Method for integrating and displaying similar comments on online post based on generative ai
The generative AI-based system effectively groups and summarizes user comments, providing quantitative insights into opinion resonance, thus enhancing user experience and information accessibility.
Patent Information
- Application Number
- PCT/KR2026/096647
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-05-16
- Filing Date
- 2026-05-07
- Publication Date
- 2026-11-19
AI Technical Summary
Existing comment display systems on online platforms fail to efficiently group and summarize user comments, leading to redundant exposure of similar opinions and lack of quantitative insight into user participation, making it difficult for users to grasp the overall flow of opinions.
A generative AI-based system that automatically classifies similar comments, generates representative summaries, displays the number of commenters per group, and sorts them based on participation, allowing for context-aware and user-customized analysis.
Enhances user understanding of comment flow and opinion weight by integrating and quantifying subjective opinions, improving information accessibility and user satisfaction by providing core opinions in a structured manner.
Smart Images

Figure KR2026096647_19112026_PF_FP_ABST
Abstract
Description
Generative AI-based method for integrating and displaying similar comments on online posts
[0001] The present invention relates to a technology for processing comments on content provided online (e.g., articles, blogs, social media posts, videos, advertisements, etc.), and more specifically, to a technology that uses generative AI to automatically classify similar comments according to pre-set criteria, generates a representative summary comment for each group, displays the number of comment writers included in the group, and sorts and displays them based on the number of writers.
[0002] Today, online content platforms (e.g., news articles, blogs, YouTube, Instagram, advertisements, etc.) collect a vast number of comments from users, which serve as an important means of inducing interaction among users and measuring the influence of content. In particular, popular content can receive hundreds to thousands of comments, leading platform users to face a situation where it is difficult to check every single comment.
[0003] Existing comment display systems typically arranged comments in simple chronological order or based on the number of upvotes. This approach leads to the repeated exposure of identical or similar opinions, making it difficult to grasp the overall flow of the comments or the core opinions. While some platforms display popular comments at the top or provide filtering features for profanity, there are still limitations in filtering out duplicate opinions or identifying meaningful groups of opinions.
[0004] Furthermore, while there has been a recent increase in attempts to summarize conversations or automatically classify documents using artificial intelligence technology, particularly Natural Language Processing (NLP), most research and implementations focus on document summarization or question-and-answer systems. Consequently, there is a lack of functionality to group or integrate content with a unique structure known as 'comments,' which are written by multiple users on the same topic.
[0005] Furthermore, even though integrating meaningful opinions presented by multiple users among numerous comments allows users to easily grasp the overall flow of opinions, existing technologies lacked the functionality to quantitatively indicate the number of users in each group who shared a particular opinion or to sort them based on the number of user participations.
[0006] Accordingly, there is a need for a system that automatically classifies and integrates similar comments, displays the number of commenters in each group along with a summary representative opinion, and sorts them based on these criteria. In particular, by utilizing generative AI, such a system has the advantage of integrating comments in a more sophisticated and context-aware manner.
[0007] The following registered Korean patents serve as prior art literature.
[0008] Republic of Korea Registered Patent 10-1946022 Method for analyzing and displaying comments and apparatus for the same
[0009] Republic of Korea Registered Patent 10-2032091 Artificial Intelligence-based Comment Sentiment Analysis Method and System
[0010] Republic of Korea Registered Patent 10-1570841 Method and System for Indicating Interests of a Post via Comments
[0011] Republic of Korea Registered Patent 10-2339721 System for Contentization of Online Comments
[0012] Republic of Korea Registered Patent 10-2427719 Comment-based distortion behavior suppression system
[0013] Republic of Korea Registered Patent 10-2511170 Method, apparatus, and computer program for providing a secondary content prediction solution through comment big data analysis
[0014] To solve the aforementioned problems, the present invention provides a method for integrating and displaying similar comments that enables users to efficiently understand a vast number of comments by automatically classifying comments with similar meanings among a number of comments written on online posts (articles, blogs, videos, advertisements, etc.) using generative AI, and summarizing and providing representative opinions of each group.
