# Phoenetics
Study of sound in human speech : physical properties, production, acoustics, articulation, etc.
# Phoneme
phoneme is the smallest sound segment that can be used to distinguish one word from another.F
US English consists of 39 to 44 phonemes

# Grapheme
a grapheme is the smallest symbol that distinguishes one written word from another. In US English, 26 letters and a space combine for 27 possible graphemes.
It is not simple to map phoeneme to grapheme , because a letter can sound different.

You can build a system that maps audio to phonemes.
With enough data, you could have your acoustic model map to words directly.

Overview of Language Model Pipeline

From an acoustic model, it might be hard to differentiate between hear and here.
Words produced by accoustic model are not absoulte choices and might need to be changed on language model.
Taking the probability of the whole sequence, considering all words could be large.

Languaage model might need only the last 3-4 words.
Repeated multiplications of small probabilities can cause underflow problems in computers when the values become to small. To solve this, we will calculate all probabilities in log space:

section 20