The problem with the word "truth" is that it have multiple meanings in different contexts which are very commonly confused, even by philosophers.
To get started, our common-sense intuition about where we should go look for truth is in logic. Logic is the study of consequence of assumptions. In other words, in logic* you assume you have certain axiom and you can then derive consequences of those axioms and other propositions. The consequences are then said to be "true". For instance if you have accepted that (A) socrates is a man and (B) all men are mortal it is then true that (C) socrates is mortal. This is the most fundamental characterization of true and the one we are all likely to resort to.
The problem is this will not do in terms of answering Ponteus question. certainly, *we* can say that: "C is true", however *logic* (as discussed up to now) is not expressing this relation, therefore *logic* cannot express the sentence "C is true" (under the above definitions). This distinction is quite technical, however it is one many philosophers are prone to re-discover in various forms and point it out as a fundamental problem of defining truth. To analyze a sentence like "C is true" (that is to say, to have a formal system which expresses a proposition is true), you need something like Tarskis semantic theory of truth which is quite technical. This would however be my answer to Ponteus: Read Tarski!
Now, a problem which now occur is none of this corresponds to our common-sense intuitions about truth in the real world. For instance suppose we flip a coin 1000 times and it comes up heads 1000 times. Would we conclude it is true the coin is biased? There is no principled way to derive this conclusion using logic** (we could just have been very unfortunate!), and so it is easy to resort to all sorts of odd speculations like there is no truths about the real world etc. etc. What occurs here is we should really ask for our *degree of belief* in the proposition "the coin is not biased"; this is a concept that can be analysed with some rigor (c.f. the Bayesian interpretation of probabilities) giving us a framework that allows us to give definite meaning to a statement such as: I have a very high belief the coin is biased because it just came up heads 1000 times out of 1000 flips.
A second problem is most of the time when we consider the real world, we are not really interested in "truth" (in a technical sense) but good models. For instance newtons laws are not "true" in any strict sense (they give wrong predictions in some instances), however they do provide a very good and valuable model for reality. It is for this reason I am a bit skeptical with a statement such as: "truth is what corresponds to reality", since it seems this both ignores the most rigorous definitions of "truth", rules out very good models of reality which nevertheless are not strictly corresponding to reality all the time (newtons laws, the theory of relativity, etc.) and introduces the additional problem that we do not have access to "what corresponds to reality". For instance how do you determine if it "corresponds" to the reality the coin from before is biased?
I think we are better off to say: "Since we are talking about the real world, we can only talk about what provides a good model for reality, and specifically the degree in which we can believe in them, and that is then what I am going to talk about. Feel free to call that "truth" or not if you like, I am off to compute the orbit of the moon".
+++
* This is not the only characterization of truth in logic.
** well ++technical stuff.
