home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!stanford.edu!bcm!convex!darwin.sura.net!paladin.american.edu!auvm!CCB.BBN.COM!BNEVIN
- Message-ID: <CSG-L%92111013540627@VMD.CSO.UIUC.EDU>
- Newsgroups: bit.listserv.csg-l
- Date: Tue, 10 Nov 1992 14:48:17 EST
- Sender: "Control Systems Group Network (CSGnet)" <CSG-L@UIUCVMD.BITNET>
- From: "Bruce E. Nevin" <bnevin@CCB.BBN.COM>
- Subject: Hamming it up
- Lines: 68
-
- (John Gabriel 921109:16:45 CST) --
-
- > A perception of the real world can be put in terms of a set of
- > propositions believd to be true (Aristotle - or if you want
- > something more modern, Pattern Recognition, Human and Mechanical -
- > Satosi Watanabe, pub John Wiley 1985).
-
- Henry Hiz wrote some interesting papers on this socalled "aletheic
- theory of truth" in the 1960s or '70s. The direction I took it in my MA
- thesis at Penn (1970) was to build an acceptability model of
- propositions rather than a truth model. Those operator-argument
- combinations actually encountered in utterances (some of them sentences,
- some not, Penni) accepted by my desired peers are ipso facto acceptable.
- Others are acceptable by analogy over the classifier vocabulary of the
- subject matter domain. The simplest case: (a) substituting a
- classifier word Ncl for a word actually occurring in a given utterance
- and (b) substituting for the classifier word Ncl each of the words that
- is acceptable in an utterance of the form N is an Ncl. This corresponds
- to some of the sublanguage analysis work of Sager, Grishman, Hirshman,
- and others. The more remote the analogical extension, the lower the
- acceptability or the expectation of encountering or producing a
- particular operator-argument dependency. From these differences follow
- the enabling conditions for the reductions that make up most of the
- messiness of language.
-
- > The distance between "your" perception and "mine" is a Hamming
- > distance(*), possibly weighted by value systems for the propositions.
- > Note that you and I can disagree about the magnitude of the weighted
- > Hamming distance because of differing value systems, unless we
- > agree on everything in which case the Hamming distance is zero
-
- The only way that you can arrive at a Hamming distance of zero in this
- sense is if you (the observer arriving at the Hamming distance of zero)
- occupy an external point of view able to look inside each of the two
- parties and inspect their perceptions from the outside. In general, the
- Hamming distance can be no more than the count of propositions that one
- person affirms and with which this person perceives that the other
- disagrees. There are ways of negotiating and testing agreements, of
- course, but they are never complete.
-
- Also, in PCT you are dealing with perceptions, not with propositions.
- (Propositions are also perceptions, that is, the linguistic perceptions
- whose control results in speech, but it is nonverbal perceptions that I
- assume are the object of agreement and disagreement here.) The division
- into propositions and "weighting by value systems" is probably at least
- in part an artifact of this confusion. If you are representing
- perceptions in your model by propositions, then it isn't a model of
- perceptual control (a PCT model).
-
- > In this framework, the PCT model has an error signal that is the
- > weighted Hamming distance,
-
- You mean in your model there is an elementary control system with a
- perceptual input signal corresponding to the number of propositions
- about which there is disagreement, weighted in some way according to the
- values associated with the propositions in a value system, and a
- reference input signal of zero, such that any perceived disagreements
- result in a non-zero error output signal, right?
-
- >and the purpose of communication,
- > diplomacy, or in the last resort war, (von Clausewitz) is to
- > change that distance either by change of perceived truth or
- > change of (even imposing a different in case of war) value system.
-
- This says that people control for agreement. Or am I missing something?
-
- Bruce Nevin
- bn@bbn.com
-