Web-based dynamic object insertion and user behavior-driven data archiving system using PDF rendering engine

The system provides real-time dynamic interaction and user behavior data management for PDFs, ensuring document integrity and automating data conversion for SEO and AI, addressing the limitations of existing PDF technologies.

KR103045404B1Active Publication Date: 2026-11-10DESIGN ART PLUS CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
KR1020260045603
Authority / Receiving Office
KR · KR
Patent Type
Patents
Current Assignee / Owner
Filing Date
2026-03-13
Publication Date
2026-11-10
Estimated Expiration
2046-03-13

AI Technical Summary

Technical Problem

PDF documents lack interactive elements and user behavior data analysis capabilities, with existing methods compromising document integrity and failing to automate data conversion into structured data for SEO and AI recommendations.

Method used

A system that overlays dynamic interaction objects in real-time during PDF rendering without modifying the original document, using a web-based PDF rendering engine to insert and manage interactive elements, and automatically collects and structures user behavior data for SEO and AI recommendations.

Benefits of technology

Enables bidirectional interaction with PDFs while preserving document integrity, automates user behavior data conversion into structured data for SEO and AI recommendations, and improves recommendation accuracy through feedback loops.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 112026030839409-PAT00001_ABST
    Figure 112026030839409-PAT00001_ABST
Patent Text Reader

Abstract

The present invention relates to a system that, when rendering a PDF document in a web browser, superimposes dynamic interaction objects in real-time at the time of rendering without modifying the original PDF file, and automatically converts collected user behavior data into structured data. The system according to the present invention includes a virtual interaction layer control unit that creates a transparent layer after PDF.js-based rendering to place interaction objects, a semantic anchoring module that determines object positions based on text context, a semantic data conversion engine that converts raw logs into structured data, and an automatic archiving manager that performs automatic content generation and SEO integration, thereby enabling the conversion of static PDF documents into dynamic data assets.
Need to check novelty before this filing date? Find Prior Art

Description

Technology Field

[0001] The present invention relates to a system that overlays dynamic interaction objects in real time without modifying the original PDF when rendering a PDF document in a web browser, and automatically converts user behavior data into structured data. Background Technology

[0002] PDF (Portable Document Format) is a document format that preserves the layout and content of a document independently of the device, and is widely used in various fields such as academic materials, corporate reports, and educational materials. In particular, in the web environment, open-source rendering engines such as PDF.js have made it possible to display PDF documents directly in the browser without separate plugins.

[0003] However, since PDF is inherently a static document format, it has limitations in supporting two-way interaction with users. Conventional technology used a method of editing the original PDF file itself to insert form fields or hyperlinks in order to add interactive elements to a PDF document. This method not only compromised the integrity of the original document but also entailed the inconvenience of having to re-insert interactive elements every time the document was revised.

[0004] Meanwhile, technologies for analyzing user browsing behavior data and utilizing it for content recommendation or search optimization are advancing on web-based content platforms. However, in the case of PDF documents, technologies for systematically collecting user behavior within the browser and converting it into structured data for use have not yet sufficiently developed. Existing web analytics tools are optimized for HTML-based web pages, making it difficult to collect detailed behavioral data (such as dwell time at specific locations on the page, text selection areas, and zoom patterns) within the PDF rendering environment.

[0005] Furthermore, no technology has yet been proposed to automate the series of processes that automatically standardize collected user behavior data and convert it into structured data usable for search engine optimization (SEO) by linking it with a content management system (CMS). The problem to be solved

[0006] The present invention aims to provide a non-destructive system capable of overlaying dynamic interaction objects in real-time at the time of web browser rendering without modifying the original PDF file.

[0007] In addition, the present invention provides a means to automatically relocate the position of an interaction object even when the document is revised through text context-based semantic anchoring.

