The 'Chinese' Theory: For and Against

[Amphigorist]

Dr. Stolfi, I was wondering what you saw as the function of the finished manuscript in the author's mind. Clearly it was not just a set of personal notes, or there would be little point in going through the expense of illuminating them. But they aren't immediately useful to anyone else. Do you think the idea was to use them to educate students with the author present to translate for them? That is the best use I can come up with, but I am mo expert on medieval education practices.

[tavie]

Of course not.  Bayes explains.  They are all so convinced that the language must be European (or gibberish, or some non-linguistic code) that they will dismiss any evidence to the contrary -- for any nit they can find, or even without looking at the evidence at all.

It really saddens me that you keep saying this kind of thing.  Every solver says something on those lines when we don't accept their solution.  Yes, absolutely everyone else is biased.  Or maybe you haven't provided enough evidence objectively speaking to convince anyone? 

JoJo, forgive me for repeating the advice I gave on another thread:  character-level statistical analyses (character and digraph frequencies, entropies, correlations, etc.) are very likely to be a waste of time.  

One of my character-level statistical analyses led you to remove a wrong criterion for identifying paragraph breaks in the Stars section.  Not a total waste of time, surely?  And when you draw conclusions about final m replacing final n, or p is k on the top row, is that really based on a word-level analysis and not a character-level one?  It's when one goes beyond the individual glyph and look at clusters, if not full word types, that problems with glyph mapping tend to emerge.  And I say that as someone who has feverishly mapped virtually every glyph to another at some point.

[Jorge_Stolfi]

That’s ONE point of view [conviction that the language is European] which, for example, doesn’t apply to me at all.

That may be your case, and that of a few others (thanks for that).  But it is evident that many people reject the Chinese theory because they think that an origin outside Europe is just absurd.

But so far, you haven’t really done that. I consider the Chinese version a possibility, but unlike you, I find the evidence rather weak—at least for now...   As I said before: Convince us with more matches and details.

I agree that the evidence I presented is not enough to convince everyone, and I owe more matches (as I promised already).

I am delaying that because if I post matches that are too uncertain or require too many aapproximations (or are just wrong), they will have the opposite effect.  

The Rooster recipe is a good example because it and parag f105v.32 are extreme outliers in the size distribution; so, if SPS is indeed a transcription of the SBJ, that parag should be te transcription of that recipe; there is no other possible candidate.  And luckily it has eight instances of the keyword 主, so the position matches with EVA versions (six daiin, two kaiin, one dain) are hard to explain away as meaningless coincidences.

But there is another (almost) uncertain pairing at the other end of the size spectrum. The (almost) shortest SBJ recipe seems to be the Flying Squirrel (FLYS) one:
  鼺鼠：［主］堕胎，令易产。［生］平谷。
The translation, according to Gemini, is something like
  Flying squirrel: [Mainly for] expelling the fetus, making labor easy. 
  Provenance: flat valleys.
Excluding the [Provenance] field (last three characters), which apparently was systematically omitted from all recipes in the SPS transcription, the FLYS recipe is only 8 hanzi long. The matching parag is therefore expected to have ~8 x 5 = 40 EVA letters.  Among the available candidate parags, there is one (f112v.35) that is 34 EVA letters long; the second-shortest one is 74 letters long. So that must be it.   It does not have a daiin at the proper place, but it has a dain:
     
(The font used by my plotting function is missing the character 鼺 which should be just above 鼠 in the plot.  Please medidate intensely until you see it with your mind's eye.)

Agreed, that is not a convincing example.  But there is another crib in that recipe: 令 = "makes", in the second item of the uses list. I am not sure yet what is its "canonical translation" in Voynichese, but based on other tentative matchings it seems to be cheody, cheydy, or some variation thereof.  And indeed f112v.35 has a chedy at just the right spot:
   

There is a complication, though: according to Gemini there is an even shorter recipe in the SBJ, with only 6 hanzi:
  铁：［主］坚肌，耐痛。
  Iron: [Mainly for] strengthening muscles, pain endurance.
(According to Gemini, that is not some iron supplement, but warm iron blocks to be applied externally to muscles.  Se non è vero è molto ben allucinato.)

The only applicable crib I have for that IRON recipe is 主, and by that criterion IRON it is a fair match to f112v.35 too:
   

So, is parag f112v.35 the transcription of IRON, or of FLYS? 

