[ATDrake]

From what I've read (mostly @ W3C and various related usage evangelists), it would be semantically correct to use <i> in situations where you're reproducing portions of text which are for some unobviously related reason (such as purposes of style or visual separation) italicized.

I've seen anthologies and poetry books where brief intro text and/or comments on the work were italicized, while the actual story/poem was in regular font style, to make it clear they were separately written. Also epigraphs and the like.

If the dialogue/narrative seems to be emphasized for reasons related to contextual meaning (foreign language phrases, someone being emphatic in their speech/observations, etc.), then <em> or <span> with style seems to be the way to go. (I freely admit that I'm lazy and rarely bother using either.)

<cite> for citations will give the same stylistic effect but greater semantic usefulness if one day they make better search engines which can take advantage of that.

Yes, <strong> goes where you would normally use bold, as being one step even more emphatic than italics.

Hope this helps.

ETA: I should add that a lot of the evangelism was rather tilted towards deprecating <i>, <b>,<u> etc. entirely in favour of completely semantic markup.

But a vague consensus seems to be that if you don't know/can't tell why the text was italicized/bolded/underlined in the first place, just leave it as is and don't try to second-guess the original typesetter's intentions; just maybe add a few class attributes so you can easily group and distinguish the italicized blocks from the italicized single lines, or whatever.