Presented By O’Reilly and Cloudera
Make Data Work
March 5–6, 2018: Training
March 6–8, 2018: Tutorials & Conference
San Jose, CA

Word embeddings under the hood: How neural networks learn from language

Patrick Harrison (S&P Global)
2:40pm3:20pm Thursday, March 8, 2018
Average rating: ****.
(4.33, 3 ratings)

Who is this presentation for?

  • Data scientists and machine learning engineers

Prerequisite knowledge

  • A basic understanding of machine learning concepts

What you'll learn

  • Understand how word vector models capture semantic relationships, the nuts and bolts of neural network modeling, including neuron layers, activation functions, loss functions and gradients, how back propagation really works, etc., and how natural language data can be used for neural network modeling


Since their introduction in the early 2010s, word vector embedding models have exploded in popularity and use. They are one of the key breakthroughs that have enabled a new, state-of-the-art approach to natural language processing based on deep learning. But despite their impact, relatively few practitioners understand how word vector models work under the hood to capture the semantic relationships within natural language data and produce their remarkable results.

Patrick Harrison opens up the black box of a popular word embedding algorithm and walks you through how it works its magic. Patrick also covers core neural network concepts, including hidden layers, loss gradients, backpropagation, and more.

This talk is based on an excerpt from the forthcoming book Deep Learning with Text from O’Reilly Media.

Photo of Patrick Harrison

Patrick Harrison

S&P Global

Patrick Harrison started and leads the data science team at S&P Global Market Intelligence (S&P MI), a business and financial intelligence firm and data provider. The team employs a wide variety of data science tools and techniques, including machine learning, natural language processing, recommender systems, graph analytics, among others. Patrick is the coauthor of the forthcoming book Deep Learning with Text from O’Reilly Media, along with Matthew Honnibal, creator of spaCy, the industrial-strength natural language processing software library, and is a founding organizer of a machine learning conference in Charlottesville, Virginia. He is actively involved in building both regional and global data science communities. Patrick holds a BA in economics and an MS in systems engineering, both from the University of Virginia. His graduate research focused on complex systems and agent-based modeling.

Comments on this page are now closed.


Picture of Patrick Harrison
03/11/2018 11:25am PDT

Hi Mario,

I just now provided the slides to O’Reilly. Hopefully they’ll be posted soon — stay tuned!


Mario Gamboa-Cavazos | CEO
03/11/2018 1:10am PST

Hi Patrick,

It was great to be at your presentation last week at Strata San Jose. I was wondering if you will be posting your slides to the conference website? I would love to review them again.

Many thanks!