL directly to verify the RSS format.
[0057] If access fails, iterate through the preset common RSS path suffix libraries (including " / rss", " / feed", and " / atom.xml"), concatenate the URLs one by one, and attempt to access them.
[0058] Perform RSS format validation on the returned content (verify whether it contains core fields such as "title", "item", and "pubDate"), determine the valid RSS source, and display it to the user for confirmation.
[0059] Step A2, "obtaining the latest content from a valid RSS feed and preprocessing it," specifically includes:
[0060] The Promise.all concurrency mechanism is used to retrieve the latest content from all valid RSS feeds.
[0061] The RSS parser converts heterogeneous XML / JSON data into a standardized internal data structure, extracting key information such as "title, content, publication time, author, and link".
[0062] Perform preprocessing: unify character encoding, standardize timestamps (convert to the system's internal time format), and clean up redundant HTML tags.
[0063] Step A3, "Calculate the time window based on the push cycle type, and filter out RSS content published within the time window to filter out expired RSS content," specifically includes:
[0064] Read the push period type and calculate the corresponding time window (e.g., "1 day" corresponds to 24 hours before the current time).
[0065] Iterate through all RSS posts, extract the "pubDate" field and convert it to a timestamp, then filter out expired posts that are outside the specified time window.
[0066] Record the number of articles before and after filtering to ensure the timeliness of the content.
[0067] Step A4, "Dynamically grouping based on the token processing limit of the AI screening model," specifically includes:
[0068] The number of tokens for each article is counted using OpenAI's official token calculation algorithm.
[0069] Based on the token processing limit of the AI screening model (e.g., 1000 tokens), a greedy algorithm is used to group the tokens: the total number of tokens in each group is accumulated in real time, and a new group is automatically created when the total number of tokens exceeds the limit due to the addition of new articles.
[0070] Enable the integrity protection mechanism to ensure that a single article is not split into different groups (if the token of a single article is close to the limit, it will be grouped independently).
[0071] Step A5, "Perform structured filtering on each group of RSS content," specifically includes:
[0072] Build filter suggestions: Combine grouped RSS content with user-defined filter conditions (such as "keep AI industry-related articles").
[0073] A predefined JSON Schema (defining an array structure of "filtered Resources" containing "rssUrl" and "data" fields) is used to enable strict mode to call the AI filtering model.
[0074] Perform filtering concurrently on each group of content, obtain the filtering results in a standard format, and filter out invalid content.
[0075] Step A6, "Reorganize the filtering results of all groups, recalculate the number of tokens, and regroup according to the token processing limit of the AI-generated model," specifically includes:
[0076] Merge the filtering results of all groups, and filter out empty results and invalid data items.
[0077] The number of tokens for the filtered content is recalculated, and the content is regrouped according to the token processing limit of the AI generation model (e.g., 800 tokens). Related articles from the same RSS source are prioritized to be assigned to the same group to ensure the continuity of the content theme.
[0078] Step A7, "combining generated prompts with automated article generation for each group of RSS content after regrouping," and step A8, "calling the title generation model to generate matching article titles for the generated articles," specifically include:
[0079] Build and generate prompts: integrate the reorganized grouped content with user-defined generation requirements (such as "formatted by 'title-summary-original link'").
[0080] The AI-generated model is invoked to generate structured articles using a non-streaming output mode.
[0081] For each generated article, a dedicated title generation model is invoked to generate a matching title of no more than 10 characters based on the article content.
[0082] Step A8, "Save the complete article data to the database and send a notification to subscribers," specifically includes:
[0083] Construct an article entity object (including fields such as "Article ID, Subscription ID, Title, Content, Generation Time, and Whether Read") and save it to the database. Specifically: Construct a complete article entity object, including core fields such as article content, title, subscription ID, and creation time; record the article's generation metadata, including task creation time, execution timestamp, and generation method identifier (scheduled / manual); save the article's creator and updater information; establish a complete data traceability chain; set the article's initial status to unread; and provide users with a new content notification function.
[0084] Query all users following this subscription task and send new article notifications (such as in-app messages or emails). Specifically: Query the list of all users following the subscription account, establish a many-to-many relationship between articles and users, create a personal article record for each user, support personalized reading status tracking, implement user permission verification, and ensure that only valid and active users can receive article pushes.
[0085] In another possible implementation, the method also includes an "exception handling mechanism": when a single RSS feed access fails or an AI model call times out, a warning log is logged but the overall process is not interrupted; when all content is filtered out, a log is generated and the user is notified that "there is no content matching the criteria in the current period".
