Abstract (by author):

Language and speech technology have advanced rapidly in the last decades. This is mainly due to the large volumes of text and speech that are now stored electronically as well as the increased computing power available to train statistical models on large data sets. However, the models used are often surprisingly naive from a linguistic point of view. Only very basic insights about language are incorporated. In this talk, we will explore the reasons for this state of affairs and look into a crystal ball: Will future models in language processing continue to be simplistic or will they mimick the way humans process language?