Students answer problem sets by label: 1(a), 1(b), 2. Graders mark them the same way. A conversion that runs the parts together into one paragraph keeps every word and loses the assignment.
In our example, question 1 asks whether three series converge, as parts (a), (b) and (c), and question 2 stands alone. On the web page the parts are a list inside question 1, each keeping its printed label. A screen reader announces a list of three items, then reads each label before its formula. Part (b):
Heardthe sum from n equals 1 to infinity of the fraction with numerator 1 and denominator the square root of n end fraction
The reading gives the structure before the details: a sum, then a fraction, then what sits in its denominator. A student listening knows the shape of the problem before working it, which is what a glance at the page gives everyone else.
The PDF keeps the same structure in its tags, so a screen reader in a PDF viewer finds a list there too, not lines that only look like one. The problem set's PDF and web page are on Slides and problem sets.