NLP Meetup : Convolutional Neural Networks for NLP

What we'll do
This talk will introduce Convolutional Neural Networks (CNNs) in the context of Natural Language Processing (NLP). We will first go over the concept of convolutional layers, how and why they allowed breakthroughs in the field of computer vision. We will then explore how we can apply CNNs to textual data and what is their advantage over Recurrent Neural Networks (RNN). We will analyze in detail an architecture proposed in 2015 by Zhang et al.: Character-level convolutional Networks for Text Classification (ref: https://arxiv.org/abs/1509.01626) that was the first one to introduce character-level convolutions. Finally we will go over some more advanced techniques like convolutional Sequence to Sequence learning for text translation or how to combine both CNNs and LSTMs. The talk will include some live-coding examples in the form of jupyter notebooks that will be made available to the attendees.

Speaker Bio:
Thomas Delteil (https://github.com/thomasdelteil/) is an applied scientist currently employed at Amazon in the AWS Deep Learning team. He has a background in machine learning and software engineering, and was previously working for Microsoft Cloud AI team as an applied scientist. He holds a MSc from Imperial College London in Advanced Computing and another MSc from ISAE-Supaero, Toulouse in Aerospace Engineering.

Agenda:
17:45 - Doors Open & Mingle
18:15 - Doors Closed & Introductions
18:30 - Talk
19:25 - Questions & Discussion
19:45 - Close