Eidetic Representations of Natural Language

The field of natural language processing (NLP) researches computational models of human language, with the goal of enabling machines to perform language tasks that are thought to require human-like intelligence (such as reading and understanding text, or conversing with humans). The recent wave of scientific breakthroughs in NLP is powered by neural language models; in essence, these models "read" very large collections of text and so acquire general world knowledge, which they can then use to perform language tasks.

In this project, we conduct fundamental research in the field of neural language modeling to address three principal limitations of current approaches. Our goal is to create the "next generation" of neural language models and significantly advance the state-of-the-art in NLP. Our research is made publicly available through the Flair framework (https://github.com/flairNLP/flair) to enable further research or application to use cases.