[0086] As an optional implementation, the method also supports "personalized configuration": users can customize filter prompts (such as "prioritize keeping articles related to deep learning"), generation prompts (such as "adopt academic style layout"), AI model parameters (such as generation temperature 0.7, Token limit) and article structure requirements (such as "include chapter divisions and abstracts").
[0087] The following specific embodiment illustrates the method provided in this application. This embodiment takes "AI industry RSS subscription and article generation" as the scenario. The user needs are: to subscribe to the RSS feeds of "Minority Report" (https: / / sspai.com) and "SimonWilison's Weblog" (https: / / simonwillison.net), generate AI industry-related articles on a "1-day" cycle, and format the articles as "title-summary content-original link". The push time is 11:00 every day.
[0088] Before executing a subscription task, users must first complete the system configuration process, including subscription task configuration and system parameter configuration.
[0089] Subscription Task Configuration: Task Name: "Daily RSS Summary of the AI Industry"; Push Cycle: 1 day; Basic RSS Source URL: https: / / sspai.com, https: / / simonwillison.net; Filtering Prompts: "Retain articles related to the AI industry, including topics such as AI tool updates, LLM technology, and RAG systems"; Generation Prompts: "Arranged in the order of 'title-summary content-original link,' the summary content should extract the core viewpoints and avoid generalities"; AI Model Selection: Filtering Model, Article Generation Model, Title Generation Model; Execution Time: 11:00 AM daily.
[0090] System parameter configuration: Maximum number of filter model tokens: 1000; Maximum number of article generation model tokens: 800; Common RSS path suffix library: [" / rss"," / feed"," / atom.xml"]; System time zone: UTC+8.
[0091] Subsequently, corresponding to step A1 of the aforementioned method, the @Cron decorator is used to check the tasks to be executed every hour (EVERY_HOUR); when the system time is 2025-09-15 11:00:00 (UTC+8), the database is queried to match the task "AI Industry Daily RSS Summary"; the integrity of the task configuration is verified (AI model is available, RSS source is not invalid), the task is added to the execution queue, and the task creator information and model configuration are obtained in batches.
[0092] Corresponding to step A2 of the aforementioned method, read the base URLs of the RSS feed: https: / / sspai.com and https: / / simonwillison.net.
[0093] For https: / / sspai.com: Direct access failed. Traverse the suffix library and concatenate the URL: concatenate " / feed" → access https: / / sspai.com / feed. The returned content contains "title:minority", "item", and "pubDate". The RSS format verification passed, and it was determined to be a valid RSS source.
[0094] For https: / / simonwillison.net: Direct access failed. Traverse the suffix libraries and concatenate the URL: concatenate " / atom / everything" → access https: / / simonwillison.net / atom / everything. The returned content conforms to the RSS format, so it is determined to be a valid RSS source.
[0095] Show the user the recognition results (2 valid RSS feeds, including feed titles), and proceed to the next step after the user confirms.
[0096] Then, Promise.all is used to concurrently retrieve the content from two valid RSS feeds:
[0097] The Minority Report RSS feed (https: / / sspai.com / feed) returned 5 articles, including "Making Weight Loss a Habit: How I Lost 50 Pounds in Two Years Using a Lifestyle Approach", "Minority Report Daily: iPhone Air China Release Delayed", and "Major AI Tool Update: Comprehensive Analysis of ChatGPT's New Voice Function".
[0098] Simon Wilison's Weblog (https: / / simonwillison.net / atom / everything) returns four articles, including "Models can prompt now" and "Building a RAG system with Claude and TypeScript".
[0099] Then the following preprocessing is performed:
[0100] The unified character encoding is UTF-8;
[0101] Standardized timestamp: Convert "2025-09-15T03:28:09Z" to UTC+8 time "2025-09-1511:28:09";
[0102] Clean up HTML tags: Remove "" from content <strong>Tags such as "etc."
[0103] Ultimately, the heterogeneous RSS data is transformed into a standardized internal data structure.
[0104] Corresponding to step A3 of the aforementioned method, the push period "1 day" is read, and the time window is calculated as: 2025-09-14 11:00:00 to 2025-09-15 11:00:00 (UTC+8).
