One funny word on f114r.30

[Jorge_Stolfi]

One or two errors per paragraph or page could be explained by context.

For most purposes, one may define "error" as any token that differs from the most common form of the intended word type.  If an author usually writes "horce" when he means a horse, but at some point he writes "horse" with exactly the same meaning -- for any reason -- this is equivalent to an "error" for many purpoes.

One or two errors per page may be a high estimate for a manuscript written produced in a "manuscript factory", by a professional scribe, in a language that he understands and has settled orthography.  One or two errors per paragraph may still be too much if the scribe is working alone (like a scholar copying a book himself) instead of in a "factory".  I guess that most surviving Medieval manuscripts with running text have either of these two origins.

But there is reason to suspect that it is not the case for the VMS. There is evidence (not conclusive, sure ) that the Scribe did not understand the text at all.  There is also the large number of single-use characters (weirdos), the variations in the repetitive sequences of You are not allowed to view links. Register or Login to view. and f66r, the large variation in glyph shapes, and more.

I recently posted a quote from the diary of the Lewis and Clark expedition.  The English text has many spelling errors, even for common words like "breakfast" and "today", and many truncated or abbreviated words.  I did not count, but they may be quite a bit more than two per paragraph.  But the error rate goes through the roof when they try to transcribe Native names.  I recall about half a dozen different spellings of "buffalo" and of the name of the same tribe.

And the VMS text may not be a proper written language, but only a shorthand -- where some vowels and consonants are often omitted or conflated, in an inconsistent manner.

And the orthography may allow variation.  The writer may be free to use certain abbreviations (like "secũd9" for "secundus") or spell the words in full, at his whim. (I believe that m is usually an abbreviation of iin, for example.) 

Or the encoding of certain features like stress or tone may be optional, present only when the writer thinks that it would be necessary for proper reading.

Thus the rate of "errors" in the VMS could be as high as one in 10 words or more.  IF the rate is indeed that high, (1) attempts at analysis and solutions that assume an error-free text will fail, and (2) proving that an eventual solution is correct will require a LOT of work.

All the best, --stolfi