[0008] In addition, the present invention provides a fully automated pipeline that automatically collects and structures user behavior data and utilizes it for automatic content generation, SEO structured data integration, and AI recommendations. means of solving the problem

[0009] The system for inserting dynamic interaction objects and automatically generating user behavior-based data using a web-based PDF rendering engine according to the present invention performs the insertion of dynamic interaction objects and the automatic generation of user behavior-based data for a PDF document in a web browser, and comprises a user terminal accessing the service through a web browser, a database storing user behavior data, document metadata, and an overlay map, and a service providing server that provides web-based PDF dynamic interaction and data archiving services. The service providing server includes a PDF rendering control module that controls PDF.js-based rendering to transmit PDF files and overlay map data to the user terminal, a virtual interaction layer control unit that creates a transparent layer on top of the canvas with the same size and position as the canvas after PDF.js completes rendering the PDF onto the canvas, and places interaction objects defined in the overlay map on the transparent layer, a coordinate transformation engine that transforms the coordinate data of the overlay map to fit the current viewport, a semantic anchoring module that stores a target text string and surrounding context text as anchor information when placing an object and determines the object position based on the stored target text string when rendering the PDF, and performs noise filtering, coordinate normalization, and context binding on the collected raw logs. It includes a semantic data transformation engine that performs operations and outputs structured data, an automatic archiving manager that automatically generates content based on structured data and links with SEO structured data, and an overlay map repository that stores document-specific object placement information.

[0010] In addition, the service provider server may further include a user behavior data collection unit that captures user behavior through an event listener and transmits it to the server, and the user behavior data collection unit may support both real-time transmission and batch transmission methods.

[0011] Additionally, the service provider server may further include an AI recommendation integration module that forms a feedback loop by configuring feature data based on user-specific behavior profiles and document-specific characteristic profiles and transmitting them to an AI recommendation engine, receiving recommendation results and reflecting them on the user screen, and re-collecting user behavior data regarding recommended content to improve the recommendation model.

[0012] Additionally, the service provider server may further include an interest area detection module that divides each PDF page into grid cells, aggregates behavioral data in each cell to calculate a weighted interest score, and merges adjacent Hotspot cells to visualize them as a heatmap.

[0013] Additionally, the semantic anchoring module can perform an exact match search using the stored target text string during PDF rendering, perform a similar match by applying edit distance and cosine similarity in parallel if an exact match fails, and perform a fallback function that temporarily places the data using the stored absolute coordinates if the similar match also fails.

[0014] In addition, the coordinate transformation engine converts PDF coordinates to HTML coordinates based on the formula P_html = M Х P_pdf + T, and can recalculate coordinates in real time when zooming in or out.

[0015] In addition, the automatic archiving manager operates via a document registration time trigger that automatically extracts document metadata upon new PDF upload to create an initial content record, and a data accumulation trigger that updates content when user response data reaches a certain threshold, and can automatically insert structured data in JSON-LD format into the generated content.

[0016] Additionally, the service provider server may further include a document replacement detection module that determines PDF replacement through file hash value comparison, automatically executes a semantic anchor re-search process based on an existing overlay map upon detection of replacement, and reports the relocation results to the administrator.

[0017] In addition, the service provider server may further include an administrator interface that includes an object placement editor, which allows selecting an object type from a palette and placing it on a PDF preview by dragging and dropping, and automatically detecting the text context of the location when placing it to automatically set a semantic anchor.

[0018] In addition, the virtual interaction layer control unit creates an independent transparent layer for each page of the PDF to activate only the layer of the currently displayed page, and can support both an immediate creation method upon page loading and a delayed creation method upon scrolling. Effects of the invention

[0019] According to the system for dynamic interaction object insertion and user behavior-based data automatic generation utilizing a web-based PDF rendering engine according to the present invention, it is possible to implement bidirectional interaction with the user by combining virtual interaction objects at the time of rendering while ensuring the integrity of the original PDF file.

[0020] In addition, the present invention enables device-independent object placement through text context-based semantic anchoring, and can reduce the management burden by performing automatic relocation even when documents are revised.

[0021] In addition, the present invention can convert static PDF documents into dynamic data assets by automating the entire process from the collection of user behavior data to standardization, archiving, SEO integration, and AI recommendation.