That parag has 8 EVA before the dain.  That is 2 letters shorter than expected from FLYS entry, which has has two hanzi before the 主.  But that is 3 letters longer than expected from the IRON entry, which has a single hanzi before the 主.  

I believe FLYS is the right one, because, among other things, another SBJ file (not from the ZHB) has IRON as a mere sub-entry of a larger entry that includes the entry IRFL for 铁落 = "iron forging scale" and IRES for  铁精 = "iron essence" (apparently iron oxide dust that collects in the chimney of a iron smelting funrnace).  In my ZHB file IRFL and IRON are next to each other, but I cannot tell whether they are one entry or two.  Again, I asked Gemini to clarify the situation but he could not give me a trustworthy answer.  But if they were transcribed together as one parag, that parag should be about 70 EVA letters long, hence not f112v.35

But that should give an idea of why I have not created and posted a full hanzi-Voynichese dictionary yet.  Is pched.she the Voynichese for 鼺鼠, or for 铁? Same for the other hanzi in those two recipes.  Or maybe f112v.35 is neither of those two entries, so pched.she could be anything else...

For other recipes things may be worse.  For a recipe of middling size, the expected-size criterion is passed by dozens of parags.  (For a 31-hanzi recipe, my program currently considers parags with 137 to 180 EVA letters -- and there are 98 of them).  For those recipes, my program often get several "best matches" that are all more or less plausible.  The skeptics would then argue that, with so many candidate parags, it is not remarkable to find one that matches the cribs with the allowed leeway.

Moreover, excluding the Rooster, three recipes have 3 occurrences of 主, 23 have two occurrences, and the other 340 have only one.   If someone is not convinced by the matching of the eight cribs of Rooster, it would be pointless to show them one of those.

But here is one "middling size" entry (BLAR) whose corresponding parag seems to be well-defined:

(A)  |蜂子：　　　　　|fēng zǐ:                  |Bee larva:                            
(A1) |　［味］甘，平。 |··[wèi] gān, píng         |··[Nature] sweet, neutral.            
(A3) |　［主］　　　　  |··[zhǔ]                   |··[Mainly for]                       
(A31)|　　风头，　　　 |····fēng tóu,             |····wind-caused headache,             
(A32)|　　除蛊毒，　　 |····chú gǔ dú,            |····eliminating magic toxins,         
(A33)|　　补虚羸，　　 |····bǔ xū léi,            |····repairing depletion and emaciation,   
(A34)|　　伤中。　　　 |····shāng zhōng.          |····injury to digestive system.       
(A4) |　［久服］　　　  |··[jiǔ fú]                |··[Prolonged consumption]            
(A41)|　　令人光泽，　 |····lìng rén guāng zé,    |····makes one's skin lustrous,        
(A42)|　　好颜色，　　 |····hǎo yán sè,           |····improves complexion,             
(A43)|　　不老。　　　 |····bù lǎo.               |····prevents aging.                   
(B)  |大黄蜂子：　　　|dà huáng fēng zǐ:         |Giant hornet larva:                   
(B3) |　［主］　　　　  |··[zhǔ]                   |··[Mainly for]                       
(B31)|　　心腹胀满痛，|····xīn fù zhàng mǎn tòng,|····distention, fullness, and pain in torso,
(B32)|　　轻身，　　　 |····qīng shēn,            |····lightens the body,                
(B33)|　　益气。　　　 |····yì qì.                |····boosts the vital energy           
©  |土蜂子：　　　　|tǔ fēng zǐ:               |Ground-nesting wasp larva:            
(C3) |　［主］　　　　  |··[zhǔ]                   |··[Mainly for]                       
(C31)|　　痈，　　　　 |····yōng,                 |····carbuncles,                       
(C32)|　　肿。　　　　 |····zhǒng.                |····abscesses.                        
(A5) |　［一名］蜚零。 |··[yī míng] fēi líng.     |··[Another name] 蜚零.                
(A6) |　［生］山谷。　  |··[shēng] shān gǔ.        |··[Provenance] mountain valleys.

Based on its size (44 hanzi after excluding the omitted fields), the matching parag should have about 220 EVA letters.  This entry has three 主 keywords. Using only those three cribs, my program consders the 43 parags with size in 199 to 251, and finds only one with three daiin-like strings (in fact, three daiin) at about the right places: f114r.4 (236 letters)

      <f114r.4>  badness 0.355  236 letters
                   13(+2)    0.039 fdeechdyopche
      daiin 0.000 131(-2)    0.016 ypchedyodalychedyqopcheokaiinshedypodairochedalloiinchedyqokaiinchdydaiindchdoseedolchdolkchedychokaiinchdyqoeedyokchedydaiiinchedy
      daiin 0.000  62(-2)    0.033 ykeedyokeeedychedolchdaiinykardarycheoldchedydkchsaiinchdedyqo
      daiin 0.000  15(+4)    0.267 okchedaiinchain

Here is the plot:
   
I should discuss this match more (including show what happens when I ask for more cribs) but that has to wait for the next post.

