Engineering Serendipity through Recommendations of Items with Atypical Aspects

TL;DR

ATARS introduces a three‑stage pipeline to engineer serendipity in recommendations by extracting atypical aspects from reviews, mapping them to user‑dependent utility, and re‑ranking items by a serendipity score that blends surprise with relevance. It leverages large language models for both extraction (ExtrAtA) and utility estimation, aided by dynamic in‑context learning to improve judgments, and it validates the approach with new datasets (ExtrAtA and synthetic profiles) and end‑to‑end ranking experiments. Results show serendipity‑based rankings correlate strongly with ground truth and outperform traditional star‑based rankings in capturing user satisfaction, supporting a shift toward serendipity as a core objective. The work provides a flexible framework and publicly available resources to spur further research on recommendation strategies that balance accuracy with meaningful, surprising user experiences.

Abstract

A restaurant dinner or a hotel stay may lead to memorable experiences when guests encounter unexpected aspects that also match their interests. For example, an origami-making station in the waiting area of a restaurant may be both surprising and enjoyable for a customer who is passionate about paper crafts. Similarly, an exhibit of 18th century harpsichords would be atypical for a hotel lobby and likely pique the interest of a guest who has a passion for Baroque music. Motivated by this insight, in this paper we introduce the new task of engineering serendipity through recommendations of items with atypical aspects. We describe an LLM-based system pipeline that extracts atypical aspects from item reviews, then estimates and aggregates their user-specific utility in a measure of serendipity potential that is used to rerank a list of items recommended to the user. To facilitate system development and evaluation, we introduce a dataset of Yelp reviews that are manually annotated with atypical aspects and a dataset of artificially generated user profiles, together with crowdsourced annotations of user-aspect utility values. Furthermore, we introduce a custom procedure for dynamic selection of in-context learning examples, which is shown to improve LLM-based judgments of atypicality and utility. Experimental evaluations show that serendipity-based rankings generated by the system are highly correlated with ground truth rankings for which serendipity scores are computed from manual annotations of atypical aspects and their user-dependent utility. Overall, we hope that the new recommendation task and the associated system presented in this paper catalyze further research into recommendation approaches that go beyond accuracy in their pursuit of enhanced user satisfaction. The datasets and the code are made publicly available at https://github.com/ramituncc49er/ATARS .

Figures (10)

• Figure 1: Example interaction with the proposed recommender system, where items with atypical aspects relevant to the user are promoted at the top of the ranking ( before), in order to engineer a serendipitous experience ( after).
• Figure 2: Example of the 3 steps of the Atypical Aspect-Based Recommender System (ATARS), where the left pane shows the user query and the user profile, given as input to the system. Middle pane shows the original star-based ranking of a sample of cafes, together with a sample of reviews showing atypical aspects extracted in Step 1 and their user-dependent utilities extracted at Step 2 (between brackets). Right pane shows the re-ranking done based on the serendipity scores computed at Step 3 (shown on the connecting arrows).
• Figure 3: ATARS system architecture.
• Figure 4: ATARS system architecture: Atypical Aspect Extraction (ExtrAtA).
• Figure 5: ATARS system architecture: Utility Classification.