[0022] In addition, the present invention can automatically detect hotspots based on user behavior data and continuously improve recommendation accuracy by forming a feedback loop for AI recommendation results. Brief explanation of the drawing

[0023] FIG. 1 is a block diagram showing the overall configuration of a system according to one embodiment of the present invention. FIG. 2 is a conceptual diagram showing the structure of a virtual interaction layer according to one embodiment. FIG. 3 is a flowchart illustrating a semantic anchoring process according to one embodiment. FIG. 4 is a flowchart illustrating a conversion pipeline from unstructured data to structured data according to one embodiment. FIG. 5 is a flowchart illustrating an automatic archiving and SEO conversion process according to one embodiment. FIG. 6 is a block diagram showing an AI recommendation linkage and feedback loop according to one embodiment. FIG. 7 is a flowchart illustrating a hotspot detection process according to one embodiment. FIG. 8 is a flowchart showing the flow of a main process according to one embodiment. FIG. 9 is a flowchart illustrating a document replacement detection and automatic relocation process according to one embodiment. Specific details for implementing the invention

[0024] Embodiments of the present invention will be described below with reference to the attached drawings. Unless otherwise specifically defined or mentioned, terms indicating direction used in this description are based on the state shown in the drawings. Additionally, throughout each embodiment, the same reference numeral indicates the same component. Meanwhile, the thickness or dimensions of each component shown in the drawings may be exaggerated for convenience of explanation and do not imply that it must actually be constructed according to the corresponding dimensions or ratios between components.

[0026] Referring to FIG. 1, the overall configuration of a system according to one embodiment of the present invention will be described. FIG. 1 is a block diagram showing the overall configuration of a system according to one embodiment of the present invention.

[0027] A dynamic interaction object insertion and user behavior-based data automatic generation system utilizing a web-based PDF rendering engine according to the present invention includes a user terminal (10), a communication network (20), a database (30), an access control unit (40), and a service providing server (100).

[0028] A user terminal (10) is a client device that accesses the service through a web browser and may include a PC, mobile phone, tablet, etc. The user terminal (10) transmits and receives data with a service provider server (100) through a communication network (20).

[0029] The communication network (20) is a network for transmitting and receiving data between a user terminal (10) and a service provider server (100).

[0030] The database (30) is a storage that stores user behavior data, statistical data, document metadata and overlay maps, etc.

[0031] The access control unit (40) performs role-based authority management, data collection consent processing, data encryption, and anonymization. The access control unit (40) separates the authority of general users and administrators by applying role-based access control (RBAC). Additionally, the access control unit (40) displays a data collection consent interface before collecting user behavior data, enables or disables collection based on consent, and records the consent history. HTTPS / TLS-based encryption is applied during data transmission, and anonymization processing is performed during statistical aggregation.

[0032] The service providing server (100) is a main server system that provides web-based PDF dynamic interaction and data archiving services. The service providing server (100) includes a PDF rendering control module (110), a virtual interaction layer control unit (120), a coordinate transformation engine (121), a semantic anchoring module (122), a user behavior data collection unit (130), a semantic data transformation engine (140), an automatic archiving manager (150), an AI recommendation integration module (160), a hotspot detection module (170), a statistical aggregation processing unit (180), an administrator interface (190), an overlay map storage (200), and a document replacement detection module (210).

[0033] Below, each component of the service provision server (100) is described in detail.

[0034] The PDF rendering control module (110) is a server-side module that controls client-side PDF.js-based rendering. When the PDF rendering control module (110) receives a request for a PDF document from a user terminal (10), it checks whether an overlay map associated with the PDF exists. If an overlay map exists, it transmits the overlay map data along with the PDF file to the user terminal (10).

[0035] A virtual interaction layer control unit (120) will be described with reference to FIG. 2. FIG. 2 is a conceptual diagram showing the structure of a virtual interaction layer according to one embodiment.

