Results of the summarizing process for the arXiv paper: 2402.16073v2

In personalized recommender systems, embeddings are commonly utilized to encode customer actions and items. Retrieval is then conducted in the embedding space through approximate nearest neighbor search. However, this approach presents challenges such as potential restrictions on the diversity of captured interests within user embeddings and the necessity for costly real-time infrastructure to maintain their up-to-date status. To address these issues in a practical industrial setting, a method was proposed that dynamically updates customer profiles and generates personalized feeds every two minutes by leveraging precomputed embeddings and their corresponding similarities. This method was implemented and tested at Bol, a prominent e-commerce platform serving the Netherlands and Belgium, to personalize promotional items. The results were promising with enhanced customer engagement and experience leading to a notable 4.9% uplift in conversions. The process involved utilizing historical customer interactions data alongside an encoder model to generate query embeddings for items in the catalogue. These query embeddings were then matched against item embeddings stored in an index vector database using approximate nearest neighbor search (ANNS) to retrieve relevant recommendations. Furthermore, the system incorporated elements of ranking algorithms, diversity considerations, business rules integration, and merging recommendations based on recent customer interactions. By combining these components effectively within a near real-time framework that updates every two minutes, the method successfully delivered tailored feeds to users while addressing concerns related to maintaining diverse user interests and minimizing infrastructure costs associated with real-time updates. Authors Binyam Gebre, Karoliina Ranta, Stef van den Elzen, Ernst Kuiper, Thijs Baars, and Tom Heskes present this innovative approach in their paper titled "Pfeed: Generating near real-time personalized feeds using precomputed embedding similarities," offering valuable insights for enhancing personalization strategies in e-commerce platforms through efficient use of embedding techniques and dynamic feed generation processes.

• - Personalized recommender systems commonly use embeddings to encode customer actions and items
• - Retrieval is done in the embedding space through approximate nearest neighbor search
• - Challenges include restrictions on diversity within user embeddings and costly real-time infrastructure maintenance
• - A method was proposed to address these challenges by dynamically updating customer profiles and generating personalized feeds every two minutes using precomputed embeddings
• - The method was implemented at Bol, an e-commerce platform, resulting in enhanced customer engagement and a 4.9% uplift in conversions
• - Historical customer interactions data is used with an encoder model to generate query embeddings for items, which are matched against item embeddings using ANNS for recommendations
• - The system incorporates ranking algorithms, diversity considerations, business rules integration, and merging recommendations based on recent interactions
• - By updating near real-time every two minutes, the method successfully delivers tailored feeds while addressing concerns about diverse user interests and minimizing infrastructure costs

Summary1. Recommender systems help suggest things you might like based on what you've done before. 2. They find these suggestions by looking at a special way of representing your actions and the things you like. 3. Sometimes it's hard to make sure the suggestions are different enough and don't cost too much to do quickly. 4. A new way was made to solve these problems by updating your profile often and making personalized suggestions every two minutes. 5. This new way was used in a shop called Bol and made customers more interested, leading to more people buying things. Definitions- Personalized: Made just for you - Recommender systems: Tools that suggest things you might like - Embeddings: Special codes that represent actions or items - Nearest neighbor search: Finding similar things - Infrastructure maintenance: Taking care of all the technology needed - Precomputed: Figured out in advance - Conversions: People buying things after seeing them - Interactions data: Information about what people do online

Personalization has become a crucial aspect of e-commerce platforms, with the aim of providing tailored recommendations to customers based on their interests and preferences. In order to achieve this, personalized recommender systems often utilize embeddings to encode customer actions and items. However, this approach presents challenges such as limited diversity in captured interests within user embeddings and the need for costly real-time infrastructure updates. To address these issues in a practical industrial setting, a team of researchers from Bol, a prominent e-commerce platform serving the Netherlands and Belgium, proposed a method that dynamically updates customer profiles and generates personalized feeds every two minutes by leveraging precomputed embeddings and their corresponding similarities. This innovative approach was presented in their paper titled "Pfeed: Generating near real-time personalized feeds using precomputed embedding similarities." The research team consisted of Binyam Gebre, Karoliina Ranta, Stef van den Elzen, Ernst Kuiper, Thijs Baars, and Tom Heskes. Their work offers valuable insights into enhancing personalization strategies in e-commerce platforms through efficient use of embedding techniques and dynamic feed generation processes. The Methodology The proposed method involved utilizing historical customer interactions data alongside an encoder model to generate query embeddings for items in the catalogue. These query embeddings were then matched against item embeddings stored in an index vector database using approximate nearest neighbor search (ANNS) to retrieve relevant recommendations. In addition to ANNS, the system also incorporated elements of ranking algorithms, diversity considerations, business rules integration, and merging recommendations based on recent customer interactions. By combining these components effectively within a near real-time framework that updates every two minutes, the method successfully delivered tailored feeds to users while addressing concerns related to maintaining diverse user interests and minimizing infrastructure costs associated with real-time updates. Implementation at Bol To test the effectiveness of their method in a practical industrial setting, the research team implemented it at Bol – one of Europe's largest online retailers. The platform serves millions of customers in the Netherlands and Belgium, making it an ideal testing ground for their approach. The team used a dataset consisting of customer interactions with items on the Bol platform over a period of three months. This data was used to train an encoder model that generated query embeddings for each item in the catalogue. These embeddings were then stored in an index vector database along with their corresponding similarities. Results and Impact The results of implementing this method at Bol were promising, with enhanced customer engagement and experience leading to a notable 4.9% uplift in conversions. The dynamic feed generation process every two minutes ensured that customers received personalized recommendations based on their recent interactions, resulting in increased user satisfaction and retention. Moreover, by leveraging precomputed embeddings and ANNS, the system was able to efficiently retrieve relevant recommendations without the need for costly real-time infrastructure updates. This helped reduce operational costs for Bol while still providing high-quality personalized feeds to its customers. Conclusion In conclusion, the paper "Pfeed: Generating near real-time personalized feeds using precomputed embedding similarities" presents a practical solution for addressing challenges faced by personalized recommender systems such as limited diversity within user embeddings and costly real-time infrastructure updates. By leveraging precomputed embeddings and ANNS within a dynamic feed generation process every two minutes, this method successfully delivered tailored feeds to users while improving customer engagement and conversion rates at Bol – one of Europe's largest e-commerce platforms.