Trend-Based Joint Embeddings for Dynamic Item Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models trained on historical data are not well-suited for industries with rapidly changing trends, such as fashion, as they fail to account for current trends and user preferences, leading to irrelevant suggestions for user interfaces like online catalogs.
Innovation Solution
A trend-based joint embedding model is fine-tuned using data from multiple modalities, including social media, sales data, and expert opinions, to enhance user selection by identifying compatible items through counterfactual queries and natural language conversations, balancing historical associations with current trends and user attributes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If machine learning models are trained only on historical data, then they can learn associations between items, but they fail to account for current trends and user preferences, leading to irrelevant suggestions
Solution Approach 1:
The patent combines historical data with current trend data from multiple sources (social media, news, expert opinions) to train the machine learning model. This merging of data sources allows the model to learn both stable historical associations and current trending patterns, resolving the contradiction between adaptability to trends and reliability of suggestions.
Solution Approach 2:
The system dynamically updates the model training data by continuously incorporating current trend data from multiple modalities. This dynamic approach allows the model to adapt to changing trends while maintaining the foundation of historical associations, thereby improving suggestion relevance without sacrificing trend adaptability.
2Reliability
If multiple data sources in multiple modalities are used to fine-tune the model, then current trends and user preferences are captured, but the system complexity increases
Solution Approach 1:
The patent introduces an intermediary processing layer that aggregates and harmonizes data from multiple sources and modalities before feeding it to the model. This intermediary structure manages the complexity of integrating diverse data sources while enabling the model to capture current trends and user preferences effectively.
3Adaptability or versatility
If the model is fine-tuned based on current trend data, then it can provide trendy suggestions, but it may lose the learned associations from historical data
Solution Approach 1:
The system performs preliminary training on historical data to establish fundamental item associations before fine-tuning with current trend data. This preliminary action ensures that the model retains learned historical associations while adapting to current trends through subsequent fine-tuning, resolving the contradiction between trend adaptability and association accuracy.
Data Source
AI summary
Methods, systems, and computer program products for facilitating user selection using trend-based joint embeddings are provided herein. A method includes obtaining a selection of an item in an online catalog; determining a compatible item of the plurality of items at least in part by providing the selected at least one item and at least one previously selected item corresponding to the user to a trend-based machine learning model, wherein the trend-based machine learning model is trained on historical data associated with the item in the online catalog and fine-tuned based on current trend data from multiple data sources; receiving feedback in response to outputting the at least one compatible item; identifying one or more attributes related to the at least one compatible item based on the feedback; and using the trend-based machine learning model to determine at least one additional compatible item based on the one or more attributes.