[tei-council] certainty revised

[Martin Holmes]

Hi Lou,

>> --------------------
>>
>> The example of the abbreviation expansion of SGML looks like this:
>>
>> You will want to use
>> <choice>
>>   <expan xml:id="CE-e1">Standard
>>     Generalized Markup Language</expan>
>>   <expan xml:id="CE-e40">Some Grandiose Methodology for Losers</expan>
>>   <abbr>SGML</abbr>
>> </choice> ...
>>
>> <!-- ... -->
>> <certainty target="#CE-e1" locus="value" degree="0.9"/>
>> <certainty target="#CE-e40" locus="value" degree="0.5"/>
>>
>> I suspect that the second @degree value should be "0.1", shouldn't it? 
>> Don't the two have to add up to 1?
>>
> 
> Um, no I don't think so. They are independent probabilities, I think. 
> Unless @given is used, of course.

Aren't they mutually exclusive? And if so, shouldn't their probabilities 
add up to 1? (Apologies if I'm being dumb here -- I'm completely 
ignorant when it comes to stats.)

>> ---------------------
>>
>> Finally, I find myself a little puzzled by the application of 
>> <precision> to @notAfter, @notBefore, etc. My habit in using these 
>> attributes has been to decide once and for all, based on known 
>> information and logical deduction, what the _actual_ earliest and latest 
>> possible dates would be. It seems a little odd to me to say that 
>> something is "not after 1857", but there's a 50% chance that it might 
>> be. If you're not sure it's @notAfter="1857", then you need to move your 
>> @notAfter value forward until you can be sure, don't you?
> 
> That would seem like a sensible practoce to me too. I observe only that 
> people like to be more nuanced (or vague) sometimes.
> 
> 
> In my
>> experience, there always _is_ a genuine value for @notBefore or 
>> @notAfter -- in the case of an action by a person, for instance, the 
>> date before which they were definitely not born, the date after which 
>> they were definitely dead, or the point at which they would have been so 
>> old as to break all known longevity records. After all, @notAfter is 
>> defined as specifying "the latest possible date".
>>
> 
> yes, but....
> 
>> One example from the text:
>>
>> <residence from="1857-03-01" notAfter="1857-04-30">From the 1st of March to
>> some time in April of 1857.
>> <precision match="@notAfter" degree="0.5"/>
>> </residence>
>>
>> Now, if "some time in April" is true, then @degree should surely be 
>> "1.0". 
> 
> No, the precision implies only to the @notAfter value
> 
> 
> If "some time in April" is doubtful, then surely so is "the 1st
>> of March". 
> 
> Why? You might have documentary evidence for the 1st of March, for example.
> 
> But if there is doubt, that doubt presumably originates in
>> external evidence, which ought at least to be adduced, and which itself 
>> would presumably give a more realistic value for @notAfter.
>>
> 
> Or it might not!
> 
> I think this example is quite plausible as it stands, but would you find 
> it more convincing if the @notAfter were replaced by a @to ?

I think if you find it OK, then we're simply differing with regard to 
the way I prefer to use these attributes versus the way other people 
might prefer to use them. I'm a bit literal-minded, so to me notAfter 
means "not after"; if it's perfectly OK for it to mean "probably not 
after", then I have no quibble.

Cheers,
Martin