Knowledge Extraction for Question Titling

This article describes the work carried out over the database of questions belonging to the different opinion polls carried out over the last 50 years in Spain. Approximately half of the questions are provided with a title while the other half is untitled. It is described the work and techniques implemented in order to automatically generate the titles for the corpus of untitled questions. The process of title generation is performed over very short texts and generated titles are subject to strong stylistic conventions and should be fully grammatical pieces of Spanish.