WWW 2015: eXclusive Insight through Swiss Glasses

This week our team conquered the beautiful city of Florence. WWW 2015, hosted in the Fortezza da Basso, was as every year rich in papers and attendees. This year, for the first time, an entire track was devoted to entrepreneurs. Data privacy, ethics, crowdsourcing, Big Data, and Linked Data were popular topics… as were research papers highlighting data trends.

Andrei Broder gave an compelling keynote on Friday on the history of Web search. He reported a number of recent statistics from Google: 20 billion pages crawled every day, a 100 petabyte index with 60 trillion addresses, 3 billion daily searches, and 15% new queries. Andrei also suggested to rename the search box ”request box”, given its ever expanding capabilities.

We share below our subjective view on a few talks we found interesting at the conference.

Luca Aiello presented a the paper on the evolution of conversations in the age of email overload. Luca highlighted many interesting facts extracted from a corpus of 16 billion emails. It was fun and often really surprising! [PDF]

Alberto’s presentations on how to fix domains and ranges of linked data properties at LDOW raised a number interesting questions. As early results show some promise (with >90% precision) he plans to continue working on the topic (stay tuned!). [PDF] [SLIDES]

Marcin presented a piece of work on executing provenance-enabled queries in a triplestore. He showed how such queries can be executed faster because of the selectivity of the provenance predicates. [PDF] [SLIDES]

A panel on Knowledge Graphs was organized towards the end of the industry track. Chris Welty (Google) and Yuqing Gao (Bing) shared their insight on how two leading search engines create, curate and make use of their knowledge graphs in order to provide a better search experience as well as new services. Paul Groth (Elsevier) and Lora Aroyo (Tagasauris) provided a complementary view on how to use KGs in an enterprise setting.

Djellel presented an analysis of Amazon Mturk in his paper ”The Dynamics of Micro-Task Crowdsourcing - The Case of Amazon MTurk”; his study unveils many interesting trends in micro-task crowdsourcing as observed over 5 years of data collected from the platform. [PDF] [SLIDES]

Ali Elkahky presented “A Multi-View Deep Learning Approach for Cross Domain User Modeling in Recommendation Systems”, a method for content recommendation based on deep learning. [PDF]

Michael Luggen entertained the attendees of LDOW with his presentation and his live demo of a graph-aware and adaptive UI engine for Linked Data. Uduvudu is built for non-specialist users and allows them to describe recurring sub-graph patterns. The user can flexibly and automatically extract, transform, and visualize such patterns. [PDF] [DEMO]

Masumi Shirakawa presented “N-gram IDF: A Global Term Weighting Scheme Based on Information Distance”. The paper introduces a redesigned IDF for n-grams by connecting it to an information distance. [PDF] [DEMO]

Time in Florence was fruitful as well as delightful; we hope to see you all @ WWW2016rendez-vous est pris pour Montréal.

XI students presenting their work WWW 2015

Check also our friends’ thoughts on WWW2015: Thomas Steiner from Google Inc., Paul Groth from Elsevier Labs,  Amy Guy from the University of Edinburgh, and Daniel Garijo from Universidad Politécnica de Madrid.