All the best, --stolfi

[JoJo_Jost]

@ Stolfi You gave yourself the title “Annoying Skeptic”—maybe it’s a little bit of karma that so many people are now such “annoying skeptics” of your theory.

But this is all about “convincing” evidence.

And of course, your examples leave plenty of room for interpretation—I don’t need to tell you that; you said it yourself. I’d really be interested to know what you, as an annoying skeptic, would say about the theory if someone else presented it to you. It’s a shame we can’t find out. 

1. For example, in the last example, there are 6 “daiin” and one “daiiin”—that is, 7 other matches—compared to only 3 matches. So what are the other “daiin”? Homophones?

2. The text also contains “qokaiin”—which, according to your interpretation = “and” “kaiin” - “kaiin” was also one of the options you listed. So there are actually 8 possibilities with 3 matches—so that’s not a good example!

3. “qokaiin” also appears right next to “daiin” in the top variant. In that case, there would even be 2 matches side by side, one has no correspondence.

Unfortunately, this reminds me too much of the mental overfitting seen in other solvers: errors are given low weight, while matches are heavily overweighted. These are exactly the things we so often criticize here—and that you yourself have criticized often enough. (But I actually think that we all suffer from this problem to a greater or lesser extent.)

I hope you understand my objection; it’s not an attack, but rather the things I’m wondering about.

But I'll just keep waiting for more evidence, because I can imagine what a lot of work that must be: an imprecise source text, not speaking the language, and then all the quirks of written Chinese compared to the spoken language—plus the unmanageable VMS with possibly (!) many omissions, and then the Chinese text, where a few paragraphs might still be missing —getting all of that to work together is certainly extremely time-consuming and an enormous amount of work—and if it’s in Chinese, then that’s a huge challenge. We should acknowledge that, too. Not an easy task!

All the best

Jost

[eggyk]

Here is the plot:

I should discuss this match more (including show what happens when I ask for more cribs) but that has to wait for the next post.

All the best, --stolfi

So, why wouldn't something like this be as valid? 2 strong matches linking 主 to "okche" (one is opche, a clear spelling error where the author mistakenly used p instead of k), one faint/weak match (likely due to the uneven spacing of characters on the top row). But more importantly, no leftover instances of "okche" that match to nothing. 

   

Or a different angle: "Lorem ipsum daiin", where DAIIN is added to lorem ipsum text after every 15 words (after the first). It looks promising. 

   

Allowing spelling errors is not great, but it can at least be justified to a certain extent. However, I do not know what your justification is for all of these non matches. In your plot, only 50% of daiins match to 主. The fact that there are 2 instances where daiin blatantly doesn't match to anything, and one instance of a double daiin where none is present in the SBJ is surely strong evidence that there isn't a match. 

In order for this match to be considered a match, in my view 2 criteria must be satisfied: 

1) instances of daiin convincingly match to 主
AND
2) instances of daiin do not consistently fail to match to anything

The plot potentially satisfies the first. It fails the second. 

You can work out the expected match rate based on random chance, I suppose, but any dictation/copying scenario is not subject to random chance. You have to take into account the probability of the author incorrectly writing daiin (a common word) 3 times, when he is supposed to have a strong enough grasp of the language to be able to consistently understand and differentiate tones. That is difficult to believe.

It is also worth mentioning that using random chance is also inappropriate in the sense that words in languages follow structural patterns, which certainly applies to chinese and may apply to voynichese. It's entirely plausible that the VMS is the same type of written content as the SBJ (recipe books for instance), but that recipes themselves have broad shared patterns across many/all languages. I'm reminded of the herbal passage transcribed in You are not allowed to view links. Register or Login to view. that also writes its content in a similar way: 

