Addressee1, Addressee2, ..., AddresseeM: This is the message body. -- Sender1, Sender2, ..., SenderN
In this way, the fragment is readable by humans and yet has a probability of matching a pattern that a computer can recognize.
On the other hand, this violates the principle of using natural language to convey natural meaning. Moreover, conversations that are transactional are less interesting than a unified, meaningful work. Also, if one wants to enforce structure to make life easier for computers, it's probably better to use limiting technology. WebLogs are good for this kind of exchange. Wikis are not as good because the text is so free.
Then again,, if you really want to emphasize computer parsing, say for DataMining, this could be useful. But an XML system is probably better as you can quickly retrieve syntactic and semi-semantic information. Also, a probability of matching is not good enough for computers at the current state of the art. The best structuring features on a wiki revolve around the LinkPattern as it one of the few structures enforced by the underlying technology.
Finally, one should note the thermodynamic law of human laziness. If (it's clear that) there's no benefit for doing extra work, the extra work will not be done. ContentOverForm!