[0105] Iterate through all articles and filter out those whose publication time falls within the specified time window:
[0106] Three articles were retained from the Minority Report: "Make weight loss a habit..." (2025-09-15 11:28:09), "Minority Report: iPhone Air..." (2025-09-15 08:59:36), and "Major update of AI tools..." (2025-09-15 18:00:00).
[0107] Simon Wilison's Weblog retains two posts: "Models can prompt now" (2025-09-15 04:25:21) and "Building a RAG system..." (2025-09-15 01:00:00).
[0108] Three expired articles (published before 2025-09-14 11:00:00) were filtered out.
[0109] Corresponding to step A4 of the aforementioned method, calculate the number of tokens for each article: Minority Report article 1 has 180 tokens, Minority Report article 2 has 160 tokens, and Minority Report article 3 has 200 tokens. Simon's article 1 has 420 tokens, and Simon's article 2 has 380 tokens.
[0110] Grouping by the maximum Token limit (1000) of the filtering model: Group 1 consists of 3 articles from Minority Report (180+160+200=540) + 1 article from Simon (420), totaling 960 Tokens (≤1000); Group 2 consists of 2 articles from Simon (380 Tokens, ≤1000). After confirming that no articles were split, the grouping is complete.
[0111] Corresponding to step A5 of the aforementioned method, filter prompts are constructed for each group, such as the prompt for group 1: "RSS subscription source data: [Group 1 article content]. Filtering criteria: retain articles related to the AI industry, including topics such as AI tool updates, LLM technology, and RAG systems." Then, strict mode is enabled, and the filtering model is called concurrently to process the two groups of content.
[0112] Group 1 Filtering Results: Retain "Major AI Tool Update: Comprehensive Analysis of ChatGPT's New Voice Function" (Minority Report) and "Models can prompt now" (Simon);
[0113] Group 2 screening results: "Building a RAG system with Claude and TypeScript" (Simon) is retained;
[0114] Filter out articles that are not related to AI, such as "weight loss" and "iPhone Air".
[0115] Corresponding to step A6 of the aforementioned method, first merge the two sets of screening results: a total of 3 AI-related articles; then recalculate the number of tokens: Minority Report Article 3 (200), Simon Article 1 (420), Simon Article 2 (380); and then group them a second time according to the upper limit of the article generation model token (800): Generation group 1 is Minority Report Article 3 (200) + Simon Article 1 (420), totaling 620 tokens (≤800); Generation group 2 is Simon Article 2 (380 tokens, ≤800).
[0116] Corresponding to steps A7-A8 of the aforementioned method, first construct the generation prompt for Generation Group 1: "Subscribe to source data: [Generate Group 1 article content]. Formatting requirements: format in the order of 'title-summary content-original link', summarize the core viewpoints, and avoid generalities."; call the article generation model to generate the article, and call the title generation model to generate the corresponding article title. After generating the article and the corresponding title, construct the article entity object and save the complete article to the database; then query the list of subscribers for this task (50 people in total) and send a notification via in-site message: "Your subscribed 'AI Industry Daily RSS Summary' has been updated with 2 articles, click to view: [link]".
[0117] In this embodiment, the following objectives were successfully achieved:
[0118] RSS source identification success rate: Both basic URLs successfully identified valid RSS sources, with a success rate of 100%.
[0119] Time filtering accuracy: 3 out of 5 articles were selected as valid articles, with a timeliness accuracy rate of 100%.
[0120] Filtering precision: 2 non-AI-related articles were filtered out, with a filtering precision of 100%.
[0121] High intelligence: Through a dual AI model collaboration mechanism, it realizes intelligent filtering of RSS content and automatic generation of high-quality articles, shortening the content planning work that traditionally requires several hours of professional editing to the minute level, and the level of intelligence reaches the industry-leading level.
[0122] High accuracy: Adopting OpenAI Structured Outputs' strict output control technology, it ensures the consistency of AI processing results in terms of format and accuracy of content, improving screening accuracy by more than 80% and generating articles with quality close to that of human editors.
[0123] High personalization: It supports users to customize filtering conditions, generation style, push cycle and other multi-dimensional personalized configurations, truly realizing a personalized content service for each user, and significantly improving user satisfaction.
[0124] High concurrency: Through intelligent grouping and the Promise.all concurrent processing architecture, it supports the simultaneous processing of hundreds of RSS feeds and multiple AI tasks, and the system's processing capacity is 5-10 times higher than traditional solutions.
