PGHPython: DyNet: the Dynamic Neural Network Toolkit by Graham Neubig

Machine learning is an extremely useful tool for implementing applications that require some sort of "artificial intelligence," such as language understanding, speech processing, image processing, or many other tasks. In particular, recently there has been a large amount of progress made in this area due to the introduction of "deep learning," which uses "neural networks" to implement these machine learning applications, and has recently caused great progress in the area. In this talk I'll give a very brief overview of this field, then give a tutorial on the DyNet toolkit, which is designed to make it possible to implement neural networks for these applications simply and efficiently. I'll show examples of how to implement models for text processing (specifically using an example of sentiment analysis: determining whether a review is positive or negative), that will demonstrate how we can start with very simple models and make ones that are progressively more sophisticated with a small amount of effort.

Before coming to the tutorial, it would be ideal if participants could install the Python bindings for the DyNet toolkit:

http://dynet.readthedocs.io/en/latest/python.html

It might also be useful to download some tutorial materials here, which will form the basis of the talk:

https://github.com/neubig/yrsnlp-2016