How Does a Computer Learn to Read?

Believe it or not, computers are illiterate. Despite appearances, a computer can’t “read” a word, understand it’s meaning, and place it in context next to adjacent words. To a computer, the words you read on the monitor are just a way of transmitting data in a way that humans can understand. The computer itself is operating on a much lower level of abstraction, because it can only store information as bits — ones and zeros that must be strung together to represent something more complex.

Consider the letter a. A computer doesn't have any way to store an a in memory, because computer memory is a string of switches that only have two states: on (1) or off (0). So, the computer has to find a way to represent the letter a with those switches. To solve this problem, humans have created various systems that match letters and symbols with decimal numbers. In ASCII for example, a is matched to 97. Now, instead of the realm of letters, we're in the realm of numbers, and computers are great with numbers. A 97 can in turn be converted into binary format rather than decimal format, and we get 01100001, which can be represented by the aforementioned switches in computer memory. To humans, this means a, but as far as the computer is concerned, letters don't exist.

Setting aside the intricacies of digital storage, let’s consider why this matters in text analytics. Although…