The setting is a typical research talk. Your room is filled will reasonably intelligent people but not everyone is exactly in your specialty. (Think conferences.)
For such a talk, I am a strong opponent to giving citations using the standard alpha style (as in BibTeX). It’s at most three alphabets followed by two digits, like this: [AMR03].
Maybe it’s just me, but I’ve met some speakers that would say “A M R Oh Three proved a seven-third plus four epsilon over three approximation…” Now that’s really helpful to the rest of us, isn’t it?
And if the speaker is any good, she would actually say the three lastnames in full. I don’t want to offend anyone, but my experience is that a significant fraction of people in the room would have never picked up the words Amir, Krauthgamer and Rao from *my* speech. It’s not just because of my strong accent (and I confess), but also the fact that sometimes we don’t really know how to pronounce a name properly.
My believe is that we should just spell out the complete last names on at least one of the slides. And if none of your slides has enough space for spelling out [Amir-Krauthgamer-Rao 03], then maybe your slides are a bit too dense?
P.S. In case you are interested, [AMR03] is Constant Factor Approximation of Vertex-Cuts in Planar Graphs in STOC. The classic result in vertex-cuts for planar graphs is the 1979 Lipton-Tarjan `O(sqrt n)` separator theorem. But there is no guarantee that their algorithm will return a cut that has a size close to the optimal. This paper achieves that guarantee, with some technicalities.
