How to train your open source HAL 9000
The current state of opensource speech to text processing.
Schedule

Presented by:

B71889b8d80ac7c9e597d089b29d6f24

Aaron Chantrill

from The Foundry, 1000 F Street, Bellingham, WA

My earliest memory is of lying in my crib after being passed around by a group of adults and thinking to myself, "When I can talk, I'm going to ask them not to stick their thumbs under babies armpits when they pick them up". I have always been fascinated with cognition and communication. As a kid, I was more interested in writing video games and figuring out how they worked than with playing them. At Johns Hopkins in the late 80's, I got to work with professor Sejnowski on some early neural networks classifying phonemes from recorded speech. Over the last twenty years I have worked in machine learning, database administration, and web design with a focus on free and opensource software. I currently work as a medical informaticist. I am deeply concerned about the renormalization of expectations around privacy.

In this tutorial, we will get a copy of Jasper Voice Assistant up and running on Debian Stretch.

Jasper is an open source virtual assistant written in Python. It runs well on a Raspberry Pi 2 or 3. The particular version of Jasper we will be installing is one I have been working on for several months and includes a plugin for playing z-machine text adventure games.

Jasper can use local speech to text and text to speech systems, making it a good platform for the privacy concious tinkerer who likes the idea of a computer you can talk to, but prefers not to have random snippets of household conversation shared with Alphabet, Amazon or Apple.

This tutorial includes setting up and configuring the PocketSphinx and DeepSpeech Speech To Text and Festival Text To Speech systems, and writing a simple Jasper SpeechHandler plugin in Python.

Through this tutorial, we will talk about privacy concerns with large closed-source speech to text processing systems, the current state of open source speech to text processing, the importance of public voice datasets and how to contribute to them, and how speech recognition intersects with artificial intelligence.

Please bring a laptop with you with vagrant and virtualbox installed. Please download the aaronchantrill/JasperVB vagrant file (vagrant box add aaronchantrill/JasperVB). This box has an 18GB dynamically allocated hard drive and uses the standard vagrant insecure SSH key and passwords. This file is about 11GB. I will publish a video of the complete process (still taking 3 hours to compile) and a video of how to setup and test the vagrant box.

I plan to spend the first hour introducing the common strategies used by HMM and ANN speech to text systems, exploring the files used, building some custom dictionaries and doing some live demonstrations. The second hour will be spent creating a Jasper configuration file (naming the assistant), writing a simple SpeechHandler plugin, and talking about future development.

This tutorial was designed at The Foundry, Bellingham’s makerspace.

Date:
2018 April 29 - 05:00
Duration:
2 h
Room:
CC-201 TUT1
Conference:
LinuxFest Northwest 2018
Language:
Track:
Code
Difficulty:
Medium

Happening at the same time:

  1. Building and using your own inexpensive, Docker-based pico-cluster
    2. Start Time:
    2018 April 29 05:00

    Room:
    CC-202 TUT2

  2. How to train your open source HAL 9000
    3. Start Time:
    2018 April 29 05:00

    Room:
    CC-201 TUT1

  3. State of Video Copyright & Free Culture
    4. Start Time:
    2018 April 29 05:30

    Room:
    G-103

  4. Linux 101: Just the basics
    5. Start Time:
    2018 April 29 05:30

    Room:
    HC-108

  5. Securing Docker on the Cheap - Part 1 - Fundamentals
    6. Start Time:
    2018 April 29 05:30

    Room:
    CC-208

  6. Applying Object Character Recognition and Kubernetes to Twitch
    7. Start Time:
    2018 April 29 05:30

    Room:
    CC-200

  7. Switching to the BSDs: A crash course in FreeBSD, FreeNAS, TrueOS and OpenBSD
    8. Start Time:
    2018 April 29 05:30

    Room:
    CC-236

  8. Presentation Skills for Open Source Folks
    9. Start Time:
    2018 April 29 05:30

    Room:
    CC-115

  9. Linux in a Microsoft world
    10. Start Time:
    2018 April 29 05:30

    Room:
    CC-114

  10. Maintaining local forks
    11. Start Time:
    2018 April 29 05:30

    Room:
    CC-235

  11. Tables and columns and rows! Oh my!
    12. Start Time:
    2018 April 29 05:30

    Room:
    HC-103 Postgres

  12. Linux 102: Choosing the right distro for you
    13. Start Time:
    2018 April 29 06:45

    Room:
    HC-108

  13. Securing Docker on the Cheap - Part 2 - Vulnerabilities
    14. Start Time:
    2018 April 29 06:45

    Room:
    CC-208

  14. Introduction to the Clojure Programming Language
    15. Start Time:
    2018 April 29 06:45

    Room:
    CC-236

  15. Endless Summer of Code: Building the Foundation of a Community Through GSOC
    16. Start Time:
    2018 April 29 06:45

    Room:
    CC-115

  16. Linux email clients give me a migrane
    17. Start Time:
    2018 April 29 06:45

    Room:
    CC-114

  17. Let's uninstall GRUB
    18. Start Time:
    2018 April 29 06:45

    Room:
    CC-235

  18. Brave New World
    19. Start Time:
    2018 April 29 06:45

    Room:
    G-103

  19. End-to-end Encryption in Nextcloud
    20. Start Time:
    2018 April 29 06:45

    Room:
    CC-200

  20. Linux Professional Institute: Exam Lab - Session 2
    21. Start Time:
    2018 April 29 07:00

    Room:
    HC-112 LPI