On machine learning in linguistics: how artifical neural networks compare to human language processing

Abstract

Deep learning models in the field of natural language processing (NLP) are now able to successfully translate, transcribe and produce texts of a high quality. Since language was thought to be a species-specific ability of mankind for so long, new and old questions arise in the field of NLP. The main question that I will be trying to answer in this paper is: Do artificial neural networks and humans process language in the same way? In the essay, firstly the topic of human language processing is discussed and explained. After, a number of researches will be listed and explained, that tackle NLP in the field of deep neural-networks (DNN). The results from these researches prove to be surprising and optimistic. A lot of DNN methods seem able to handle difficult language contraints. We conclude that while DNN methods are doing very well in the field of language, a DNN needs a to have bias of a hierarchical structure to come close to human language processing.