[0036] The virtual interaction layer control unit (120) performs transparent layer creation, object placement, and page-by-page layer activation and deactivation management. The virtual interaction layer control unit (120) includes a coordinate transformation engine (121) and a semantic anchoring module (122).

[0037] The virtual interaction layer control unit (120) automatically creates a transparent HTML div layer with the same size and position as the canvas at the top (z-index) of the canvas after PDF.js finishes rendering the PDF onto the canvas. The layer creation method supports both an immediate creation method when the page loads and a lazy loading method when the page is reached during scrolling.

[0038] The virtual interaction layer control unit (120) places each object (button, form, quiz, etc.) defined in the overlay map onto the transparent layer according to the coordinates calculated by the coordinate transformation engine (121). Each object is rendered as an HTML5 element, and the coordinates are recalculated to synchronize the position when PDF pages are switched or zoomed in / out.

[0039] The virtual interaction layer control unit (120) creates an independent transparent layer for each page of the PDF, activates only the layer of the currently displayed page, and deactivates the others.

[0040] The coordinate transformation engine (121) transforms PDF coordinates into HTML coordinates based on the formula P_html = M Х P_pdf + T. Here, P_html is the coordinate in HTML, M is the transformation matrix, P_pdf is the original PDF coordinate, and T is the translation vector. The coordinate transformation engine (121) recalculates the coordinates in real time when zooming in or out to maintain the accuracy of the object's position.

[0041] Referring to FIG. 3, the semantic anchoring module (122) will be described. FIG. 3 is a flowchart showing a semantic anchoring process according to one embodiment.

[0042] The semantic anchoring module (122) performs text context-based anchor setting, exact match matching, similar match matching, and fallback re-search.

[0043] In the anchor setting step, when the administrator places an object, the semantic anchoring module (122) stores the surrounding context text (N characters before and after) along with the target text string at the location as anchor information. At the same time, the page number and absolute coordinates (for fallback) of the text are also stored.

[0044] In the normal matching step, when rendering the PDF, the semantic anchoring module (122) extracts the entire text using the getTextContent() API and then performs an exact match search using the stored target text string. If a match is found, the object is positioned based on the bounding box coordinates of the text.

[0045] In the similarity matching step, if a perfect match fails, the semantic anchoring module (122) performs a similarity comparison using the stored surrounding context text. The similarity comparison algorithms include Levenshtein Distance and Cosine Similarity. If the similarity is greater than or equal to a set threshold, it is automatically relocated to the corresponding location; if it is less than the threshold, it is marked as a placement failure and a manual relocation notification is sent to the administrator.

[0046] In the final fallback stage, if all text matching fails, the semantic anchoring module (122) temporarily places it at the existing location using the stored absolute coordinates (page number and coordinates).

[0047] Referring to FIG. 4, the user behavior data collection unit (130) and the semantic data conversion engine (140) will be described. FIG. 4 is a flowchart showing a conversion pipeline from unstructured data to structured data according to one embodiment.

[0048] The user behavior data collection unit (130) captures user behavior (click, scroll, hover, text selection, form input, etc.) through an event listener and transmits it to the server. The transmission method supports both real-time transmission (WebSocket) and batch transmission (periodic or event count).

[0049] The semantic data transformation engine (140) is a pipeline that performs noise filtering, coordinate normalization, context binding, and structured data output on the collected raw logs.

[0050] In the noise filtering step, the semantic data conversion engine (140) removes meaningless repetitive events (multiple clicks on the same coordinate within a short period of time, bot / auto scroll patterns). It also filters out events below a minimum dwell time threshold (e.g., hovers of less than 100ms) by determining them to be meaningless.

[0051] In the coordinate normalization step, the semantic data conversion engine (140) converts screen pixel coordinates into percentage-relative coordinates relative to the page to ensure device independence. It also inversely calculates the current zoom level (scale) and converts it into the original PDF coordinate system.

[0052] In the context binding step, the semantic data conversion engine (140) maps the PDF text context (sentence or paragraph at the corresponding location) corresponding to the event occurrence coordinates. If it is an interaction with an interaction object, the object ID and type are linked as metadata.

