`atWord` wasn't used at all and spaces should be eaten by the lexer And added `unspecified` token
QueryLexeter does not parse "atWord" because I couldn't find what it is used for.