9 Personalized search

This chapter covers

The personalization spectrum between search and recommendations
Implementing collaborative filtering and personalization using latent features from users’ signals
Using embeddings to create personalization profiles
Multimodal personalization from content and behavior
Applying clustering-based personalization guardrails
Avoiding the pitfalls of personalized search

The better your search engine understands your users, the more likely it will be able to successfully interpret their queries. In chapter 1, we introduced the three key contexts needed to properly interpret query intent: content understanding, domain understanding, and user understanding. In this chapter, we’ll dive into the user understanding context.

We’ve already focused on learning domain-specific context from documents (chapter 5) and on the most popular results according to many different users (chapter 8), but it’s not always reasonable to assume that the “best” result is agreed upon across all users. Whereas signals-boosting models find the most popular answers across all users, personalized search instead attempts to learn about each specific user’s interests and to return search results catering to those interests.

9.1 Personalized search vs. recommendations

9.1.1 Personalized queries

9.1.2 User-guided recommendations

9.2 Recommendation algorithm approaches

9.2.1 Content-based recommenders

9.2.2 Behavior-based recommenders

9.2.3 Multimodal recommenders

9.3 Implementing collaborative filtering

9.3.1 Learning latent user and item features through matrix factorization

9.3.2 Implementing collaborative filtering with Alternating Least Squares

9.3.3 Personalizing search results with recommendation boosting

9.4 Personalizing search using content-based embeddings

9.4.1 Generating content-based latent features

Summary