[0125] High timeliness: The intelligent time filtering mechanism based on the push cycle ensures the timeliness and relevance of the content, avoids interference from outdated information, and achieves a content timeliness accuracy rate of over 95%.
[0126] High reliability: A comprehensive fault tolerance mechanism and anomaly handling system are established, so that a single point of failure does not affect the overall service, and the system availability reaches more than 99.9%.
[0127] Reduce costs: Reduce manual content preparation time by 90%, lower content operation costs, and improve the return on investment for enterprise content production.
[0128] Strong scalability: The modular design architecture facilitates functional expansion and technology upgrades, and can quickly adapt to new RSS source types, AI models, and changes in business needs.
[0129] Based on the same inventive concept, this application also provides a system for implementing the AI model-based intelligent filtering and automated article generation method for RSS subscription sources described above. The solution provided by this system is similar to the implementation scheme described in the above method. In an exemplary embodiment, such as... Figure 2 As shown, an AI-based intelligent filtering and automated article generation system for RSS feeds is provided, including the following functional modules:
[0130] The subscription task timed processing module is used to obtain subscription task configuration information, periodically check tasks to be executed and identify delayed tasks for batch processing, and monitor task execution status; the subscription task configuration information includes the RSS source base URL, push cycle type, filter prompt words and generated prompt words.
[0131] The source intelligent identification and data acquisition module is used to perform intelligent identification and verification on the RSS source base URL, obtain the latest content of the valid RSS source and preprocess it, and convert heterogeneous RSS data into a standardized internal data structure. The intelligent identification and verification of the RSS source base URL includes first directly accessing the RSS source base URL. If the access fails, it traverses the preset common RSS path suffix library to attempt URL concatenation. The content returned after concatenation is then checked for RSS format validity to determine the valid RSS source.
[0132] The multi-dimensional time filtering module is used to calculate a time window based on the push cycle type and filter out RSS content published within the time window to filter out expired RSS content.
[0133] The content dynamic grouping processing module is used to calculate the number of tokens for the selected RSS content, and dynamically group them according to the token processing limit of the AI screening model to ensure that a single article is not split and the total number of tokens in each group does not exceed the processing limit.
[0134] The structured AI filtering module is used to call the AI filtering model using a preset JSON Schema, and perform structured filtering on each group of RSS content in combination with the filtering prompts to obtain structured filtering results.
[0135] The reorganization and regrouping module is used to reorganize the filtering results of all groups, recalculate the number of tokens, and regroup them according to the token processing limit of the AI-generated model.
[0136] The article automation generation module is used to call the AI generation model and combine the generation prompts to automatically generate articles for each group of RSS content after regrouping.
[0137] The article storage and push module is used to call the title generation model to generate a matching article title for the generated article, save the complete article data to the database, and send notifications to subscribed users.
[0138] certainly, Figure 2 The architecture shown is merely exemplary; it can be omitted as needed when implementing different functionalities. Figure 2 One or at least two components of the system shown.
[0139] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 3 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs in the non-volatile storage media to run. The database stores user-configured subscription task settings and generated article content. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it can implement the AI-based RSS subscription source intelligent filtering and automated article generation method provided in the previous embodiment.
[0140] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0141] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.
[0142] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.
[0143] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.
[0144] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.
[0145] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).
[0146] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.
[0147] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0148] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.< / strong>
Claims
1. A method for intelligent filtering and automated article generation of RSS subscription sources based on an AI model, characterized in that, include: Obtain the subscription task configuration information, periodically check the tasks to be executed and identify delayed tasks for batch processing, while monitoring the task execution status; The subscription task configuration information includes the RSS source base URL, push cycle type, filter prompts, and generated prompts; The system performs intelligent identification and verification on the base URL of the RSS source, obtains the latest content of the valid RSS source and preprocesses it, and converts heterogeneous RSS data into a standardized internal data structure. The intelligent identification and verification on the base URL of the RSS source includes first directly accessing the base URL of the RSS source. If the access fails, it traverses a preset library of common RSS path suffixes to attempt URL concatenation. The system then verifies the RSS format validity of the content returned after concatenation to determine the valid RSS source. Calculate the time window based on the push cycle type, and filter out the RSS content published within the time window to filter out expired RSS content. Calculate the number of tokens in the selected RSS content, and dynamically group them according to the token processing limit of the AI screening model to ensure that a single article is not split and the total number of tokens in each group does not exceed the processing limit. The AI filtering model is invoked using a preset JSON Schema, and the filtering prompts are combined to perform structured filtering on each group of RSS content to obtain structured filtering results; The filtering results of all groups are reorganized, the number of tokens is recalculated, and the groups are regrouped according to the token processing limit of the AI-generated model. The AI generation model is invoked, and the generated prompts are used to automatically generate articles for each group of RSS content after regrouping. The generated article is used to generate a matching article title using a title generation model, and the complete article data is saved to the database. At the same time, a notification is sent to the subscribing users.