1. Herb name: Pentadatillus; others call it Palma Christi, others Castos Oro guardian of the garden, others
2. Priapiscus, and others Girasol; the Greeks call it Tebeet, and the Gauls call it Perga Major. Herb grows
3. in gardens, fields, and mountains, similar to hellebore. Herb has broad leaves,
4. split like a wild fig tree, and it bears a certain round pod like wild burdock.
5. In this small pod, something is enclosed that resembles a certain worm,
6. common name: cetha. This herb is called Girasol because its leaves turn
7. according to the turning of the sun; just as the sun beams brightly, so these leaves turn themselves with the sun. The flower
8. does the same. The crushed seed of this plant's black pods, when drunk with wine,
9. warm, relieves [indecipherable]. [Indecipherable] is spherical, and [indecipherable]
10. madness, and when eaten it is beneficial against bloating/flatulence, which it solves. Likewise, its seed gives
11. the virtue of walking invisibly, according to what is found in Dioscorides, who
12. says expressly: 'Take nine seeds of this herb during the harvest, and plant them
13. inside the head of a black [indecipherable], and place it in the earth.'

Herb name(s)  --->   where it grows  --->  how it looks/acts  --->   how its used to treat   ---> how to use the seeds

I would speculate that a match to this type of content would also provide spurious "above chance" matches to other similar content, based on structure and topic alone.

[Jorge_Stolfi]

JoJo, forgive me for repeating the advice I gave on another thread:  character-level statistical analyses (character and digraph frequencies, entropies, correlations, etc.) are very likely to be a waste of time.  

One of my character-level statistical analyses led you to remove a wrong criterion for identifying paragraph breaks in the Stars section.  Not a total waste of time, surely?  And when you draw conclusions about ... p is k on the top row, is that really based on a word-level analysis and not a character-level one?  It's when one goes beyond the individual glyph and look at clusters, if not full word types, that problems with glyph mapping tend to emerge.  And I say that as someone who has feverishly mapped virtually every glyph to another at some point.

Apologies for coming off as too negative.  (But please note that I said "very likely",  not "guaranteed".)  

Character-level statistics do indeed have utility in the right contexts.  I have used them myself when developing my word structure model, which includes the proposal that p/f are "capital" forms of ke/te (or possibly more; and still not sure which is which).  And, of course, when they do yield useful insights, it would be silly to dismiss these just because they are "character-based".

But I still stand by my advice.  It is not a light comment: it comes from 30 years of watching people spend inordinate amounts of time doing character statistics that did not yield any useful conclusions.  

Or, worse, that led to wrong conclusions.  Like the attempts to identify the language (or prove that it is not a natural language) by comparing character entropies of Voynichese to those of sample texts.

Many people seem unaware that character statistics are not diagnostic properties of language, topic, or author.  They do depend on those attributes, in the sense that they will usually give different values on plaintexts that differ in those attributes.  However, the dependency is quite complicated and indirect, and character statistics are obliterated by even small changes in the spelling or encryption. 

Character statistics can lead to cracking of a simple substitution cipher applied to a plaintext with mostly "official" orthography.  But even slight complexifications -- like context- of position-dependent nulls and one-to-many letter encodings -- will render that tool useless.  (IIUC, this was the case of the Copiale cipher.)  

As I noted before, if the words of a text have very unequal frequencies (as implied by Zipf's law), the frequency of a character or digraph in a text is largely determined by its occurrence or absence in the most common words.  The assertion that "th" is more common at the start of an English word than expected by chance may be true, but it is not very informative.  It is even somewhat misleading, because it seems to suggest that "th" is sort of a fungus that grows indiscriminately on the edges of words on islands with rainy climate and warm beer.  The statement some of the most common English words begin with "th", while technically weaker, should give a better idea of what is going on.

Many others have noted that character and digraph frequencies around line breaks are different from those in the middle parts of lines. @JoJo recently pointed out anomalous statistics about initial Sh, Ch, and q. @DG97EEB posted on patterns of the initial letters of successive lines. These anomalies are real, but they are not  properties of those characters.  They are properties of certain words and sets of words, which mostly happen to use those characters.

So, if you happen to notice anomalies in character or n-gram statistics, you should not stop there, or try to deduce character-based "rules" from that data.  You should check which words are responsible for the anomaly, and try to reformulate the observation in terms of words.  

It is likely that you will find that (say) it is not that q is more common than expected at line start, but rather that the words qokeedy and okeedy and keedy are more enhanced there.  Or that words with the qo prefix tend to be longer than those without it -- and line breaking naturally occurs more often before longer words.

about final m replacing final n

Character statistics indeed suggested that m could be an abbreviation.  But the theory that it is an abbreviation for iin (not n) came mainly from observing that the word dam, specifically, was very common at the end of lines -- suggesting that it was an abbreviation for the word daiin, which is common outside of that context.

