[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [Elist Home]
Subject: possible Base anchor Idiosyncrasies RE: [humanmarkup-comment] FirstWorking Draft of HM.Requirements
Avoiding MUST here is a logical way to keep HumanML from having declared itself invalid in the likely case of incompatible conventions coming from elsewhere. (In my Notes version, wording was softened to "MUST be designed to", I believe.) There may be cases where it would be poor design to adopt some "standards" or counter-productive to have to change ours because someone else changes theirs. (Prior discussion noted this also.) Further, as Len points out, "pure physical characteristics" are not always germane to the significance of communications in a given interpretive context. At 01:44 PM 25-03-2002 -0600, Len wrote: >"The Primary Base HumanML Schema MUST include a HumanIdentifier Element that is compatible with and interoperable with [many] standards, ...." > >...identification ... relies not much on pure physical characteristics .... The process ... >is typically matched to the need .... [And various, changing systems are in use.] > > >So our human identifier code elements have to be fairly limited in the primary schema. As a feature at the base level perhaps IDIOSYNCRASIES would do. Special-needs feature specs would then be slottables at the Secondary level. As example, at Secondary level we might have the Name(s) of communication agents. Per provenience of the communique and interpretive purpose, Name(s) could including Form(s) of Address style names, such as relationship-based ones like Son, Mom, Boss, and Sir, not to mention telnet addresses. At a lower level a rich set of options for different applications might be pulled in, as Len notes. If these specs borrowed in were not totally compatible with each other, the user can adjust as needed without jeopardizing the HumanML itself. >Physical characteristics for LE systems will also be a >small set and not nearly as detailed as they would be for say, medical applications and perhaps even the VR systems. > >Even the seemingly obvious and simple means such as recording names become complicated if >inherited in a module that has a cultural >component. Anthropometric means are better and biometric means are best. > >len For many genres, there is no single speaker: consider media news or folk tales. No personal name. There may be, however, significant facts about the provenience of the communique. A place in the Base Schema is needed for attaching whatever info there is. HumanML also needs built-in flexibility for dealing with mixed-agent systems. Some of the communication Agents in Human-Computer Interactions such as we aim to deal with might have physical existence but NO biometrics. Yet like live beings, they may have idiosyncrasies relevant to communication that we want to characterize in like wise. For instance, all the communication agents may have characteristic times-of-reaction. (Variations in turn-taking intervals in different types of conversation in different groups are highly significant. Lengthy times in CHI may be differently interpreted than in geolocal human groups, though. Artificial agents may have voice profiles with parameters affecting significance of message components. There may well be idiosyncratic writing style(s) or rendering to be characterized. (Cf. HTML, Unicode name adaptation.) Suppose we put something on the order of Idiosyncrasies ("Metrics" seems too vague to me -- "Provenience"? "Eccentricities"? ;) at the Primary level. Then, expansion into communication detail appropriate to the particular expression and interpretation circumstances and purposes would be done in the Secondary level. And "identifier code elements ... be fairly limited in the primary schema", as Len says. SC
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [Elist Home]
Powered by eList eXpress LLC