EXPEDIA GROUP TECHNOLOGY — DATA

Expedia Group @ RecSys 2021

Our recent scientific contributions to the Lodging recommendation domain

Published in

Expedia Group Technology

6 min readOct 21, 2021

At Expedia Group™️ we believe travel is a force for good and we power global travel for everyone, everywhere through our technology platform supporting the broadest offering in the travel industry.

Because of the scale of Expedia Group™️ — 2019 bookings were over $107 billion while serving hundreds of millions of travelers — very small percentage improvements are significant. As such, developing and employing the most advanced algorithms for our Recommendation Systems is key to best serving our travelers.

Below we introduce the three contributions in the field of Recommender Systems our Machine Learning team presented at ACM RecSys 2021. The conference was held in Amsterdam and Expedia Group™️ was proud to be a Silver sponsor.

Pictures by Jamie Fenn, Christina Morillo and Eva Darron from Unsplash.com under Unsplash license: Pictures of a landscape, a female developer in front of a computer and a plane to illustrate the Technology serving the Travel domain.

Expedia Group RecTour Research Dataset [paper]

RecTour (RECommenders in TOURism) RecSys 21 Workshop

Adam Woznica, Jan Krasnodebski

This document provides details on the dataset that Expedia Group™️ released to the RecTour community at the 15th ACM Conference on Recommender Systems. This dataset is based on real traveler lodging searches and bookings on Brand Expedia websites, which have been anonymized to protect the identities of consumers and suppliers. The intention is to provide the recommendation system research community, and more specifically travel researchers, an open and rich dataset for their work. The motivation for this dataset was multiple requests originating from Expedia Group™️ sponsored competitions, where participants wanted to use the data that was provided for research purposes.
This dataset was designed to meet that specific demand while preserving confidentiality.

Illustration of data collected through the UI: search parameters, applied filters, property information, clicks and bookings

Juggler: Multi-Stakeholder Ranking with Meta-Learning [paper]

MORS (Multi-Objective Recommender Systems) RecSys 21 Workshop

Tiago Cunha, Ioannis Partalas, Phong Nguyen

Online marketplaces must optimize recommendations with regards to multiple objectives, in order to fulfil expectations from a variety of stakeholders. This problem is typically addressed using Pareto Theory, which explores multiple objectives in a domain and identifies the objective vectors which yield the best performance. However, such an approach is computationally expensive, and available commonly only through domain-specific solutions, which is not ideal for online marketplaces and their ever-changing business dynamics.

We tackle these limitations by proposing a Meta-Learning framework to address the Multi-Stakeholder recommendation problem, which is able to dynamically predict the ideal settings on how business rules should be mingled into the final recommendations. The framework is designed to be generic enough to be leveraged in any item ranking domain and requires only the definition of a policy, i.e. a set of multi-objective metrics the meta-model should optimize for. The model finds the mapping between the search context and the corresponding best objective vectors. This way, the model is able to predict in real-time which is the best solution for any unforeseen search, and therefore adapt the recommendations on a search level. We show that under this framework, the range of models one is able to build depends only on how many policies can be defined, thus offering a virtually unlimited way to address multi-objective problems.

The experimental results showcase the generalization abilities of this framework and its highly predictive performance. Furthermore, the simulation results confirm the ability to approximate a policy’s expectation in most cases and hints at the potential to use this framework in many other item recommendation problems.

Diagram of the Juggler framework for training and inference phases. — Juggler framework

Hotel2Vec: Learning Hotel Embeddings from User Click Sessions with Side Information [paper]

RecTour (RECommenders in TOURism) RecSys 21 Workshop

Ioannis Partalas, Anne Morvan, Ali Sadeghian, Shervin Minaee, Xinxin Li, Brooke Cowan, Daisy Zhe Wang

We propose a new neural network architecture for learning vector representations of items with attributes, specifically hotels. Unlike previous works, which typically only rely on modeling of user-item interactions for learning item embeddings, we propose a framework that combines several sources of data, including user clicks, hotel attributes (e.g., property type, star rating, average user rating), amenity information (e.g., if the hotel has free Wi-Fi or free breakfast), and geographic information that leverages a hexagonal geospatial system as well as spatial encoders. During model training, a joint embedding is learned from all of the above information. We show that including structured attributes about hotels enables us to make better predictions in a downstream task than when we rely exclusively on click data. We train our embedding model on more than 60 million user click sessions from a leading online travel platform and learn embeddings for more than one million hotels. Our final learned embeddings integrate distinct sub-embeddings for user clicks, hotel attributes, and geographic information, providing a representation that can be used flexibly depending on the application.

An important advantage of the proposed neural model is that it addresses the cold-start problem for hotels with insufficient historical click information by incorporating additional hotel attributes, which are available for all hotels.
We show through the results of an online A/B test that our model generates high-quality representations that boost the performance of a hotel recommendation system on a large online travel platform.