Transcriptional Implicature

Using a Transcript to Reason about an Exemplar

Claus Huitfeldt, Univ. of Bergen

C. M. Sperberg-McQueen, Black Mesa Technologies

9 August 2017

http://blackmesatech.com/2017/08/MLCD/

Project context

What does markup mean? and how?
When markup means “T is a transcript of E”, what does that mean?
What is the propositional content of a transcript?
Can those propositions be formalized?
- If so, how?
- If not, why not?

Transcriptional implicature

Can we translate project's descriptions of transcription practice into formulae?

Of course. But results disappointing.

Gaps. Silent assumptions.

Transcriptional implicature: the set of inferences from T licensed by implicit rules.

Default transcriptional implicature

We can approximate:

Reciprocity: There is a one-to-one relation between normal tokens in an exemplar E and normal tokens in its transcript T.
Purity: Every normal token in T transcribes something in E.
Completeness: Every normal token in E is transcribed by something in T.
Type similarity: corresponding tokens in E and T have the same type, or non-identical but similar types.

The current work

Goal: put this account to a simple not-too-hard test.

Understand statement of principle.
Rephrase in terms of purity, completeness, type similarity.
Formalize.
Show that it works for obvious inferences. (Remember the 150 pages in PM devoted to proving 1 + 1 = 2?)

Newberry Melville

Northwestern-Newberry Edition of Herman Melville's Moby Dick includes two-page facsimile of Melville's notes in a volume of Shakespeare, with transcript.

List of symbols used.

The transcriber translated

Alt heading: Newberry Melville reformulated

“[”, “]”, “<”, “>”, “?” (before words), “xxxx” (etc.) are special tokens.
Inside [...], italic material is special.
Words with and without underlining are type-distinct in E; words in roman and italic are type-distinct in T; underlining in E and italics in T are type-similar.
Later additions are type-distinct.
Overlined words, overwritten words, other words are type-distinct.
“About n undeciphered letters” is a type, instantiated differently in T and E.
Words prefixed with “?” in T are type-distinct and are probably type-identical to their exemplars.

Formalization

(∀t t:tokens(T))(special-transcript-token(t) ⇔ (square-bracket(t)

∨ angle-bracket(t)

∨ prefixed-question-mark(t)

∨ xxx-sequence(t)

∨ (italic(t) ∧ within-square-brackets(t))

∨ (italic(t) ∧ within-angle-brackets(t))

∨ prefixed-question-mark(t)

∨ page-furniture-token(t)))

You didn't really think we were going to get through an entire talk without showing you any existential quantifiers, did you?

Test 1

To prove: The first line of the first page reads “A seaman figures in the Canterbury Tales”.

More formally: The line in E transcribed by the line in T labeled “1” has [that] string value.

<pre> line_number(L, "1"), % let L be the line labeled "1" exemplar_transcript(E, L), % let E be its exemplar type_token(T, E), % let T be the type of the exemplar </pre>

Test 2

To prove: The dots to the left of the first line are [or: are not] significant.

Results

Prolog formalization produces erroneous results.

White space snafu: String value of first line is (in v1) not “A seaman figures in the Canterbury Tales” but “ A seaman figures in the Canterbury Tales”

Add a rule? “Whitespace adjacent to impure tokens is itself impure.” (Is that in fact true? Is the whitespace after a line number part of the line number? or a different impurity? or ...?)
According to the rules,
[on verso of last leaf of Volume VII, page [524]]
means, in context,
The string value of the first line of the first page of E is ‘[524]’.

Hypothesis is false as formulated.

Not very false, just a little. We still believe it. But false is false.

Future work

Another example.

Then another.

Keep on until we stop being surprised.