Presented By O’Reilly and Cloudera
Make Data Work
September 11, 2018: Training & Tutorials
September 12–13, 2018: Keynotes & Sessions
New York, NY

Deep learning methods for natural language processing

Garrett Hoffman (StockTwits)
9:00am–12:30pm Tuesday, 09/11/2018
Data science and machine learning
 Location: 1A 21/22 Level: Intermediate
Secondary topics:  Deep Learning, Text and Language processing and analysis
Average rating: ****.
(4.75, 4 ratings)

Who is this presentation for?

  • Data scientists, machine learning engineers, software engineers, and data engineers

Prerequisite knowledge

  • A basic understanding of Python, machine learning, and feed-forward neural networks
  • Familiarity with TensorFlow (useful but not required)

Materials or downloads needed in advance

  • A laptop with the most recent version of Docker installed
  • Clone the course GitHub repository (Before the tutorial, follow the instructions in the course GitHub repository to set up your working environment. Note: It is not required that you follow along during the sessions. Approximately half of the session is focused on the models themselves and the underlying mathematics, with the other half involving going through implementations of the models in code. All of the models will be run in advance since they take a few hours to train.)

What you'll learn

  • Learn how deep learning is applied to NLP-related tasks and how to implement these models in Python using TensorFlow
  • Understand practical considerations for applying deep learning methods to business problems

Description

Garrett Hoffman walks you through deep learning methods for natural language processing and natural language understanding tasks, using a live example in Python and TensorFlow with StockTwits data. Methods include word2vec, recurrent neural networks and variants (LSTM, GRU), and convolutional neural networks. You’ll explore use cases and motivations for using these methods, gain a conceptual intuition about how these models work, and briefly review the mathematics that underly each methodology.

Outline:

Representation learning for text with word2vec word embeddings: The CBOW and skip gram models; how to train custom word embeddings; how to use pretrained word embeddings, such as those trained on Google News

Traditional recurrent neural networks (RNNs): Why these types of models often perform better than traditional alternatives; variants to traditional RNNs, such as long short-term memory (LSTM) cells and gated recurrent units (GRUs); why these models provide improvements in accuracy

Convolutional neural networks (CNNs): Why CNNs that are traditionally applied to computer vision are now being applied to language models; advantages that these have over RNNs; how RNNs can be used to learn generative models for text synthesis and the applications of this method

Photo of Garrett Hoffman

Garrett Hoffman

StockTwits

Garrett Hoffman is a director of data science at StockTwits, where he leads efforts to use data science and machine learning to understand social dynamics and develop research and discovery tools that are used by a network of over one million investors. Garrett has a technical background in math and computer science but gets most excited about approaching data problems from a people-first perspective—using what we know or can learn about complex systems to drive optimal decisions, experiences, and outcomes.

Comments on this page are now closed.

Comments

Picture of Garrett Hoffman
Garrett Hoffman | DIRECTOR, DATA SCIENCE
09/11/2018 10:57am EDT

Hi All! The Github repo linked in the materials section above is updated with the final slides! Hope you all enjoyed!

Picture of Sophia DeMartini
Sophia DeMartini | SENIOR SPEAKER MANAGER
09/03/2018 5:52pm EDT

HI Alberto,

Yes, you’ll need to purchase a Gold or Silver pass to the conference, and will have to register for the specific tutorial you’d like to attend. (You do not need to pre-register for sessions).

Thank you,
Sophia

Alberto Andreotti |
09/03/2018 12:14pm EDT

Is it necessary to register in advance for the event? Thanks.

Alberto Andreotti |
09/03/2018 12:03pm EDT

Looking forward to see your presentation on this really exciting topic :-)