The 'Chinese' Theory: For and Against

[Jorge_Stolfi]

Jorge, I read your paper about the "rooster paragraph" and responded to it at length. (For those interested, my issues with supposing the text has been altered can be found You are not allowed to view links. Register or Login to view. and You are not allowed to view links. Register or Login to view., and my difficulties extending glosses can be found You are not allowed to view links. Register or Login to view.. My other comments thereabouts in the thread are salient too.)

Sorry if I did not respond to all those concerns; there were many posts on this thread on those days, and this "wonderful" forum software will mark unread posts 10, 11, 12, 13 as "read" if you read just 10 and reply to it.

But I hope that the evidence will eventually be enough to make those supposed weaknesses irrelevant.

At present, I am asking for the 11 new identifications so I can evaluate them. If they are not ready for presentation, I understand, but then they cannot be relied upon as proof for the rest of us

I have already checked more than 30 SBJ entries and found matches for more than 20 (the expected ratio).  But with varying degrees of confidence.

You are not allowed to view links. Register or Login to view..  The main information is at the very end.  

There are a number of "paragraph evaluation blocks" in that page. Their format is described below.  They were computed for all "good" (non-smashed) parags of the SPS, but only a few of those with best (lowest) scores are shown.

Most of that HTML page is generated by a bunch of python functions that I wrote over the past month or so.   But some parts had to be generated manually, including the detailed parsing of the SPS entry and its matching English translation, the identification of omitted fields, the hanzi keywords to use in the structural matching, the patterns for the possible Voynichese translations, etc.  And then the various "best matches" had to be inspected to see whether they look like real or just coincidences.

There are still some bugs and shortcomings in my functions and in the formatting of those pages.  I am working on that.

IMPORTANT: the alignment of the SPS parag words with the SBJ entry hanzi, shown at the very end of that page, is tentative.  Only the matching of the keywords is reasonably certain; for the words between the keywords, the correct alignment may be off by one word or more.  And even for the keywords there may be other possible assignments; that one shown is the one which gave the smallest badness score.
All the best, --stolfi

---

In that report there are a number of "paragraph evaluation blocks" that detail the comparison of the SBJ entry with a candidate parag of the SPS.


Each block refers to a specific parsing of the SBJ entry into "hits" (instances of the known keywords, like 主) and "gaps" (the strings before, between, and after those instances).

Each candidate SPS parag was tentatively parsed in the same way, into "hits" and "gaps", by looking for RE patterns that describe the possible Voynichese spellings of those same keywords.  For example, the pattern for the 主 keyword is 'daiin|laiin|dair' when trying for a match under "strict" criteria, or '[dlkrs][ao]iin|dain|[dlkrs][ao]ir' if that strict attempt fails and the search is repeated with more "liberal" criteria.

The first line of each block is the locus ID of the candidate parag's head line, like "f114.1", followed by a numric "badness score" that summarizes how badly that parag matches the SBJ entry in question.  More precisely, how different are the actual lengths of the gaps in the SPS parag from the values predicted from the corresponding gaps in the SBJ entry, assuming the ratio of ~5 EVA letters per hanzi. The score would be zero for a perfect match (all gap sizes match the perdicted values).

The second line has {N+2} numbers, where {N} is the number of specified keywords (such as '主') in that SBJ entry which were matched in the SPS parag.  Those numbers are the percent errors (deviation from predicted value ranges) of the parag's total gap length and of the lengths of the {N+1} individual gaps before, between, and after the identified {N} keywords.""")

The third line gives the actual total gap length and the {N+1} actual gap lengths.

The fourth line gives the predictions for the numbers on the third line. Namely, the gap sizes of the SBJ entry times 5. They are ranges rather than numbers to account for rounding errors.

The following {N+1} lines show the pasrsing of the SPS parag defined by keyword patterns.  The left column shows the "hits" (the actual substrings of the parag that matched the keyword patterns), the right column is the gap after each hit.

If the parag had multiple possible matches for the specified keyword patterns, the block shows the choices of the {N} matches that gave the lowest badness score.

All the substrings of the parag that matched the keyword patterns are highlighted in bold, even those that were not considered hits for score computation.
 
All sizes are in EVA characters.