Sounds and Words as Data

Close your eyes for a moment and listen. Maybe you hear traffic outside, a fan humming, someone talking. Your brain knows what each sound is because you have heard similar sounds many times before. Machines can learn to recognize sounds and words too — but first those sounds and words have to be saved as data.

Saving Sounds as Data

Sound is made of vibrations in the air — waves that travel into your ears. A microphone can capture those waves and turn them into numbers that a computer can save and read. Think of a recording of your voice. The microphone listens very fast — thousands of times per second — and writes down a number each time. That long list of numbers IS the recording. It is sound data. A machine can study millions of voice recordings and learn to understand spoken words. That is how voice assistants work — they have been trained on enormous amounts of sound data.

Words written on a page are data too — text data. Every word you type in a message, every sentence in a book, every caption under a photo — all of that is text data that a machine can study. Machines trained on billions of words can learn a lot about language. They learn that 'happy' and 'joyful' mean similar things. They learn that a sentence usually ends with a period. They learn what kinds of words go together. The machine does not understand words the way you do — but it finds patterns in the data, and those patterns are very useful.

Here is something cool: machines can work with all these kinds of data together. A video is image data AND sound data at the same time. A social media post might have text data, image data, and even the time it was posted — number data. The more kinds of data a machine can read, the more it can learn about the world.