Interactive Content Node Recommendation via Embedding Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing technologies for creating interactive content require users to plan and produce every node, which is time-consuming, especially for large-scale content, as they lack efficient methods for recommending next nodes, leading to prolonged production times.
Innovation Solution
A method and system that calculates embedding values for specific nodes and other interactive content nodes, using a similarity calculation model to recommend next nodes based on similarity, allowing users to focus on generating content without manually planning each step.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If users manually plan and produce every node in interactive content using a builder tool, then the content can be fully customized and complete, but the production time becomes excessively long especially for large-scale content
Solution Approach 1:
The system pre-calculates and stores embedding values for all nodes in the database before they are needed for content creation. When a user creates interactive content, the system can quickly retrieve these pre-computed embeddings and perform similarity comparisons to recommend next nodes, eliminating the need for real-time computation and significantly reducing production time while maintaining content quality
Solution Approach 2:
The patent introduces embedding vectors as an intermediary representation between the textual/content features of nodes and the recommendation system. These embeddings serve as a mediator that enables efficient similarity comparison and node recommendation, allowing the system to suggest appropriate next nodes without requiring users to manually design every connection, thus reducing production time while preserving content completeness
2Quantity of substance
If users create large-scale interactive content with many nodes, then the content becomes more comprehensive and valuable, but the time and effort required for production increases significantly
Solution Approach 1:
The system enables semi-automated content creation where the recommendation engine automatically suggests next nodes based on similarity calculations, and users can accept these recommendations with minimal effort. This self-service approach allows users to create large-scale content with many nodes efficiently, as the system handles the complex task of node selection and connection planning, thereby maintaining high productivity while enabling comprehensive content creation
3Loss of time
If the system recommends next nodes based on similarity calculations using embedding values, then the production time is reduced, but the complexity of the recommendation system increases
Solution Approach 1:
The system transforms the complex task of node recommendation into a parameter-based similarity comparison problem using embedding vectors. By representing nodes as fixed-dimensional vectors and using established similarity metrics (cosine similarity, Euclidean distance), the system reduces the complexity of the recommendation logic while maintaining effectiveness. This parameter-based approach enables fast recommendation with reduced planning time, as the complex semantic understanding is captured in the embedding parameters
Data Source
AI summary
Disclosed are a method, a server and a computer-readable medium for recommending nodes of an interactive content, in which, when receiving recommendation request information for requesting a recommendation node for a specific node included in an interactive content from a user generating the interactive content, a first embedding value for a first set including the specific node is calculated, and a second embedding value for each second set including each of a plurality of nodes of each of one or more other interactive contents included in the service server is calculated, so as to calculate a similarity between the first embedding value and the second embedding value and provide the user with a next node, as a recommendation node, of a node corresponding to the second embedding value determined based on the similarity.