2. The method for intelligent filtering and automated article generation of RSS subscription sources based on an AI model according to claim 1, characterized in that, The preset library of common RSS path suffixes includes " / rss", " / feed" and " / atom.xml". The RSS format validity check includes verifying whether the returned data contains the fields "title", "item" and "pubDate".
3. The method for intelligent filtering and automated article generation of RSS subscription sources based on an AI model according to claim 1, characterized in that, The time window is calculated based on the push cycle type, specifically as follows: if the push cycle type is "1 day", the time window is 24 hours before the current time; if it is "7 days", the time window is 168 hours before the current time; if it is "30 days", the time window is 720 hours before the current time, and the time zone information of the system deployment is adapted synchronously during the calculation process.
4. The method for intelligent filtering and automated article generation of RSS subscription sources based on an AI model according to claim 1, characterized in that, When calculating the number of tokens for the selected RSS content, the official OpenAI token calculation algorithm is used. When performing dynamic grouping, a greedy algorithm is used to accumulate the total number of tokens for each group in real time. When the addition of new articles causes the total number of tokens to exceed the processing limit of the AI screening model, a new group is automatically created.
5. The method for intelligent filtering and automated article generation of RSS subscription sources based on an AI model according to claim 1, characterized in that, The AI filtering model is invoked using a preset JSON Schema. Specifically, the "filtered Resources" array structure of the filtering results is defined through the Schema. Each element in the array contains "rssUrl" and "data" fields. The "data" field contains an array of "items". Strict mode is enabled when calling the model to ensure that the output fully conforms to the Schema format.
6. The method for intelligent filtering and automated article generation of RSS subscription sources based on an AI model according to claim 1, characterized in that, Before obtaining the subscription task configuration information, the method further includes: receiving the subscription task name, description information, AI model selection instructions and task execution timestamp input by the user, constructing a complete subscription task configuration and storing it in the database.
7. A system for intelligent filtering and automated article generation of RSS feeds based on an AI model, characterized in that, include: The subscription task timed processing module is used to obtain subscription task configuration information, periodically check tasks to be executed and identify delayed tasks for batch processing, and monitor task execution status. The subscription task configuration information includes the RSS source base URL, push cycle type, filter prompts, and generated prompts; The source intelligent identification and data acquisition module is used to perform intelligent identification and verification on the RSS source base URL, obtain the latest content of the valid RSS source and preprocess it, and convert heterogeneous RSS data into a standardized internal data structure; the intelligent identification and verification of the RSS source base URL includes first directly accessing the RSS source base URL, and if the access fails, traversing the preset common RSS path suffix library to attempt URL concatenation, and performing RSS format validity verification on the content returned after concatenation to determine the valid RSS source; A multi-dimensional time filtering module is used to calculate a time window based on the push cycle type and filter out RSS content published within the time window to filter out expired RSS content. The content dynamic grouping processing module is used to calculate the number of tokens for the filtered RSS content, and dynamically group them according to the token processing limit of the AI filtering model to ensure that a single article is not split and the total number of tokens in each group does not exceed the processing limit. The structured AI filtering module is used to call the AI filtering model using a preset JSON Schema, and perform structured filtering on each group of RSS content in combination with the filtering prompts to obtain structured filtering results; The reorganization and regrouping module is used to reorganize the filtering results of all groups, recalculate the number of tokens, and regroup according to the token processing limit of the AI-generated model. The article automatic generation module is used to call the AI generation model and combine the generation prompt words to automatically generate articles for each group of RSS content after regrouping. The article storage and push module is used to call the title generation model to generate a matching article title for the generated article, save the complete article data to the database, and send notifications to subscribed users.
8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the AI-based intelligent filtering and automated article generation method for RSS subscription sources according to any one of claims 1-6.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the AI-based intelligent filtering and automated article generation method for RSS subscription sources as described in any one of claims 1-6.
10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the AI-based intelligent filtering and automated article generation method for RSS subscription sources as described in any one of claims 1-6.