Please enable JavaScript.
Coggle requires JavaScript to display documents.
voice recognition and speech synthesis (Finding a voice - The Economist…
voice recognition and speech synthesis
Finding a voice - The Economist
approaches to language technology
Neural-network-based approach (DNN)
qualities similar
to those of the human brain: “neurons” are connected in
software, and connections can become stronger or weaker
in the process of learning
Technology
GPUs: graphical processing units
TPUs: Google's Tensor Processing Units
rules-based approach
write rules to analyse the text of a sentence in the
language of origin, breaking it down into a sort of abstract
“interlanguage” and rebuilding it according to the rules of
the target language
“brute force” approach
application
of statistical methods
software scouring vast amounts of data,
looking for patterns and learning from precedent.
statistics-based approach
phrase-based approach
systems
speech recognition systems
text-to-speech engines
machine translation
phrase-based machine translation
speech recognition
problems/challenges
have a "true" conversation
machine translation proceed sentence by sentence
long sentences can be hard to translate
neural-net based systems struggle with rare words
training data are rare for many language pairs (e.g. smaller languages like Greek-Urdu)
computers don't "understand" the real world
privacy issues
disruption of jobs
involved companies
big player
Google
Google speech assistant
Google translate
Microsoft
Cortana
Skype
IBM
Watson
Apple
Siri
Amazon
Alexa
Other
Nuance
startups
Lilt
Smartling
Corticol.io
Datalingvo