Unsupervised Learning from Sequential Data with Recurrent Neural Networks

Project

Unsupervised Learning from Sequential Data with Recurrent Neural Networks

Principal Investigator

Alexander Rush

Harvard's School of Engineering and Applied Sciences

Oracle Principal Investigator

Guy Steele, Software Architect
Michael Wick, Principal Member Technical Staff

Summary

We investigate unsupervised or lightly-supervised methods and algorithms to uncover structure in sequential data. Our main focus is on parsing natural language, especially to transfer parsers from one language to another, thereby leveraging structure shared across languages to reduce the amount of necessary labelled data.