The Web Conference 2018 took place from 23 to 27 April 2018. This and the previous post are an account of the contributions made by WDAqua.

After the first two days of workshops, the main conference tracks started on the 25th of April and with these the plenary sessions and keynotes. Speakers and panel members were top level and covered key topics around the Web and Artificial Intelligence. How to deal with extremism and polluting agents in general on the Web, and whether enforcing some sort of control would be a good strategy against that were among the questions raised during the compelling keynote of Luciano Floridi, of the University of Oxford. Find the video of his talk here.

Similar topics were central also in the discussion of the amazing panel of day 2. Tim Berners-Lee (MIT, W3C) and the panel chair, Wendy Hall (University of Southampton), pointed out the danger represented by the bias inherent to some Artificial Intelligence applications and the implications of that. On the other hand, Vinton Cerf drew the attention on the fact that whereas applications are developed to facilitate users’ lives – to reduce friction as put it by the keynote speaker, Ruhi Sarikaya – the existence of several, sometimes competing, applications in various ecosystems actually increases friction and may actually be detrimental for users. Furthermore, Vinton Cerf suggested that the negative behaviour of technology should be tackled by investing on education, rather than by regulatory control, as we are not fully aware yet of all the possible effects of the technologies available at present and in the near future. This indirectly addressed some of the remarks made by Luciano Floridi on the first day.

On the second day of the main conference, Kuldeep Singh presented his paper “Why Reinvent the Wheel: Let’s Build Question Answering Systems Together”.

The moment we combine efforts we can achieve better results than the baseline approaches. @OyeKuldeep Presenting Frankenstein framework @WDAqua #TheWebConf pic.twitter.com/BM8EITReh7

— chabrowa (@kacprzakemilia) April 26, 2018

This is the outcome of Kuldeep’s and other WDAqua members’ work and addresses the question of how to flexibly integrate different components in a Question-Answering pipeline, in order to optimise its performance. The answer to that is an approach that uses machine learning to select the best performing components for the input questions. This approach is implemented in Frankenstein, a QA framework able to select QA components and compose QA pipelines. We knew Frankenstein could infuse life into creatures, but apparently from now on it can contribute to (Artificial) Intelligence!

If you feel enticed by this description and want to know more, the paper can be found here or you can watch Kuldeep’s presentation below.