Same  for my current belief that final ir is mostly  a scribal error for iin.  I first suspected it from the frequencies of the  final "elements" iin, iiin, ir, and iir in my word structure model.  But it became more certain when I saw the words dair and kaiin (but not taiin!) show up often in the SBJ-SPS matchings where daiin was expected.  Then I realized that, in moderately sloppy handwriting, the pair in could be easily be confused with r, and d could be confused with k (but not with t!) 

All the best, --stolfi

[Jorge_Stolfi]

So, why wouldn't something like this be as valid? 2 strong matches linking 主 to "okche" (one is opche, a clear spelling error where the author mistakenly used p instead of k), one faint/weak match (likely due to the uneven spacing of characters on the top row). But more importantly, no leftover instances of "okche" that match to nothing.

That is a very bad matching. 

First because, as you note, the spacing of the proposed Voynchese words (in EVA letters) is too different from the spacing of the 主 (in hanzi scaled by ~5.338).  In a good crib matching the "gap errors" (discrepancies between actual and predicted gap sizes) should be mostly less than ~5 EVA = ~1 hanzi.  

The hanzi gaps before, between, and after the 主 in the recipe are 2, 25, 12, and 2 hanzi. The predicted gap sizes in the SPS are then 11, 133, 64, and 11 EVA letters.  In the matching to f114r.4 that I posted, with the three  主 mapped to those three daiin, the observed gap sizes are 13, 131, 62, and 15 EVA letters.  

So the gap errors for my matching are +2, -2, -2, and +5 EVA letters, which correspond to +0.4, -0.4, -0.4, and  +1.0 hanzi.   Your proposed matching would have gap errors -2, -25, +21, and -1 EVA letters.  Note that, statistically, a measure of the badness of a matching should be something like the sum of the squares of the gap errors. 

Second, the cribs and their allowed "spelling variants" must be the same for all recipes and paragraphs.  The mapping  主 = ~daiin was determined from the Rooster recipe, and there is no viable alternative there.  So the only freedom in this case is the choice of which three daiin-like strings should be paired with the three 主.  

Or a different angle: "Lorem ipsum daiin", where DAIIN is added to lorem ipsum text after every 15 words (after the first). It looks promising.

This is a slightly more pertinent objection, but it still has problems.

That specific "lorem" parag is obviously a bad match because it is way too long; but that could be fixed if you used a an appropriate shorter filler, to match the total size of f114r.4 (236 EVA letters).  

Also note that by conveniently fixing the size of the first and last gaps, you left only two gaps with "random" size; and considering  that the total size is fixed, that leaves only variable to be matched by chance.

A better check would be to generate 43 parags with 236 EVA letters, with six daiin scattered among filler text, in random positions (including the initial and final gaps); and then check how many would allow the three 主 to be paired with three of those six daiin with total square gap size error less than or equal to (+2)^2 + (-2)^2 + (-2)^2 + (+5)^2 = 37.

However, I do not know what your justification is for all of these non matches. In your plot, only 50% of daiins match to 主. The fact that there are 2 instances where daiin blatantly doesn't match to anything, and one instance of a double daiin where none is present in the SBJ is surely strong evidence that there isn't a match.

Homophones are possible explanation, but they should be relatively rare.

A more likely explanation for some of those unpaired daiin is that they are parts of bigger words that are not related to 主.  Like HIS is a word on its own but also part of THIS and HISTORY and CHISEL.

Another, perhaps even more likely explanation, is that the encoding of  sounds to Voynichese is many-to-one.  For example, here is the syllable-by-syllable reading of that recipe in modern Mandarin:
  fēng zǐ wèi gān píng zhǔ fēng tóu chú gǔ dú bǔ xū
  léi shāng zhōng jiǔ fú lìng rén guāng zé hǎo yán
  sè bù lǎo dà huáng fēng zǐ zhǔ xīn fù zhàng mǎn
  tòng qīng shēn yì qì tǔ fēng zǐ zhǔ yōng zhǒng
The syllable fēng occurs four times (three as the reading of = "bee,wasp,hornet" in the three titles, and once as the reading of 风 = wind).  The syllables zhǔ (as the Mandarin reading of 主) and zǐ (as the reading of 子 = "child,egg,larva,etc.") occur three times each.  All other syllables occur only once each.But now consider what happens if we omit the tone marks: 
  feng zi zhu feng tou chu gu du bu xu lei shang
  zhong jiu fu ling ren guang ze hao yan se bu lao
  da huang feng zi zhu xin fu zhang man tong qing
  shen yi qi tu feng zi zhu yong zhong
Now feng, zhu, and zi still occur 4, 3, and  3 times each; but bu, fu, and zhong now occur twice each. Each of them corresponding to two very different hanzi.

