Florent Garcin

Personalized news recommendation with context trees

The proliferation of online news creates a need for filtering interesting articles. Compared to other products, however, recommending news has specific challenges: news preferences are subject to trends, users do not want to see multiple articles with similar content, and frequently we have insufficient information to profile the reader. In this paper, we introduce a class of news recommendation systems based on context trees. They can provide high-quality news recommendations to anonymous visitors based on present browsing behaviour. Using an unbiased testing methodology, we show that they make accurate and novel recommendations, and that they are sufficiently flexible for the challenges of news recommendation.

Offline and online evaluation of news recommender systems at swissinfo.ch

We report on the live evaluation of various news recommender systems conducted on the website swissinfo.ch. We demonstrate that there is a major difference between offline and online accuracy evaluations. In an offline setting, recommending most popular stories is the best strategy, while in a live environment this strategy is the poorest. For online setting, context-tree recommender systems which profile the users in real-time improve the click-through rate by up to 35%. The visit length also increases by a factor of 2.5. Our experience holds important lessons for the evaluation of recommender systems with offline data as well as for the use of the click-through rate as a performance indicator.

Reporting incentives and biases in online review forums

Online reviews have become increasingly popular as a way to judge the quality of various products and services. However, recent work demonstrates that the absence of reporting incentives leads to a biased set of reviews that may not reflect the true quality. In this paper, we investigate underlying factors that influence users when reporting feedback. In particular, we study both reporting incentives and reporting biases observed in a widely used review forum, the Tripadvisor Web site. We consider three sources of information: first, the numerical ratings left by the user for different aspects of quality; second, the textual comment accompanying a review; third, the patterns in the time sequence of reports. We first show that groups of users who discuss a certain feature at length are more likely to agree in their ratings. Second, we show that users are more motivated to give feedback when they perceive a greater risk involved in a transaction. Third, a users rating partly reflects the difference between true quality and prior expectation of quality, as inferred from previous reviews. We finally observe that because of these biases, when averaging review scores there are strong differences between the mean and the median. We speculate that the median may be a better way to summarize the ratings.

Predicting Online Performance of News Recommender Systems Through Richer Evaluation Metrics

We investigate how metrics that can be measured offline can be used to predict the online performance of recommender systems, thus avoiding costly A-B testing. In addition to accuracy metrics, we combine diversity, coverage, and serendipity metrics to create a new performance model. Using the model, we quantify the trade-off between different metrics and propose to use it to tune the parameters of recommender algorithms without the need for online testing. Another application for the model is a self-adjusting algorithm blend that optimizes a recommenders parameters over time. We evaluate our findings on data and experiments from news websites.

A Study of Forward Error Correction Schemes for Reliable Transport in Underwater Sensor Networks

Underwater communications is a very challenging topic due to its singular channel characteristics. Most protocols used in terrestrial wireless communications can not be directly applied in the underwater world. A high bit error rate and low propagation delay make the design of reliable transport protocols especially awkward. In this paper, we first propose four schemes that combine forward error correction mechanisms at the bit and/or packet level to increase the reliability in a non-cooperative scenario. The broadcast property of the underwater environment allows us to extend them to a cooperative setting. Based on our analyses, we introduce ADELIN: an adaptive reliable transport protocol for underwater sensor networks. We suggest an architecture for implementation and compare our protocol to other schemes. We show that it succeeds in a better probability and energy tradeoff for both single- and multi-hop communications.

PEN RecSys: a personalized news recommender systems framework

We present the Personalized News (PEN) recommender systems framework, currently in use by a newspaper website to evaluate various algorithms for news recommendations. We briefly describe its system architecture and related components. We show how a researcher can easily evaluate different algorithms thanks to a web-based interface.

Focal: a personalized mobile news reader

Traditionally in mobile apps, news articles and recommendations are presented to the users as an ordered list. This ordering often reflects the freshness of the stories. Although most users are satisfied with such presentation, some users have different expectations and want to read stories related to some specific topics. In this demo, we depart from the classic list-view layout and aim at exploring other ways to present news stories to the users. We introduce Focal, a personalized mobile news reader, which implements a fisheye-inspired interface. We briefly describe its system architecture and interface.