Pavel Levin’s research while affiliated with Tel Aviv University and other places

What is this page?


This page lists works of an author who doesn't have a ResearchGate profile or hasn't added the works to their profile yet. It is automatically generated from public (personal) data to further our legitimate goal of comprehensive and accurate scientific recordkeeping. If you are this author and want this page removed, please let us know.

Publications (4)


Figure 3: 2-D visualization of city embeddings learned by Cleora. Each color denotes a diï¿¿erent hotel country.
Figure 5: 2-D visualization of city embeddings learned by Cleora. Each color denotes a different hotel country out of top 20 countries. Taken from [6].
Figure 7: Alternative Destinations Recommendations
Dataset description
Top performing Booking.com challenge teams
Booking.com Multi-Destination Trips Dataset
  • Conference Paper
  • Full-text available

July 2021

·

1,081 Reads

·

12 Citations

·

Pavel Levin
Download

Dataset description
Challenge key dates
Top 10 performing teams
Booking.com WSDM WebTour 2021 Challenge

March 2021

·

1,495 Reads

·

7 Citations

·

Kostia Kofman

·

Pavel Levin

·

[...]

·

Guy Nadav

The ACM WSDM WebTour 2021 Challenge focuses on a multi- destinations trip planning problem, which is a popular scenario in the travel domain. The goal of the challenge is to make the best recommendation of an additional trip destination. To encourage research on this field, Booking.com provided a unique dataset based on millions of real anonymized bookings. More than 800 participants have signed up for the contest. Best performing team achieved Accuracy @ 4 of 0.5939, using a blend of Transformers, GRUs, and feed-forward multi-layer perceptron. Additional leading teams implemented advanced state-of-the-art solutions to tackle the problem.



Efficient Image Gallery Representations at Scale Through Multi-Task Learning

May 2020

·

17 Reads

Image galleries provide a rich source of diverse information about a product which can be leveraged across many recommendation and retrieval applications. We study the problem of building a universal image gallery encoder through multi-task learning (MTL) approach and demonstrate that it is indeed a practical way to achieve generalizability of learned representations to new downstream tasks. Additionally, we analyze the relative predictive performance of MTL-trained solutions against optimal and substantially more expensive solutions, and find signals that MTL can be a useful mechanism to address sparsity in low-resource binary tasks.

Citations (3)


... In our experiments, we utilized several multimodal datasets that include both visual and sequential information to evaluate the effectiveness of the proposed SelfAM-Vtrans model. Specifically, we employed four datasets: TripAdvisor Dataset (Nilizadeh et al., 2019), Expedia Dataset (Goldenberg and Levin, 2021), Yelp Dataset (Asghar, 2016), and Open Images Dataset (Kuznetsova et al., 2020). The TripAdvisor and Expedia datasets provide user-generated reviews and travel-related data such as geographic locations, reviews, and ratings of hotels, restaurants, and tourist attractions, capturing textual information relevant to user preferences. ...

Reference:

A multimodal travel route recommendation system leveraging visual Transformers and self-attention mechanisms
Booking.com Multi-Destination Trips Dataset

... Neural network-based multitask learning is widely used in practice. e same network structure is used to extract features and design different loss functions for joint training to achieve noise reduction and performance improvement [23,24]. e network structure is divided into two branches: the left side is the one-dimensional convolutional layer, maximum pooling layer, and dropout layer, and the right side is the BiLSTM layer, attention mechanism layer, and fully connected layer, in that order. ...

Efficient Image Gallery Representations at Scale Through Multi-Task Learning
  • Citing Conference Paper
  • July 2020