Genre in Semantic Networks: A study of the Lexicon of News Articles

Our project aims at understanding text genres within the domain of the news. Advances in computational methods and availability of digital corpora has ushered in a new age of empirically testing intuitions about genres and styles, in particular the automatic classification of a document to its genre. At the same time, identifying systematic patterns of difference between genres, both quantitative and qualitative, can bring insight to our understanding of discourse from a linguistic perspective, and perhaps the social motivations that underlie our speech patterns as well. From a more applied perspective, discovering the key differences between genres can be of enormous use to natural language processing (NLP) tasks that work with corpora from a range of genres by identifying biases associated with particular genres or potential artifacts of genre imbalance in datasets. Other tasks such as Part-of-Speech (POS) tagging, parsing, word sense disambiguation (WSD), and information retrieval can also be significantly improved when taking genre into account.