A language model is all you need
Deep learning models are said to have revolutionised the natural language processing field in the last couple of years in much the same way they did in the computer vision field about 7 years ago. At the heart of this revolution are more powerful models that learn to generate and predict text. They are trained on raw, unlabelled data and their ability to understand the structure and semantics of natural language seems to have a profound impact on many natural language processing tasks such as text classification, sentiment analysis, etc. In this talk, I will describe what a language model is, some of its variants, and a little about its history. Finally, I will give examples of how they can be used to solve business problems and how they relate to our product at Seal Software.
Originally a mathematician with a PhD in complex analysis, Aron Lagerberg is a data scientist working with neural networks at the Contract Analytics company Seal Software. As parts of one of the best machine learning teams in town, he and his teammates try to use state of the art deep language models to understand legal documents, and to extract the essential information from those documents.