[0015] Another objective of the present invention is to provide a comment integration and sorting method that displays the number of comment writers included in each integrated comment group and sorts the comment groups based on this, thereby intuitively reflecting the weight of opinions and the degree of user participation, and enabling users to easily grasp the flow of opinions.
[0016] Another objective of the present invention is to provide a generative AI-based comment classification, integration, and display method that enables flexible comment analysis suitable for the situation and the provision of user-customized results by inputting preset criteria applicable during the comment classification and integration process into the generative AI in the form of a prompt, thereby allowing the criteria to be reflected in the clustering and integration results.
[0017] To achieve the aforementioned purpose, the present invention relates to a system for processing user comments on online content, comprising: an AI command input module (110) that utilizes generative AI to receive an analysis command in the form of natural language input by a condition setter and to interpret the command to convert it into a setting value or filter condition applicable to clustering; a comment collection module (120) that collects a number of comments written by a commenter on the online content; an embedding and clustering module (130) that utilizes generative AI to convert the collected comments into numeric vectors through a natural language processing-based embedding technique (e.g., word embedding, sentence embedding, etc.) and to automatically group similar comments according to filtering criteria or clustering criteria interpreted in advance by the condition setter; a summary generation module (140) that utilizes generative AI to automatically generate representative summary comments by integrating key content for each cluster; a writer count calculation module (150) that calculates the number of comment writers included in each cluster and stores it as metadata; and a sorting and display module that sorts clusters based on the writer count and displays them to the user. A comment integration and display system is provided, characterized by including a module (160) and a function (170) that allows all individual comments within a cluster to be viewed by clicking on a representative summary comment integrated by cluster.
[0018] According to the present invention with the above configuration, the following effects are achieved.
[0019] By automatically analyzing comments written by multiple users on online posts using generative AI, grouping similar comments, and summarizing the key points of each group to display as representative comments, users can intuitively grasp the flow of the overall opinion without having to read through the vast number of comments one by one.
[0020] In addition, by displaying the number of commenters included in each comment group and sorting them based on this criterion, it is possible to provide users with quantitative insights into which opinions have actually resonated with many people, thereby significantly improving information accessibility and user satisfaction.
[0021] In particular, the present invention quantifies people's subjective opinions into numerical values (e.g., number of commenters) and visually represents them, thereby enabling users to intuitively understand the weight of opinions and efficiently consume key opinions from more content within a short period of time.
[0022] Furthermore, in the clustering and integrated processing methods, condition setters can specify classification criteria or filtering conditions in the form of commands, enabling customized comment analysis tailored to the situation and allowing for flexible analysis focused on specific topics or sentiments.
[0023] The present invention has the effect of simultaneously improving information quality and customer experience (UX) within a platform by eliminating redundant comment expressions and providing only core opinions in an integrated manner, thereby increasing content consumption efficiency and allowing the exchange of opinions among users to unfold in a more structured manner.
[0024] Figure 1 is a configuration diagram of a system for integrating and displaying similar comments of online posts based on generative AI according to the present invention.
[0025] FIG. 2 is a flowchart showing a comment processing process according to one embodiment of the present invention.
[0026] FIG. 3 is an example of a user interface (UI) screen of a sorting result of representative comments and the number of authors according to an embodiment of the present invention.
[0027] Embodiments of the present invention are described in detail below with reference to the attached drawings. In order to clearly explain the invention in the drawings, parts unrelated to the explanation are omitted, and the same reference numerals are used for identical parts throughout the specification.
[0028] Figure 1 is a configuration diagram of a system for integrating and displaying similar comments of online posts based on generative AI according to the present invention.
[0029] As described above, the system includes a terminal (101) for inputting user comments and verifying results, a comment integration server (102) that collects, analyzes, and integrates received comments to display representative opinions, and a generative AI model (200) responsible for text embedding and integrated generation, and these are interconnected through a communication network (Internet, etc.).
