Peggy van der Kreeft, from Deutsche Welle Innovation, represented SUMMA at the Media Fast Forward (MFF2016) event, organized by the VRT (Flemish Radio and Television company) in Brussels on 8 December 2016. Dieter Boen, VRT Innovation Manager, introduced the event and talked about VRT efforts on innovation collaboration and its tools, such as VRT Sandbox, Open VRT, and the Live-IP Studio. Trying out the new technologies on the floor, in the production process itself, is vital, as it reveals the issues and challenges. Some 400 people registered for the event, the majority coming from Flanders, as most of the sessions were in Flemish. Nevertheless, it also provided a somewhat broader, international perspective, starting with the keynote speech by Phil Fearnley, BBC Director of Homepage & My BBC, on the BBC transformation journey, personalization of content and user engagement.
Other interesting talks by Pierre Marchand, Teradata Business Analytics Consultant, and Wouter Quartier, VRT Strategy Manager, addressed the importance of proper forecasting in media, in particular in the light of recent events such as Brexit and the U.S. elections. We need a different view on how to measure public opinions, including sentiment analysis tools that use resources that were left unexplored so far. Current analysis methods and sources obviously do not reflect the majority of users. Innovative and explorative social media analysis tools are essential to get a grip on public opinion. Various analysis methods were presented and discussed. A lively panel discussion debated on the increasing role of video-on-demand.
Especially interesting for SUMMA was a demonstration by VRT Research and Innovation Manager Marieke Lycke on the recently finished project STON, focusing on speech and language technology subtitling in Dutch. It showed automatically generated subtitles in Dutch for VRT content, implementing a tool developed by a consortium, including the Belgian university KU Leuven and the University of Ghent, applied to VRT on-demand-videos. The demo showed differences in quality, based on content type – for instance, documentaries with studio-recorded spoken text by one reporter clearly results in much better speech-to-text output than live reporting with interviews and background noise/voices. The implementation of automatic punctuation (only periods thus far) and person and organization recognition and identification was shown and the related issues discussed.
Although this tool is focusing only on Dutch (and in particular the Flemish version of the Dutch language), a language not covered by SUMMA, the demonstration of such language tools are interesting to determine the state-of-the-art of technologies used in SUMMA, and the (projected) use and research undertaken at other (European) broadcasters. It also provides good comparative material for content providers such as Deutsche Welle for assessing the outcome of similar technologies in the SUMMA platform, in this case speech-to-text and identifying issues of particular interest to content providers.