Now check the pronunciation of pinyin "x", "q", "c", "zh", "z", "sh", and see how many of them you can tell apart...

Needless to say, I don't know what is the language that the SBJ was dicitated in, but surely it was not modern Mandarin. Thus, it is quite possible that the Author was unable to distinguish the syllables for 主 from the syllables for three other hanzi, and wrote all six syllables as daiin.

In order for this match to be considered a match, in my view 2 criteria must be satisfied: 1) instances of daiin convincingly match to 主, AND 2) instances of daiin do not consistently fail to match to anything

Then I am afraid that you will never see the VMS deciphered...

You have to take into account the probability of the author incorrectly writing daiin (a common word) 3 times, when he is supposed to have a strong enough grasp of the language to be able to consistently understand and differentiate tones.

We don't know whether he was able to distinguish all the tones and other phonemes of the spoken language.  He probably was, because he must have been aware that tones were important, and he would not have undertaken this project if he was not confident that he could distinguish them.  

The SBJ uses ~1000 distinct hanzi, and in my transcription of the SPS,ignoring commas, I count 927 distinct words that occur at least twice. (I count also ~2300 hapaxes, but many are probably "errors" of various sorts, including the result of omitted or bogus spaces).  The count would be a bit higher if the four missing folios were not missing; on the other hand, many of those 927 may be variant spellings (like dain or kaiin for daiin). 

Thus these numbers number may not prove anything, but they are at least compatible with the mapping from hanzi to Voynichese words being mostly one-to-one.

However, the Author may have omitted tone information where he could tell the meaning even without it, just from the context.  Like when people in English write "read" for two different words with different sounds: usually they can tell from context which spoken word those letters stand for.

It's entirely plausible that the VMS is the same type of written content as the SBJ (recipe books for instance), but that recipes themselves have broad shared patterns across many/all languages.

 
That is true; but, as shown, in my comparisons I require that the cribs be at similar distances, not just in the same order.

As for "other languages", it has long been known that the statistics and word structure of Voynichese are incompatible with "European" languages, including Arabic and Hebrew.  The word structure could be explained by a codebook cipher; but the "sentence" structure seems harder to explain under the "European language" assumption.

All the best, --stolfi

[rikforto]

I will get to the outstanding answer to my last post when I get to it, but this is simpler to address:

There is a complication, though: according to Gemini there is an even shorter recipe in the SBJ, with only 6 hanzi:
  铁：［主］坚肌，耐痛。

Do you have a specific edition that has this as a full entry? Not a work, but an edition?

[ololololo]

Only one recipe is presented as proof, and everything that @eggyk has said here is correct, because you are relying on a single recipe, and it proves nothing.

The more direct evidence (and clear correspondences for daiin) you have, the more convincing your theory will look. It’s simple...

[Jorge_Stolfi]

As promised, here is the outcome of asking my program to find the matching SPS parag to the SBJ recipe BLAR 蜂子 = "bee larva", with more cribs.

There are several hanzi in that recipe that I am testing as possible additional cribs, besides 主 ↔ daiin = "mainly for"; but the two that I am reasonably confident about are 
气 = "vital energy", with canonical translation Chedy
令 = "makes", with canonical translation Cheody, Cheydy, or the like
When asked to look for the cribs 主,令,主,气,主, my program still says that f114r.4 is the best match. Here is the matching plot:
   
In this pairing, all hanzi cribs are paired with their "canoncal" translations, except 令 that has only a Chdy instead of a Cheody near the expected place. 

The gaps in the recipe before, between,and after those cribs are 2,12, 12, 8, 3, and 2 hanzi.  Assuming the average of 5.338 EVA lertters per hanzi, the predicted gaps in the matching SPS parag would be 11, 64, 64, 43, 16, and 11 EVA letters.  The actual gaps in f114r.4 between the assigned translations are 13, 64, 63, 40, 17, and 15 EVA letters.  The errors relative to the predicted values are +2, 0, -1, -2, 0, and +4 EVA letters.

All te best, --stolfi