+ All Categories
Home > Technology > NMI15 Jan Šedivý – Rozpoznávání řeči

NMI15 Jan Šedivý – Rozpoznávání řeči

Date post: 21-Jul-2015
Category:
Upload: new-media-inspiration
View: 96 times
Download: 1 times
Share this document with a friend
24
Intro A Speech Recognition Odyssey Jan Šedivý ČVUT FEL, dept. of Cybernetics
Transcript

Intro

A Speech Recognition Odyssey Jan Šedivý

ČVUT FEL, dept. of Cybernetics

2001 A Space Odyssey

Stanley Kubrick

Arthur C. Clarke

1968

Human Evolution

Artificial Intelligence

Extraterrestrial life

Speech generation

Speech recognition

Question answering

Intelligent dialog

Artificial Intelligence

1993 IBM Personal Dictation System

1996 VoiceType office document, isolated words

1996 Nuance first speech application

1997 Dragon Systems Naturally Speaking

1999 VoiceXML

2000 Telephony applications

2002 Enabling car control

2003 Microsoft speech - Office 2003

2008 Google - mobile speech search

2009 Nuance Acquires IBM's Speech Technology patents

2911 Deep Belief Networks

The key is the microphone

Wave shape

15 ms segments

Spectrogram

Vectors (13)

Sounds are

different

Words

Phonemes

Triphones

Triphone model - context

Pattern matching

Viterbi algorithm

Search the best word

Statistical Language Model

Speech recognition model

Machine learning

Hundreds of hours of recorded speech

Billions of words for language model

Speech - Text - Understanding

Personal assistants

Google Now– Android

Siri Apple iOS

Cortana Microsoft

• Attention word

• Web search

• Predictive notifications (traffic on your commute is bad)

• Geofencing (e.g., reminding you to make a purchase when

you're near a business)

• Event or contact based notification ("when your sister calls,

tell her happy birthday")

• Call or text contact Name

• Make calendar appointment

Siri, Google Now, Cortana

• Set interests (favorite sports teams, e.g.)

• Check weather

• Daily summaries of info of interest to you

• Directions

• App integration: start app

• App integration: internal app functions such as add to Hulu

queue

• Answers sassy questions like "Are you sexy?"

• Play music by artist or genre

Siri, Google Now, Cortana

• How to cook enchilada

• What is the population of Brazil

• How much is 100 Euro in Czech Crowns

• What is the weather in Sydney

• Stock price Microsoft

• How old is Tom Cruise

• Who is the president of the Czech Republic

• Who is his wife

• When is the next New York Rangers match

Google Now – questions examples

Web Intelligence

Machine learning

Data mining

Behavior modeling

ČVUT FEL

Jan Šedivý


Recommended