Semantics is the study of meaning. Meaning is something outside the signs we use, some concept on what signs refer to. By saying something we mean something that can be no sign. Whether we are able to refer to the reality by this is a controversial matter. In every case we refer to a concept or idea or imagination of something. This reference makes language an extreme flexible sign-system. We can refer to something from several perspectives and we can describe relations between concepts in various ways. We can even adopt several systems of concepts, as for example ideologies, and mediate between them.
Even in the world of computers, signs mean something. This meaning might be implicit. To a certain extend a computer can understand such a meaning. But for this it has to have some reference-system of concepts and their relations. If I say: “We want a four star hotel on the beach, with an excellent cuisine.” The computer, in order to understand it and respond to this, has to have concepts of “beach”, “hotel”, “cuisine” and that they should be an aggregate offer at the same place. If it can relate to this concepts, it can search a database and create the right combination of this offer. Of course, on this simple level there may be misunderstandings and irrelevant hits. But it is, for example, able to learn, what people usually book, when they write this. So it is easy for it also to understand homonyms and synonyms and slang expressions.
In the present world wide web more than 90% of the websites do not say themselves what they mean. Their content is interpreted by the users and also by machines like search engines. Human interpretation is easy, because web-pages are optimized for this. Machine interpretation is much more tentative. “Cuisine” can mean some style of food but in can also mean some place where the food is cooked. By “excellent cuisine” the furnishing of the kitchen can be meant, which is no guarantee that you get excellent dishes. Semantic web, for web publishers, is all about making clear what you mean to the machine. In return the machine can give the user much better proposals, search hits and can even find out what they mean even if they don’t exactly say it.