Machine Learning: 2 Books in 1: Machine Learning for Beginners, Machine Learning Mathematics. An Introduction Guide to Understand Data Science Through the Business Application

Speech Recognition

---

Improvements in artificial intelligence have made speech recognition more
useful very recently. Most of our smartphones now have some level of
speech recognition ability, which involves machine learning. Speech
recognition takes the audio data we give it, and it turns it into text that can
be interpreted.
The difficult thing about speech recognition is the irregularities in the way
that people speak. Like intra-class variability. You and I may have different
accents and different inflections that are hard to account for when you are
teaching a computer how to understand the human voice. If we both say the
same word with different accents, how do we teach the model to understand
us?
Speech recognition also uses neural networks to interpret data, like image
recognition. This is because the patterns in audio data would probably not
be recognizable by a human. Data scientists use sampling in order to
interpret data and make accurate predictions despite the variances in
peoples voices. Sampling is done by measuring the height and length of the
sound waves, which believe it or not can be used to decipher what the user
is saying. The recorded audio is converted into the wave map of
frequencies. Those frequencies are measured by numerical values and then
fed through the neural networks hidden layers to look for patterns.
Medicine and Medical Diagnosis
Machine learning is not just useful for digital marketing or making
computers respond to your requests. It also has the potential to improve the
field of medicine, particularly in the diagnosis of patients using data from
previous patients.

---

With as much potential as machine learning has for medical diagnosis, it
can be challenging to find patient data that is available to use for machine
learning because of the laws surrounding patient privacy. Its gradually
gaining acceptance in the field of medicine, which means data is becoming
available to data scientists. Unfortunately, up until now, it has been difficult
to have enough meaningful data to be able to make models regarding
medical diagnosis. But the technology is there and available to be used.
Machine learning could use image recognition to diagnose x-rays by taking
data from several patients to imaging scans in order to make predictions
about new patients. Clustering and classification can be used to categorize
different types of the same disease so that patients and medical
professionals can have better a better understanding of the variation of the
same disease between two patients, and their likelihood of survival.
Medical diagnosis with machine learning can reduce diagnosis errors made
by doctors or give the doctors something to offer them a second opinion. It
can also be used to predict the probability of a positive diagnosis based on
patient factors and disease features. Someday, medical professionals may be
able to look at data from thousands of patients about a certain disease to
make a new diagnosis.
But medical diagnosis is just one of the numerous ways that machine
learning could be utilized in medicine. Medical datasets remain small today,
and the science of machine learning still has a lot of unmet potential in the
field of medicine.