[0053] In the structured data output step, the semantic data conversion engine (140) structures and outputs the data according to the final JSON schema (event_id, timestamp, interaction_type, position, dwell_time, context_metadata).

[0054] Referring to FIG. 5, an automatic archiving manager (150) will be described. FIG. 5 is a flowchart illustrating an automatic archiving and SEO conversion process according to one embodiment.

[0055] The automatic archiving manager (150) performs trigger-based automatic content generation, duplicate prevention, and SEO structured data (JSON-LD) integration.

[0056] The automatic archiving manager (150) operates via a document registration time trigger and a data accumulation trigger. When a new PDF is uploaded to the system, the document registration time trigger automatically extracts document metadata (title, author, number of pages, keywords, etc.) to create an initial content record. The data accumulation trigger updates the content when user response data reaches a certain threshold (number of interactions or time cycle).

[0057] When content is automatically generated, the automatic archiving manager (150) combines document metadata, extracted text body, and accumulated user response data. It automatically converts the content into HTML, JSON-LD, or Markdown format depending on the bulletin board type of the target CMS. It automatically extracts sections of high user interest (using Hotspot data) as summary highlights and includes them in the content. It integrates universally with various CMSs based on API.

[0058] To prevent duplication, the automatic archiving manager (150) processes existing records as updates rather than new creations if content for the same PDF document already exists. It determines whether the documents are identical by comparing document hash values ​​and records the history of content changes by version.

[0059] To integrate SEO structured data, the automatic archiving manager (150) automatically inserts structured data in JSON-LD format (based on Schema.org) into the generated content. It outputs the document title, author, topic, and user response statistics (popularity range, views, etc.) as metadata that can be recognized by search engines.

[0060] Referring to FIG. 6, an AI recommendation integration module (160) will be described. FIG. 6 is a block diagram showing an AI recommendation integration and feedback loop according to one embodiment.

[0061] The AI ​​recommendation integration module (160) performs feature data configuration and transmission, recommendation result reception and application, and feedback loop. The AI ​​recommendation integration module (160) supports both internal AI module integration and external AI service integration.

[0062] When configuring feature data, the AI ​​recommendation integration module (160) aggregates user-specific behavior profiles (list of viewed documents, time spent per page, interaction object response history, interest area patterns). It also configures document-specific characteristic profiles (document metadata, popular ranges aggregated for all users, and object response statistics).

[0063] The AI ​​recommendation integration module (160) transmits feature data to the AI ​​recommendation engine via a standardized REST API. The transmitted data format is a JSON structure containing a user ID, a document ID, and a feature vector (quantified behavioral data).

[0064] The AI ​​recommendation linkage module (160) receives recommendation results (list of recommended documents, recommended objects, related content, etc.) from the AI ​​engine and reflects the received recommendation results in real time on the user's current viewing screen. For example, recommended content can be displayed on a side panel or interaction layer within the PDF viewer.

[0065] For the feedback loop, the AI ​​recommendation integration module (160) collects actions taken by the user regarding the recommended content (click, ignore, additional viewing, etc.) as behavior data. The collected feedback data is transmitted back to the AI ​​engine to form a cyclic structure that continuously improves the accuracy of the recommendation model.

[0066] Referring to FIG. 7, a region of interest (Hotspot) detection module (170) will be described. FIG. 7 is a flowchart illustrating a region of interest (Hotspot) detection process according to one embodiment.

[0067] The hotspot detection module (170) performs grid segmentation, calculation of weighted interest scores, merging of hotspot areas, and heatmap visualization.

[0068] The area of ​​interest (Hotspot) detection module (170) collects click coordinates and frequency, hover position and dwell time, scroll stop interval and stop time, text selection area, interaction object response status and elapsed time.

[0069] When dividing the area, the hotspot detection module (170) divides each page of the PDF into grid cells of a certain size. It aggregates behavioral data corresponding to each cell to calculate an interest score for each cell.

