Session 3: Cold Start and Hybrid Recommenders

Date: Tuesday, Oct 7, 16:15-18:00
Moderator: Joe Konstan

• Ensemble Contextual Bandits for Personalized Recommendation

  by Liang Tang, Yexi Jiang, Lei Li and Tao Li

  The cold-start problem has attracted extensive attention among various online services that provide personalized recommendation. Many online vendors employ contextual bandit strategies to tackle the so-called exploration/exploitation dilemma rooted from the cold-start problem. However, due to high-dimensional user/item features and the underlying characteristics of bandit policies, it is often difficult for service providers to obtain and deploy an appropriate algorithm to achieve acceptable and robust economic profit. In this paper, we explore ensemble strategies of multiple contextual bandit algorithms to obtain robust predicted click-through rate (CTR) of web objects. Specifically, the ensemble is acquired by aggregating different pulling policies of bandit algorithms, rather than forcing the agreement of prediction results or learning a unified predictive model. To this end, we employ a meta-bandit paradigm that places a hyper bandit over the base bandits, to explicitly explore/exploit the relative importance of base bandits based on user feedbacks. Extensive empirical experiments on two real-world data sets (news recommendation and online advertising) demonstrate the effectiveness of our proposed approach in terms of CTR.

• Cold-start News Recommendation with Domain-dependent Browse Graph

  by Michele Trevisiol, Luca Maria Aiello, Rossano Schifanella and Alejandro Jaimes

  Online social networks and mash-up services create opportunities to connect different web services otherwise isolated. Specifically in the case of news, users are very much exposed to news articles while performing other activities, such as social networking or web searching. Browsing behaviour aimed to the consumption of news, especially in relation to the visits coming from other domains, has been mainly overlooked in previous work. To address that, we build a BrowseGraph out of the collective browsing traces extracted from a large viewlog of Yahoo News (0.5B entries) and we define the ReferrerGraph as its subgraph induced by the sessions with the same referrer domain. The structural and temporal properties of the graph show that browsing behavior in news is highly dependent on the referrer URL of the session, in terms of type of content consumed and time of consumption. We build on this observation and propose a news recommender that addresses the cold-start problem: given a user landing on a page of the site for the first time, we aim to predict the page she will visit next. We compare 24 flavors of recommenders belonging to the families of content-based, popularity-based, and browsing-based models. We show that the browsing-based recommender that take into account the referrer URL is the best performing, achieving a prediction accuracy of 61% in conditions of heavy data sparsity.

• Item Cold-Start Recommendations: Learning Local Collective Embeddings

  by Martin Saveski and Amin Mantrach

  Recommender systems suggest to users items that they might like (e.g., news articles, songs, movies) and, in doing so, they help users deal with information overload and enjoy a personalized experience. One of the main problems of these systems is the item cold-start, i.e., when a new item is introduced in the system and no past information is available, then no effective recommendations can be produced. The item cold-start is a very common problem in practice: modern online platforms have hundreds of new items published every day. To address this problem, we propose to learn Local Collective Embeddings — a matrix factorization that exploits items’ properties and past user preferences while enforcing the manifold structure exhibited by the collective embeddings. We present a learning algorithm based on multiplicative update rules that are efficient and easy to implement. Experiments on two item cold-start use cases: news recommendation and email recipient recommendation, demonstrate the effectiveness of this approach and show that it significantly outperforms six state-of-the-art methods for item cold-start.

• Improving The Discriminative Power Of Inferred Content Information Using Segmented Virtual Profile

  by Haishan Liu, Anuj Goyal, Trevor Walker and Anmol Bhasin

  We present a novel component of a hybrid recommender system at LinkedIn, where item features are augmented by a virtual profile based on observed user-item interactions. The concept of virtual profiles is generating a representation of an item in the user feature space by leveraging the over-represented user features from users that interacted with the item. It is a way to think about Collaborative Filtering with content features. The core principle is that if the feature occurs with high probability for the users who interacted with an item (henceforth termed as relevant users) versus those who did not (henceforth termed non-relevant users), then that feature is a good candidate to be included in the virtual profile of the item in question. However this scheme suffers from the data imbalance problem, given that observed relevant users are usually an extremely small minority group compared to the whole user base. Feature selection in this skewed setting is prone to noise from the overwhelming non-relevant examples that belong to the majority class. To alleviate the problem, we propose a method to select the most relevant non-relevant examples from the majority class by segmenting users on certain intelligently selected feature dimensions. The resulting virtual profile from the method is called the segmented virtual profile. Empirical evaluation on real-world large scale recommender system at LinkedIn shows that simple strategies for the segmentation yield significantly better performance.

• Ratings Meet Reviews, a Combined Approach to Recommend

  by Guang Ling, Michael Lyu and Irwin King

  Most existing recommender systems focus on modeling the ratings while ignoring the abundant information embedded in the review text. In this paper, we propose a unified model that combines content-based filtering with collaborative filtering, harnessing the information of both ratings and reviews. We apply topic modeling techniques on the review text and align the topics with rating dimensions to improve prediction accuracy. With the information embedded in the review text, we can alleviate the cold-start problem. Furthermore, our model is able to learn latent topics that are interpretable. With these interpretable topics, we can explore the prior knowledge on items or users and recommend completely "cold" items. Empirical study on 27 classes of real-life datasets show that our proposed model lead to significant improvement compared with strong baseline methods, especially for datasets which are extremely sparse where rating-only methods cannot make accurate predictions.

Back to Program