A Brief History of Language and Speech Technology

Flashcards covering the history of language and speech technology, including key figures, inventions, and concepts in speech recognition and synthesis.

Earliest Examples of Graphic Art

Cave paintings from the Chauvet Cave (France) that date back possibly to as much as 32,000 years ago.

Mesopotamia

A true representation system, purely functional in nature, was developed around 10,000 years ago, apparently to account for goods.

Tokens

Considered to be the precursors of present-day writing systems, they appeared as early as 8000 BC.

Isaac Newton

Observed in 1665 that one could produce a progression of vowel sounds using liquid poured into a flagon.

Christian Kratzenstein

In 1773, he provided the first description of a machine that could produce vowel-like sounds using resonance tubes connected to organ pipes.

Wolfgang von Kempelen

A Hungarian scientist who published Mechanism of Human Speech in 1791 with the description of his speaking machine.

Mechanical Turk

A chess-playing automaton featuring a turbaned Turk who played a strong game of chess against human opponents.

Joseph Faber's Speaking Machine

Exhibited in 1846, this invention was demonstrated with ordinary conversation and whispered speech, and was also able to sing.

Erasmus Darwin

Also invented a machine of a similar kind to Faber's speaking machine.

William Jones

In 1796, he demonstrated that Sanskrit was related to Latin and Greek, laying the groundwork for historical linguistics.

Melville Bell

Alexander Bell’s father, who made significant contributions to the study of phonetics, including a phonetic transcription system that he called Visible Speech.

Henry Sweet

Founded the modern study of phonetics.

Homer Dudley’s Voder

Exhibited at the 1939 World’s Fair in New York and San Francisco.

Printing

The earliest such technology (Gutenberg, around 1440 - 1450).

QWERTY Keyboard

The arrangement of keys is universal in every English-speaking country, and practically universal in all countries where the Latin alphabet is used.

Morse Code

The first electronic code for representing text, invented in the 1830s, demonstrating that signals could be transmitted electronically by wire.

Automatic Speech Recognition (ASR)

The machine equivalent of the speech process in the human mind.

Word Error Rate

The percentage of words the system gets wrong.

Acoustic Model

Converts physical waves into a sequence of acoustic features, and then converts these features into phones.

Pronunciation Model

Used to figure out what word sequences might have been said from sequences of phones.

Language Model

Used to figure out what word sequence was actually most likely, given what the speaker is probably talking about.

Adaptation Algorithms

Algorithms that allow the models to adapt quickly to the speech of new speakers and to new dialects.

Frames

The waveform is first cut up into a series of short time segments.

Hidden Markov Model (HMM)

At the core of most speech recognizers used today.

Speech synthesis

Has to do with modeling the process of producing speech from linguistic input.

Speech Synthesis

Denotes the technology for getting computers to talk from any kind of input.

Text-to-Speech Synthesis (TTS)

Where the input is text and the goal is to read the text the way a native speaker does.

Unit Selection Synthesis Systems

Synthesizing units rather than synthesizing single sounds.