Siamese Wide-and-Deep Similarity Ranking for Interface Elements
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current interface systems rely exclusively on positive interactions for similar element recommendation, leading to low accuracy and relevancy in identifying similar interface elements.
Innovation Solution
Implementing a Siamese wide and deep neural network framework to generate a similarity score for candidate elements using a pair-wise ranking model, incorporating both wide and deep features, and transmitting an interface with selected similar elements to a user device.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current similar element recommendation processes are used, then the system can identify some similar elements, but the accuracy and relevancy of recommended elements is low
Solution Approach 1:
The patent transitions from traditional single-dimensional recommendation approaches to a multi-dimensional framework by integrating wide features (explicit user-item interactions) and deep features (implicit behavioral patterns) into a unified Siamese neural network. This dimensional expansion enables the system to capture both direct and indirect similarity signals, significantly improving identification accuracy and recommendation relevancy
Solution Approach 2:
The patent creates a composite recommendation system by combining multiple feature types (wide and deep features) and multiple network components (Siamese twin networks, ranking modules) into a unified architecture. This composite approach allows the system to leverage complementary strengths of different feature sources and processing methods, resolving the accuracy-relevancy tradeoff
2Device complexity
If exclusively positive interactions are used for recommendation, then the system structure remains simple, but the accuracy and relevancy of recommendations deteriorates
Solution Approach 1:
The patent segments the recommendation system into distinct functional components: a Siamese twin network for feature extraction, a ranking module for similarity scoring, and a retrieval module for candidate generation. This segmentation allows each component to specialize in processing specific feature types (positive and negative interactions separately), improving overall accuracy while maintaining manageable system complexity through modular architecture
Solution Approach 2:
The patent introduces an intermediary ranking module that mediates between the Siamese network's similarity scores and the final recommendation output. This intermediary component processes both wide and deep features through a unified framework, reconciling the complexity of multi-feature processing with the need for accurate recommendations by providing a centralized scoring mechanism
Data Source
AI summary
In various embodiments, systems and methods for generating interfaces including similar elements are disclosed. An interface request identifying an anchor element is received and a set of similar elements for the anchor element identifier is generated by implementing an inference recommendation model generated by a Siamese wide and deep training framework. The inference recommendation model is configured to receive at least one recall set of candidate elements and generate a similarity score for each candidate element in the set of candidate elements and the anchor element. An interface including at least one similar element selected from the set of similar elements is generated and transmitted to a user device associated with the interface request.