AI and NLP: Data Talks II

18:30 Doors Open
19:00 Keynote
19:10 Talk 1 (20 minutes + 10 minutes Q&A)
19:40 Talk 2 (20 minutes + 10 minutes Q&A)
20:10 Pizza and Beer Networking

Sam Davis
CEO amplified ai
Recognizing Bias Takes Effort – Data & Diversity in Machine Learning

— TALK 1 —
Eric Fandrich & Mitsuhiro Tsunoda
Financial Technology Solutions Division Senior Consultant
Nomura Research Institute

The Challenge of Cross-Border NLP – NRI’s experience bringing Facebook’s DrQA to Japan

— TALK 2 —-
Max Frenzel
Research Scientist
Cogent Labs

Variational Autoencoders for NLP: Basic Ideas, Particular Difficulties, and Recent Solutions

Deep generative models such as Variational Autoencoders (VAEs) and Generative Adversarial Networks (GANs) have played a big role in the recent advancement of unsupervised learning. They can be used to compress data, or reconstruct noisy or corrupted data, and allow us to perform sophisticated data manipulations. More subtly, but in practice often of most interest, they can uncover hidden concepts and relations in large amounts of unlabelled data. However, applying them to natural language has proven quite challenging. In this talk I will review the basic ideas behind VAEs, discuss why their application to NLP is particularly difficult, and introduce some more recent developments that have allowed us to overcome these challenges.
This talk will be loosely based on a series of articles I published on the same topic:

■Access to the community lab
Tokyo-to, Shinagawa-ku Osaki, 4-1-2 Win Gotanda Building 3F
7mins from JR Gotanda Station, West Exit
1min from Tokyu Ikegami Line Osaki Hirokoji Station

東京都品川区大崎4-1-2  ウィン第2五反田ビル 3F
JR五反田駅: 西口改札より徒歩7分