[0030] A terminal (101) is a device that allows a user to enter comments on online content such as news, blogs, videos, and advertisements, and to check the integrated results processed by the server. It is implemented in various forms such as a personal computer, smartphone, and tablet, and operates in a web or application environment.
[0031] When a user writes a comment while viewing content, the terminal transmits the comment to the server (102). At this time, the user uses a standard comment writing window, and the written comment is received and stored on the server through the comment collection module (120).
[0032] When a comment is saved, the embedding and clustering module (130) vectorizes it using a natural language processing-based embedding technique and automatically groups it using a clustering algorithm (K-means, DBSCAN, etc.) and a generative AI model (200) based on similar meanings.
[0033] For each cluster, the summary generation module (140) generates a representative comment that summarizes the core content in a natural and clear sentence using a generative AI model (200), and the author count calculation module (150) calculates the number of unique comment authors included in the cluster to quantify the degree of empathy for the opinion.
[0034] The sorting and display module (160) sorts the clusters based on the number of authors and outputs the combined comments and the number of authors to the user terminal.
[0035] Users can intuitively understand the overall flow of opinions by browsing through representative opinions by cluster, and can view all actual individual comments included in the cluster by clicking the [View Details] button at the bottom of a representative comment of interest.
[0036] Additionally, the condition setter can directly set clustering conditions through the AI command input module (110). For example, if conditions such as “classifying comments into positive, neutral, and negative opinions,” “classifying into five categories such as delivery, quality, price, customer service, and design,” “classifying into questions, praise, complaints, reviews, requests, and others,” and “classifying into direct expressions, indirect expressions, politeness, disparagement, and logical expressions” are entered, the conditions are reflected in the embedding and clustering module (130), and the comment analysis and integration results are reconstructed in real time accordingly.
[0037] FIG. 2 is a diagram illustrating the internal configuration of a comment integration server (102) according to an embodiment of the present invention, which is a core processing system that receives comments written by a user through a terminal (101), automatically analyzes and integrates them, and outputs results.
[0038] When a user writes a comment on an online post, the comment is first transmitted to a comment collection module (120). The module receives and stores raw comment data including information such as the time of writing, author ID, original text, and location of writing, and performs preprocessing (e.g., removal of special characters, removal of duplicates, etc.) to convert it into a structured form.
[0039] Afterwards, the embedding and clustering module (130) receives this data and embeds each comment at the sentence level.
[0040] This embedding utilizes deep learning-based natural language processing models such as BERT and RoBERTa to convert comments into high-dimensional vectors, and groups the converted vectors into comments with similar meanings using similarity-based clustering algorithms (K-means, HDBSCAN, etc.).
[0041] In addition, it is also possible to implement a generative AI model (200) as an auxiliary tool in the preliminary interpretation stage for setting clustering criteria or determining similarity.
[0042] In this case, even if the user uses similar expressions differently (e.g., “It’s so funny,” “That was really funny”), the comments are determined to be semantically similar and classified into the same cluster. Thanks to this processing, the present invention enables integration that recognizes the 'context of the opinion' rather than simple text matching.
[0043] When a cluster is created, a summary generation module (140) aggregates the comments within each cluster and generates a representative summary comment in a natural sentence through a generative AI model (200). This is a process in which a single concise opinion representing a common meaning is automatically derived from among numerous comments written by users.
[0044] Next, the author count calculation module (150) calculates the number of unique users included in each cluster and displays the number of users who agreed with this opinion.
[0045] The sorting and display module (160) sorts representative comments based on the number of authors and configures data to be output to the user.
[0046] The feature to view individual comments by cluster (170) is a UI interaction feature that allows the user to expand and view the actual individual comments included in the group by clicking the 'View Details' button at the bottom of the integrated comment.
[0047] FIG. 3 is a diagram illustrating the entire flow from writing a comment (101) from the user's perspective, automatic analysis (102), and displaying the integrated result.
