MAKING SPEECH RECOGNITION LESS COMPLICATED
A BASIC KNOWLEDGE ABOUT SPEECH AND THE PROCCESS OF RECOGNITION WOULD BE NESSECSERY BEFORE YOU CAN FULLY UNDERSTAND ANYRYZE PLATFORM
WHAT IS SPEECH RECOGNITION
Speech recognition is the ability of a machine or program to identify words and phrases in spoken language and convert them to a machine-readable format. Rudimentary speech recognition software has a limited vocabulary of words and phrases, and it may only identify these if they are spoken very clearly. More sophisticated software has the ability to accept natural speech.
How it works
Speech recognition works using algorithms through acoustic and language modeling. Acoustic modeling represents the relationship between linguistic units of speech and audio signals; language modeling matches sounds with word sequences to help distinguish between words that sound similar.
WHAT ANYRYZE IS DOING TO MAKE THIS PROCCESS LESS COMPLICATED
Over the past 50 years people tried to make computers understand human lan-
guage in order to make them more responsive to our needs and helpful, and to
create a bigger amount of useful features. Starting from Bell Labs and IBM exper-
iments, the greatest value in speech recognition was introduced by Google with
the innovative Speech API. But all these solutions were built on the principles from
50 years ago and soonly will reach their limits. Thus, now it is time to embrace the
brand new solutions and take computer speech recognition on the next level.
Why speech recognition matters? As soon as we educate computers to understand
natural human language, it will become a revolution in a different areas of work.
Developing, educating, and creating new neural network models is vital nowadays,
because of the possibility to make computer actually think like a human being. The
accurate speech recognition is the very first step to it. After all, this is how we com-
municate. Our mission is to create the best Speech Recognition System: highly ed-
ucated, thinking like us, understanding all our words and intention, and while doing
that, we want to get as many people as we can to participate. Afterwards, we will be
proud to say that this is the collaborative act aimed on the education of the Com-
puter that became possible not only due to the efforts of the small group of scien-
tists but also with the contribution of usual people.
Speech is a complex phenomenon. People rarely understand how is it produced
and perceived. The naive perception is often that speech is built with words, and
each word consists of phones. The reality is unfortunately very different. Speech
is a dynamic process without clearly distinguished parts. It’s always useful to get a
sound editor and look into the recording of the speech and listen to it.
Description of speech is at some degree probabilistic. That means that there are no
certain boundaries between units, or between words. Speech to text translation is
never 100% correct. That idea is rather unusual for software developers, who usually
work with deterministic systems. Therefore, it creates a lot of issues specific only to
speech technology.
Recognition process
The common way how Anryze recognizes speech is the following: we take the
waveform, split it on utterances by silences then try to recognize what is being said
in each utterance. To do that we want to take all possible combinations of words
and try to match them with the audio. We choose the best matching combination.
There are few important things in this match.
First of all, it is a concept of features. Since a number of parameters is large, we
are trying to optimize it. Numbers are calculated from speech usually by dividing
speech on frames, then, for each frame of length typically 10 milliseconds we ex-
tract 39 numbers that represent the speech. That’s called feature vector. The way
to generate numbers is a subject of active investigation, but in simple case it is a
derivative from spectrum.
Second, it is a concept of the model. Model describes some mathematical object
that gathers common attributes of the spoken word.
