C. E. Shannon in A Mathematical Theory of Communication (Bell System Technical Journal, 1948) says:
The fundamental problem of communication is that of reproducing at one point either exactly or approximately a message selected at another point. Frequently the messages have meaning; that is they refer to or are correlated according to some system with certain physical or conceptual entities. These semantic aspects of communication are irrelevant to the engineering problem.” (emphasis added)
Avoidance of the “semantic aspects of communication” remains the most popular position. Think about it. What are the common responses to heterogeneous data (semantic “noise”)?
- Let’s use my semantic (or, lacking the power to insist),
- Let’s use a common semantic.
Both are homogenization of semantically heterogeneous messages. A “McDonald’s” version as opposed to having choices ranging from Thai to Southern Barbeque (BBQ, Bar-B-Q, Bar-B-Que). Not only is there information loss, the results are bland and uninteresting.
Semantic homogenization is not the answer. Semantic homogenization is the question. The answer is NO.
I agree with Shannon, except for the minor quibble that when he writes “irrelevant to the engineering problem” I think he must have meant to write “the only interesting engineering problem.”
Comment by sam hunting — March 3, 2010 @ 1:11 pm
Paraphrasing Patrick Durusau, the usual choice is
(1) Everyone should use the same terms I want to use, or, failing that,
(2) We should all use “standard” terms.
This sounds a lot like the truism that “The only reason to participate in the development of standards is that you’re not the market leader.” Hard to argue with that, and it dramatizes the fact that the designs of standards are far more responsive to the strategic interests of vendors than to the public’s obvious interest in lowering all barriers to communication.
Standards whose designs serve the public interest, like SGML (now known as XML) face an uphill battle in the real world. SGML took the radical position that one size (one taxonomy, one vocabulary, etc.) does not fit all, and therefore that it should stay the hell out of semantics entirely. (And XML, considered as a syntax, still does exactly that. Considered as an aggregation of semantic-bearing XML vocabularies, however, it doesn’t. But all that barrier-raising stuff can be ignored, and to a perhaps remarkable extent, it *is* ignored.
Comment by Steve Newcomb — March 3, 2010 @ 3:55 pm
I believe Shannon meant exactly what he said. It is a counterpart to Quine’s injunction to deal only with the extensional in logic, not attempt to deal with the intentional.
Shannon observed that the issue is to preserve the (representation of) the message, making it unnecessary to understand it well enough to preserve the sense of the message instead, or even know the sense of the message. That left him with a tractable engineering problem having to do with error correction and the information entropy of the representation. He’s saying to deal with that he doesn’t have to know what the users intention is, because he is going to preserve the bits and let the parties at each end worry about their communicating. [Enter favorite quote from “Cool Hand Luke” here.]
Comment by orcmid — March 7, 2010 @ 5:30 pm
If you look at Fig. 8 in Shannon’s 1948 article, he posits correction data being collected between the source and transmitter as well as between the receiver and destination. So both the encoder/decoder are sources of “noise” as well.
Shannon is treating the encoder/decoder (transmitter/receiver) as parts of the channel, we know that is a naive view of communication. Each of us as encoders/decoders introduce noise into any communication.
Consider Shakespeare’s Twelfth Night for example. (Or the Wikipedia Twelfth Night for the impatient.)
A. C. Foskett’s The Subject Approach To Information talks about how our encoding/decoding adds semantic noise to communication.
Comment by Patrick Durusau — March 13, 2010 @ 5:22 pm