[0070] When calculating the interest score, the interest area (Hotspot) detection module (170) assigns weights to each type of behavior. For example, a click may have the highest weight, followed by text selection, scroll stop, and hover. A weighted sum score is calculated for each cell, and cells that exceed a threshold (more than N times the average) relative to the overall page average are determined to be Hotspots.

[0071] The area of ​​interest (Hotspot) detection module (170) merges adjacent Hotspot cells into a single Hotspot area and visualizes the merged area in the form of a heatmap, providing it to an administrator-only dashboard.

[0072] Hotspot data generated by the interest area (Hotspot) detection module (170) is transmitted as a feature to the AI ​​recommendation integration module (160). Additionally, it is used to automatically extract text from the corresponding area as a summary highlight during automatic archiving. It can suggest the placement of new interaction objects in the Hotspot area to the administrator.

[0073] The statistical aggregation processing unit (180) performs periodic batch aggregation at the document, page, and object levels. At the document level, it aggregates behavioral data of all users for a specific PDF document. At the page level, it calculates the number of views per page, the average time spent, and the bounce rate. At the object level, it calculates the click-through rate, response rate, and response distribution statistics for each interaction object.

[0074] The statistical aggregation processing unit (180) performs periodic batch aggregation (on an hourly, daily, or weekly basis) on standardized individual log data. The aggregation results are stored in a separate statistical table and managed separately from the raw logs. The output metrics include page views, average and median time spent, bounce rate, click-through rate (CTR) relative to impressions per object, response completion rate, interest trends (time series) by Hotspot area, and comparison of behavioral patterns by user segment.

[0075] The administrator interface (190) includes an object placement editor (drag and drop), a statistics dashboard, and a relocation report approval function.

[0076] In the object placement editor of the administrator interface (190), when the administrator logs into the web-based administration interface and selects a target PDF document, the selected PDF is rendered into PDF.js and displayed on the preview screen. The administrator selects an object type (button, input form, quiz, link, media, etc.) from the palette and places the selected object page by page by dragging and dropping it to a desired location on the PDF preview. When placing, the text context of the location is automatically detected and a semantic anchor is automatically set.

[0077] Selecting a placed object displays the properties editing panel. Configurable items include object size, display conditions (specific user groups, time zones, etc.), and connection behaviors (URL navigation, data submission, quiz answer verification, etc.). The behavior can be verified from the user's perspective in preview mode. The placement information (coordinates, anchor text, object type, attributes) of all configured objects is saved to the server as overlay map data (JSON).

[0078] The statistics dashboard of the administrator interface (190) provides aggregated statistics as visualization dashboards (charts, heatmaps, tables). It supports filtering by period, and drill-down viewing by document, page, and object.

[0079] The overlay map storage (200) stores document-specific object placement information (JSON) and performs multiple map operations and version management.

[0080] The overlay map stored in the overlay map storage (200) consists of a list of objects per page. Each object record includes an object unique ID, a target page number, placement coordinates (percentage-based relative coordinates), semantic anchor information (target text, surrounding context text, and absolute coordinate fallback), object type, object attributes, and creation / modification time.

[0081] The overlay map storage (200) stores a unique identifier (file hash value) of a PDF document and an overlay map by mapping them. Multiple overlay map versions can be maintained for the same PDF. Overlay map data is stored in a server-side database in JSON format, and upon a client request, the latest overlay map of the corresponding document is delivered via an API.

[0082] Referring to FIG. 9, a document replacement detection module (210) will be described. FIG. 9 is a flowchart illustrating a document replacement detection and automatic relocation process according to one embodiment.

[0083] The document replacement detection module (210) determines PDF replacement through file hash value comparison and triggers an automatic relocation process.

[0084] In the document replacement detection step, if an administrator replaces and uploads an existing PDF with a new version of the PDF, the document replacement detection module (210) automatically determines whether the replacement has been made by comparing file hash values. Unless the original PDF is replaced, the overlay map remains unchanged.

