MLT Workshop: Intro to NLP Research (Part I)

With billions of messages, comments, reviews, blog posts and tweets generated every day, text is still the predominant way of online-communication. But how can we analyze text data and derive valuable insights from it? Even though Natural Language Processing has been around for many years, it gained momentum with the advancement of deep learning and sequence-to-sequence learning.

For that reason, Machine Learning Tokyo is organizing a two-part workshop: "Intro do NLP Research" with Edison Marrese-Taylor and Jorge Andrés Balazs, NLP researchers at Matsuo Lab, an Artificial Intelligence Lab at the University of Tokyo.

Part I will be dedicated to understanding Natural Language Processing from scratch. We then move to a more practical hands-on part, where you will learn how to set up your research environment, the prerequisites for further computational experiments and the implementation of machine learning and deep learning algorithms.

Part II (end of September) will be dedicated to implementing a deep learning model for a NLP classification task in PyTorch.

*******************************************************************

All MLT workshops are free of charge.
For Part I we have 60 seats for registered participants. Please sign up with your full name and contact here: https://goo.gl/forms/oGADaMFjMCTPj5fh2

*******************************************************************

SCHEDULE

• 11:00 - 11:15: Welcome and Intro

• Suzana Ilic, MLT Lead

• 11:15 - 12:45: NLP from scratch: From String to Token to Vector

• Jorge Andrés Balazs, PhD Candidate, University of Tokyo

• 12:45 - 13:45: Lunch break

• 13:45 - 16:00: Setting up your Research Arena: tmux, SSH tunneling, ...

• Edison Marrese-Taylor, Postdoctoral Researcher, University of Tokyo

(with a short break in between)

*******************************************************************

Edison Marrese-Taylor is a post-doctoral research fellow at the University of Tokyo (http://weblab.t.u-tokyo.ac.jp/en/). He conducted his Ph.D. studies under the supervision of Prof. Yutaka Matsuo (http://ymatsuo.com/). His research topics are Computational Linguistics and Representation Learning. His work is in general concerned with the usage of machine learning in the study of different aspects of affect in text, such as opinion, emotion and sarcasm. He is also interested in the intersection between NLP and Software Engineering, working on tools to help us summarize source code changes using natural language. https://epochx.github.io/

Jorge Andrés Balazs is a Ph.D. student at the University of Tokyo under the supervision of Prof. Yutaka Matsuo, currently working on Deep Learning applied to Natural Language Processing. He's especially interested in how to exploit morphological information to help us obtain better word representations. His previous research involved predicting implicit emotion in tweets (paper), and entailment in pairs of sentences (paper). He's also a strong advocate of honing one's software engineering skills, regardless of background, to efficiently produce replicable and easy to understand code.