Seminarium KMMF "Teoria Dwoistości"

Natural Language Processing, that is, processing large amounts of natural language data using computers, is both challenging and relevant for practical applications. The topic of the seminar is language modeling. In this task the goal is to "fill in the blank" in a sentence - in other words, to predict the probability distribution of words in a given context. Language modeling is important not only because of immediate applications (eg. in optical character recognition or in predictive keyboards), but also because such models are good starting points for other language processing tasks. We will cover both statistical (count-based) and neural language models. Specifically, I will focus on how discrete objects (words and sentences) can be represented to facilitate natural language processing tasks. I will also overview the challenges of developing commercial language models in the industry setting.