Livestream Carousel Widget Ranking Model for E-Commerce
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for generating livestream recommendations on e-commerce platforms are inefficient and costly due to the need to consider numerous factors related to user actions, products, and real-time livestream data, leading to processing inefficiencies and missed opportunities for users to engage with relevant livestreams.
Innovation Solution
A system and method for automatically generating livestream carousel widgets using machine learning and user action statistics, which retrieves candidate livestreams, organizes them based on user data, and inputs this information into a ranking model or neural network to determine personalized rankings for display on user interfaces.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional methods are used to generate livestream recommendations by considering numerous factors related to user actions, products, and real-time livestream data, then comprehensive recommendations can be provided, but processing inefficiencies and unnecessary costs occur
Solution Approach 1:
The patent segments the recommendation system into multiple independent components: a recall module that retrieves candidate livestreams based on user history and product data, and a ranking module that scores these candidates using a machine learning model. This segmentation allows each module to focus on specific tasks, improving overall processing efficiency while maintaining comprehensive recommendation quality.
Solution Approach 2:
The system performs preliminary actions by pre-processing user action data, product data, and livestream data into structured formats before the recommendation generation. Candidate livestreams are retrieved and organized in advance, and the machine learning model is pre-trained with historical data, enabling fast real-time recommendations without reprocessing all raw data during execution.
2Adaptability or versatility
If conventional methods process numerous factors related to user actions, products, and real-time livestream data, then personalized recommendations can be generated, but unnecessary costs are incurred
Solution Approach 1:
The patent extracts only the most relevant features from extensive user action data, product data, and livestream data. The recall module filters candidate livestreams based on key matching criteria, and the ranking model focuses on essential features like user viewing history, product categories, and livestream metadata, discarding redundant information to reduce computational costs while maintaining personalization.
Solution Approach 2:
The system changes parameters by transforming raw data into optimized feature representations suitable for machine learning processing. User actions are converted into structured event sequences, product attributes are encoded into categorical features, and the model dynamically adjusts weighting parameters based on data freshness and relevance, reducing computational complexity while preserving personalization accuracy.
3Loss of information
If all candidate livestreams are presented to users, then complete information is provided, but users feel overwhelmed and cannot decide which livestream to watch
Solution Approach 1:
The patent implements dynamic recommendation lists that adapt to user interactions in real-time. The ranking model continuously scores candidate livestreams based on current user behavior, and the system dynamically reorders and renews the recommendation list as users interact with content. This dynamic adjustment maintains information completeness by keeping relevant livestreams available while improving ease of operation by prioritizing the most relevant options at the top.
Solution Approach 2:
The system applies partial action by presenting a curated subset of the most relevant livestreams rather than all available options. The recall module retrieves a manageable number of candidates, and the ranking module scores and ranks them, displaying only the top recommendations. This partial presentation reduces user cognitive load and decision-making difficulty while still providing access to comprehensive information through the recommendation interface.
Data Source
AI summary
Computer-implemented systems and methods for automatic generation of livestream carousel widgets are disclosed and may be configured to retrieve a plurality of candidate livestreams based on first data related to one or more users and second data related to a plurality of livestreams, organize the retrieved plurality of candidate livestreams, input third data related to a first user and fourth data related to the retrieved plurality of candidate livestreams into a ranking model, output, from the ranking model, a value for each livestream of the organized plurality of candidate livestreams, based on the outputted value for each livestream, determine a rank for each livestream of the organized plurality of candidate livestreams, generate a livestream carousel widget including a number of candidate livestreams based on the determined rank, and transmit the generated livestream carousel widget for display on a user interface associated with the first user.