Meta-User2Vec model for addressing the user and item cold-start problem in recommender systems
- PDF / 1,744,022 Bytes
- 26 Pages / 439.37 x 666.142 pts Page_size
- 77 Downloads / 176 Views
Meta‑User2Vec model for addressing the user and item cold‑start problem in recommender systems Joanna Misztal‑Radecka1,2 · Bipin Indurkhya3 · Aleksander Smywiński‑Pohl1 Received: 31 October 2019 / Accepted in revised form: 25 September 2020 © The Author(s) 2020
Abstract The cold-start scenario is a critical problem for recommendation systems, especially in dynamically changing domains such as online news services. In this research, we aim at addressing the cold-start situation by adapting an unsupervised neural User2Vec method to represent new users and articles in a multidimensional space. Toward this goal, we propose an extension of the Doc2Vec model that is capable of representing users with unknown history by building embeddings of their metadata labels along with item representations. We evaluate our proposed approach with respect to different parameter configurations on three real-world recommendation datasets with different characteristics. Our results show that this approach may be applied as an efficient alternative to the factorization machine-based method when the user and item metadata are used and hence can be applied in the cold-start scenario for both new users and new items. Additionally, as our solution represents the user and item labels in the same vector space, we can analyze the spatial relations among these labels to reveal latent interest features of the audience groups as well as possible data biases and disparities. Keywords Recommender system · Neural embeddings · Cold-start · Doc2Vec model
* Joanna Misztal‑Radecka [email protected] Bipin Indurkhya [email protected] Aleksander Smywiński‑Pohl [email protected] 1
AGH University of Science and Technology, Krakow, Poland
2
Ringier Axel Springer Polska, Warsaw, Poland
3
Cognitive Science Department, Institute of Philosophy, Jagiellonian University, Krakow, Poland
13
Vol.:(0123456789)
J. Misztal‑Radecka et al.
1 Introduction This work aims to address the cold-start problem in recommendation systems for both the new-user and the new-item situations. We focus here on defining a strategy for the complete cold-start (CCS) situation when no historical browsing data are available [in contrast to incomplete cold start when a small number of records are available (Wei et al. 2016)]. Toward this goal, we propose MetaUser2Vec, which adapts the User2Vec method of Phi et al. (2016) based on the Doc2Vec architecture of Le and Mikolov (2014) to represent new users and new items in the recommendation system in an unsupervised way. More specifically, the work presented here makes the following contributions: • We propose a hybrid Meta-User2Vec model, which enables modeling users
and items with unknown history based on their metadata as described in Sect. 3 and can be applied for both the user and the item cold-start problems. • We evaluate our proposed approach with respect to different model architectures and parameters and compare its performance with a factorization machine approach on three real-world datasets (Sect
Data Loading...