[0085] In the automatic relocation execution step, when a replacement is detected, the document replacement detection module (210) automatically executes a semantic anchor re-search process in batches based on the existing overlay map. Depending on the matching result, it is classified into automatic relocation (exact match or similar match), fallback temporary placement, and matching failure.

[0086] In the relocation result report stage, the document replacement detection module (210) reports to the administrator the status of automatic relocation success and failure relative to the total objects, as well as the matching type and similarity score for each object. It displays objects that failed to match or were similarly matched as a list requiring manual verification, and activates them when the administrator verifies and approves them in the editor.

[0087] Referring to FIG. 8, the main process flow of the system according to the present invention will be described. FIG. 8 is a flowchart showing the flow of the main process according to one embodiment.

[0088] The main process includes a service provision and rendering control step (S100), a virtual layer synchronization and object placement step (S200), a user behavior collection and data structuring step (S300), and an archiving, statistics, and AI recommendation integration step (S400).

[0089] In the service provision and rendering control step (S100), the user terminal (10) requests a PDF document from the service provision server (100) via the communication network (20). The service provision server (100) transmits the PDF file and overlay map data to the client via the PDF rendering control module (110). On the client side, the PDF is rendered on a canvas based on PDF.js.

[0090] In the virtual layer synchronization and object placement step (S200), the virtual interaction layer control unit (120) creates a transparent layer on top of the rendered canvas. The coordinate transformation engine (121) transforms the coordinate data of the overlay map to fit the current viewport. The semantic anchoring module (122) determines and places the object position based on the text context.

[0091] In the user behavior collection and data structuring step (S300), the user behavior data collection unit (130) captures the user's interaction in real time and transmits it to the server. The semantic data conversion engine (140) performs noise filtering, coordinate normalization, and context binding to output structured JSON data.

[0092] In the archiving, statistics, and AI recommendation integration stage (S400), the automatic archiving manager (150) automatically generates CMS content and integrates SEO structured data based on structured data. The statistics aggregation processing unit (180) periodically aggregates multi-user data and provides it to the administrator dashboard. The area of ​​interest (Hotspot) detection module (170) analyzes the aggregated data to determine Hotspots and generates a heatmap. The AI ​​recommendation integration module (160) transmits feature data to the AI ​​engine, receives recommendation results and reflects them on the user screen, and forms a feedback loop.

[0094] Although preferred embodiments of the present invention have been described above, the technical concept of the present invention is not limited to the preferred embodiments described above and can be implemented in various ways within the scope that does not depart from the technical concept of the present invention as embodied in the claims. Explanation of the symbols

[0096] 10: User terminal 20: Communication Network 30: Database 40: Access control unit 100: Service provision server 110: PDF Rendering Control Module 120: Virtual Interaction Layer Control Unit 121: Coordinate transformation engine 122: Semantic Anchoring Module 130: User Behavior Data Collection Unit 140: Semantic Data Transformation Engine 150: Automatic Archiving Manager 160: AI Recommendation Integration Module 170: Hotspot Detection Module 180: Statistical Aggregation Processing Unit 190: Administrator Interface 200: Overlay Map Repository 210: Document Replacement Detection Module

