Michael P. O'Mahony

Collaborative recommendation: A robustness analysis

Collaborative recommendation has emerged as an effective technique for personalized information access. However, there has been relatively little theoretical analysis of the conditions under which the technique is effective. To explore this issue, we analyse the <i>robustness</i> of collaborative recommendation: the ability to make recommendations despite (possibly intentional) noisy product ratings. There are two aspects to robustness: recommendation <i>accuracy</i> and <i>stability</i>. We formalize recommendation accuracy in machine learning terms and develop theoretically justified models of accuracy. In addition, we present a framework to examine recommendation stability in the context of a widely-used collaborative filtering algorithm. For each case, we evaluate our analysis using several real-world data-sets. Our investigation is both practically relevant for enterprises wondering whether collaborative recommendation leaves their marketing operations open to attack, and theoretically interesting for the light it sheds on a comprehensive theory of collaborative recommendation.

Detecting noise in recommender system databases

In this paper, we propose a framework that enables the detection of noise in recommender system databases. We consider two classes of noise: natural and malicious noise. The issue of natural noise arises from imperfect user behaviour (e.g. erroneous/careless preference selection) and the various rating collection processes that are employed. Malicious noise concerns the deliberate attempt to bias system output in some particular manner. We argue that both classes of noise are important and can adversely effect recommendation performance. Our objective is to devise techniques that enable system administrators to identify and remove from the recommendation process any such noise that is present in the data. We provide an empirical evaluation of our approach and demonstrate that it is successful with respect to key performance indicators.

Learning to recommend helpful hotel reviews

User-generated reviews are a common and valuable source of product information, yet little attention has been paid as to how best to present them to end-users. In this paper, we describe a classification-based recommender system that is designed to recommend the most helpful reviews for a given product. We present a large-scale evaluation of our approach using TripAdvisor hotel reviews, and we show that our approach is capable of suggesting superior reviews compared to a number of alternative recommendation benchmarks.

Google Shared. A Case-Study in Social Search

Web search is the dominant form of information access and everyday millions of searches are handled by mainstream search engines, but users still struggle to find what they are looking for, and there is much room for improvement. In this paper we describe a novel and practical approach to Web search that combines ideas from personalization and social networking to provide a more collaborative search experience. We described how this has been delivered by complementing, rather than competing with, mainstream search engines, which offers considerable business potential in a Google-dominated search marketplace.

A classification-based review recommender

Many online stores encourage their users to submit product or service reviews in order to guide future purchasing decisions. These reviews are often listed alongside product recommendations but, to date, limited attention has been paid as to how best to present these reviews to the end-user. In this paper, we describe a supervised classification approach that is designed to identify and recommend the most helpful product reviews. Using the TripAdvisor service as a case study, we compare the performance of several classification techniques using a range of features derived from hotel reviews. We then describe how these classifiers can be used as the basis for a practical recommender that automatically suggests the most-helpful contrasting reviews to end-users. We present an empirical evaluation which shows that our approach achieves a statistically significant improvement over alternative review ranking schemes.

Mining the real-time web: A novel approach to product recommendation

Real-time web (RTW) services such as Twitter allow users to express their opinions and interests, often expressed in the form of short text messages providing abbreviated and highly personalized commentary in real-time. Although this RTW data is far from the structured data (movie ratings, product features, etc.) that is familiar to recommender systems research, it can contain useful consumer reviews on products, services and brands. This paper describes how Twitter-like short-form messages can be leveraged as a source of indexing and retrieval information for product recommendation. In particular, we describe how users and products can be represented from the terms used in their associated reviews. An evaluation performed on four different product datasets from the Blippr service shows the potential of this type of recommendation knowledge, and the experiments show that our proposed approach outperforms a more traditional collaborative-filtering based approach.

Unsupervised retrieval of attack profiles in collaborative recommender systems

Trust, reputation and recommendation are key components of successful e-commerce systems. However, e-commerce systems are also vulnerable in this respect because there are opportunities for sellers to gain advantage through manipulation of reputation and recommendation. One such vulnerability is the use of fraudulent user profiles to boost (or damage) the ratings of items in an online recommender system. In this paper we cast this problem as a problem of detecting anomalous structure in network analysis and propose a novel mechanism for detecting this anomalous structure. We present an evaluation that shows that this approach is effective at uncovering the types of recommender systems attack described in the literature.

A recommender system for on-line course enrolment: an initial study

In this paper we report on our work to date concerning the development of a course recommender system for University College Dublins on-line enrollment application. We outline the factors that influence student choices and propose solutions to address some of the key considerations that are identified. We empirically evaluate our approach using historical student enrolment data and show that promising performance is achieved with our initial design.

On the real-time web as a source of recommendation knowledge

The so-called real-time web (RTW) is a web of opinions, comments, and personal viewpoints, often expressed in the form of short, 140-character text messages providing abbreviated and personalized commentary in real-time. Twitter is undoubtedly the king of the RTW. It boasts 100+ million users and generates in the region of 50m tweets per day. This RTW data is far from the structured data (ratings, product features, etc.) familiar to recommender systems research, but it is useful to consider its applicability to recommendation scenarios. In this short paper we describe an experiment to look at harnessing the real-time opinions of movie fans, expressed through the Twitter-like short textual reviews available on the Blippr service (www.blippr.com). In particular we describe how users and movies can be represented from the terms used in their associated reviews and describe a number of experiments to highlight the recommendation potential of this RTW data-source and approach.

Towards a reputation-based model of social web search

While web search tasks are often inherently collaborative in nature, many search engines do not explicitly support collaboration during search. In this paper, we describe HeyStaks (www.heystaks.com), a system that provides a novel approach to collaborative web search. Designed to work with mainstream search engines such as Google, HeyStaks supports searchers by harnessing the experiences of others as the basis for result recommendations. Moreover, a key contribution of our work is to propose a reputation system for HeyStaks to model the value of individual searchers from a result recommendation perspective. In particular, we propose an algorithm to calculate reputation directly from user search activity and we provide encouraging results for our approach based on a preliminary analysis of user activity and reputation scores across a sample of HeyStaks users.