[0048] Users view online posts (content) such as blogs, news, and videos, write their impressions or opinions as comments, and the comments are transmitted to a comment integration server (102) via a terminal (101). The server (102) receives these in real time, stores them in a comment collection module (120), and embeds the comments through an embedding and clustering module (130). The processed vectors are assigned to the most appropriate cluster by calculating semantic similarity with existing comments, or a new cluster is created.
[0049] When clustering is complete, the summary generation module (140) uses a generative AI model (200) to automatically generate a natural language summary sentence representing the meaning of the cluster. This representative comment is written in a style similar to actual user expressions and is delivered in a language familiar to the user. Subsequently, the author count calculation module (150) calculates the number of unique authors included in each cluster, and the sorting and display module (160) sorts and organizes the cluster list in order, starting with the representative comment with the most authors.
[0050] The sorted list of representative comments is transmitted to the user terminal (101), allowing the user to intuitively and quickly grasp the flow of the entire comment and the main opinions. At this time, next to the sorted representative comments, the 'number of people who agreed with this opinion' e.g. (27 people) is displayed.
[0051] When the user clicks the [View Details] button displayed at the bottom of each representative comment, they can view the actual individual comments included in the cluster through the cluster-specific individual comment viewing function (170), and the structure is provided so that the original comments as well as the integrated representative opinions can be viewed together.
[0052] Such a system can dramatically improve information accessibility and content consumption efficiency by removing redundant expressions from a vast amount of comments and integrating semantically similar opinions to provide users with only core information. In particular, by quantifying people's subjective opinions into objective figures (number of writers) and expressing them visually, it provides the effect of allowing one to grasp the weight of opinions at a glance.
[0053] Although the present invention has been described in detail with reference to preferred embodiments, this is by way of example and the invention may be implemented in other specific forms without altering the technical spirit or essential features of the invention; therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive.
[0054] Furthermore, the scope of the present invention is defined by the claims set forth below rather than by the detailed description above, and all modifications or variations derived from the meaning and scope of the claims and equivalent concepts thereof should be interpreted as being included within the scope of the present invention.
[0055] [Explanation of the symbol]
[0056] 100: Entire comment integration and display system
[0057] 101: Terminal
[0058] 102: Server
[0059] 110: AI Command Input Module
[0060] 120: Comment Collection Module
[0061] 130: Embedding and Clustering Module
[0062] 140: Summary Generation Module
[0063] 150: Author count calculation module
[0064] 160: Alignment and Display Module
[0065] 170: Feature to view individual comments by cluster
[0066] 200: Generative AI Models
Claims
1. Regarding the method of integrating and displaying similar comments on online posts, A step of providing an input tool to a user terminal to enable the writing of opinions related to online posts; A step of vectorizing multiple comments input by a user using natural language embedding techniques and clustering similar comments together based on semantic similarity; A step of generating a representative integrated comment using generative AI for each of the clustered similar comment groups; A step of calculating the number of unique comment writers included in each similar comment group and storing the figures as cluster-specific metadata; Characterized by including the step of displaying the number of authors of the above representative integrated comments and the corresponding comment group on a user terminal, and sorting and outputting multiple representative comments based on the number of authors. Method for integrating and displaying similar comments on online posts.
2. In Paragraph 1, The above similar comment clustering step is, Filtering conditions such as classification criteria, emotional state, and number of classification categories are received from the condition setter in the form of commands through the AI command input module, and After interpreting the received conditions using generative AI, Characterized by performing clustering only on a set of comments filtered according to the corresponding condition, Method for integrating and displaying similar comments on online posts.
3. In Paragraph 1, The above step for displaying the number of authors is, Calculate the number of non-duplicate user IDs within each cluster, and Characterized by visually displaying the calculated number of authors corresponding to the representative comment, Method for integrating and displaying similar comments on online posts.
4. In Paragraph 1, The above user terminal is characterized by providing an interface that allows viewing all individual comments included in the corresponding cluster when a representative comment is selected. Method for integrating and displaying similar comments on online posts.
