Agentic data science with the Data Agent Kit
cloud.google.com·7h ago
AI Research
arXiv CS.AI·Sep 3·5 min read
DeepAffinity: Long-Term Aspect Preference Prediction in eCommerce using Small Language Models
TL;DR
DeepAffinity is a novel approach for predicting eCommerce user preferences for product aspects like brand and size, utilizing Small Language Models (SLMs). It significantly enhances recommendation quality on a large-scale multinational platform by effectively capturing long-term user preferences from interaction history.
✦ Why It Matters
Implement DeepAffinity to improve your eCommerce platform's recommendation system today.
Key Takeaways
- Utilize DeepAffinity for better user preference predictions.
- Incorporate structured prompts in your SLMs for improved outcomes.
- Focus on long-term user interaction history for personalization.
- Avoid general-purpose LLMs without task-specific tuning.
How It Works
DeepAffinity leverages Small Language Models (SLMs) by using structured prompts that guide the model in understanding user preferences over time. It employs specialized prediction heads that are fine-tuned specifically for the Aspect Affinity task, allowing it to capture the nuances of user behavior and predict future choices based on historical data.