I don't know what is the proper name for what you're seeing, but it is the standard way to lexically analyze text. You divide it into tokens of specific types. For the sake of context-free parsing (the next step in the parsing chain), you only need the type of each lexeme; but further steps down the road will need to know the semantic content (sometimes called annotation) of each lexeme. For example, the context-free parser doesn't care which number $2$ is; it only needs to know that it's a number. Further down the road, it will suffice to know that this number is an integer. Even further, you would need to know that this integer is specifically $2$.
Some lexemes are atomic, for example the various operators (though in principle we could group, say, + and - together). We need to separate them because syntactically = is different from +, and + is different from * (due to operator precedence). These don't need any annotation.