Expert System took part in the recent “Data Science Award Spain 2019”, an event organized by LUCA, the business unit specialized in Big Data and Artificial Intelligence of Telefónica, one of the world’s largest telecommunications companies. José Manuel Gómez-Pérez, Director of R&D at Expert System, was invited to help judge the awards and to participate in a panel discussion, “Artificial Intelligence: under the microscope.”
At the panel, fake news was at the center of the debate. José Manuel Gómez-Pérez pointed to the important contribution of natural language processes to help identify fake news (listen to the discussion at minute 54:00-58:00 here).
In a round table moderated by Paz Noriega, Communication Manager at Telefónica, José Manuel Gómez-Pérez explained the contribution of processes related to natural language to detect and identify possible false news. Gómez-Pérez believes that “only with natural language processing is not enough to detect misinformation. Models are being worked on that combine the analysis of text with those of the reputation or credit of the sources.”
José Manuel Gómez-Pérez pointed out that, “the language of fake news has characteristics that can be captured through certain models, since it is highly emotional, controversial, aggressive and based on a simple lexicon, so that it is easy to assimilate. It is composed, to a large extent, by language of discredit towards alternative sources.” Our director of the LAB in Madrid, also delved into the ‘language models’ model that is capable of interpreting, with 92% accuracy, if something has been generated as false by another machine, only from a written text.”
Finally, José Manuel Gómez explained that his team is investigating the automatic detection of bots, based only on the content. This work is in the development phase, but it is beginning to provide positive results, although “there is still a lot of work ahead.”