Claims

Claim 1 A system for performing dynamic interaction object insertion and automatic generation of user behavior-based data for PDF documents in a web browser, comprising: a user terminal accessing a service through a web browser; a database storing user behavior data, document metadata, and an overlay map; A service provider server that provides web-based PDF dynamic interaction and data archiving services; wherein the service provider server includes: a PDF rendering control module that controls PDF.js-based rendering to deliver PDF files and overlay map data to a user terminal; a virtual interaction layer control unit that creates a transparent layer on top of the canvas with the same size and position as the canvas after PDF.js completes rendering the PDF onto the canvas, and places interaction objects defined in the overlay map on the transparent layer; a coordinate transformation engine that transforms the coordinate data of the overlay map to fit the current viewport; a semantic anchoring module that stores a target text string and surrounding context text as anchor information when placing an object, and determines the object position based on the stored target text string when rendering the PDF; a user behavior data collection unit that captures user behavior through an event listener and transmits it to a server; a semantic data transformation engine that performs noise filtering, coordinate normalization, and context binding on the raw logs collected by the user behavior data collection unit to output structured data; and, based on the structured data, automatically generates content for web publishing by combining document metadata, extracted text body, and accumulated user response data, and the generated Schema in content.A system for dynamic interaction object insertion and automatic user behavior-based data generation utilizing a web-based PDF rendering engine, comprising: an automatic archiving manager that automatically inserts SEO structured data in an org-based JSON-LD format; and an overlay map repository that stores object placement information per document; wherein the service providing server further includes an interest area detection module that divides each PDF page into grid cells, aggregates behavioral data in each cell to calculate a weighted interest score, and merges adjacent Hotspot cells to visualize them as a heatmap. Claim 2 In paragraph 1, the user behavior data collection unit is a system for dynamic interaction object insertion and automatic user behavior-based data generation utilizing a web-based PDF rendering engine that supports both real-time transmission and batch transmission methods. Claim 3 A system for dynamic interaction object insertion and automatic user behavior-based data generation utilizing a web-based PDF rendering engine, wherein the service providing server further comprises an AI recommendation integration module that forms a feedback loop to improve the recommendation model by re-collecting behavioral data regarding the user's recommended content, and wherein the service providing server configures feature data based on user-specific behavior profiles and document-specific characteristic profiles and transmits them to an AI recommendation engine, receives recommendation results and reflects them on the user screen. Claim 4 delete Claim 5 A system for inserting dynamic interaction objects and automatically generating user behavior-based data utilizing a web-based PDF rendering engine, wherein the semantic anchoring module performs an exact match search using a stored target text string during PDF rendering, performs a similarity matching by applying edit distance and cosine similarity in parallel when an exact match fails, and performs a fallback function of temporarily placing using stored absolute coordinates when similarity matching also fails. Claim 6 In claim 1, the coordinate transformation engine converts PDF coordinates to HTML coordinates based on the formula P_html = M Х P_pdf + T (where P_html is a coordinate in HTML, M is a transformation matrix, P_pdf is a PDF source coordinate, and T is a translation vector), and recalculates coordinates in real time when zooming in or out, a system for inserting dynamic interaction objects and automatically generating user behavior-based data utilizing a web-based PDF rendering engine. Claim 7 In claim 1, the automatic archiving manager operates by a document registration time trigger that automatically extracts document metadata upon new PDF upload to create an initial content record, and a data accumulation trigger that updates content when user response data reaches a certain standard, and the SEO structured data in the JSON-LD format includes document title, author, subject, and user response statistics as metadata recognizable by search engines, thereby creating a dynamic interaction object insertion and user behavior-based data automatic generation system utilizing a web-based PDF rendering engine. Claim 8 A system for dynamic interaction object insertion and automatic data generation based on user behavior utilizing a web-based PDF rendering engine, wherein the service providing server further includes a document replacement detection module that determines PDF replacement through file hash value comparison and, upon detection of replacement, automatically executes a semantic anchor re-search process based on an existing overlay map to report the relocation result to an administrator. Claim 9 A system for inserting dynamic interaction objects and automatically generating user behavior-based data utilizing a web-based PDF rendering engine, wherein the service providing server further includes an administrator interface comprising an object placement editor that selects an object type from a palette and places it on a PDF preview by dragging and dropping, and automatically detects the text context of the location at the time of placement and automatically sets a semantic anchor. Claim 10 In claim 1, the virtual interaction layer control unit creates an independent transparent layer for each page of the PDF and activates only the layer of the currently displayed page, and is a dynamic interaction object insertion and user behavior-based data automatic generation system utilizing a web-based PDF rendering engine that supports both an immediate generation method upon page loading and a delayed generation method upon scrolling.

Citation Information

Patent Citations

  • Capture and display of annotations in paper and electronic documents

    KR1020090069300A

  • Annotation management in a pen-based computing system

    US20050055628A1