Online publishers rely on different techniques to trap web visitors, clickbait being one such technique. Besides being a bad habit, clickbait is also a strong indicator for fake news spreading. Its presence in online media leads to an overall bad browsing experience for the web consumer. Recently, big players on the Internet scene, such as search engines and social networks, have turned their attention towards this negative phenomenon that is increasingly present in our everyday browsing experience. The research community has also joined in this effort, a broad band of detection techniques being developed. These techniques are usually based on intelligent classifiers, for which feature selection is of great importance. The work presented in this paper brings our own contributions to the field of clickbait detection. We present a new language-independent strategy for clickbait detection that takes into consideration only features that are general enough to be independent of any particular language. The methods presented in this paper could be applied to web content written in different languages. In addition, we present the results of a complex experiment that we performed to evaluate our proposed method and we compare our results with the most relevant results previously obtained in the field.

No abstract