20-11-2026, 09:45 AM
What is faceted notation?
Let's say that you are a keen gardener. The weather probably plays a big part in the decisions that you make about what you plant and when.
Frequently, you find that the general weather forecasts are inaccurate, so you decide to track the weather for yourself.
At first, you decide to keep a diary of events, but after a few weeks you realise that the diary is growing at an exponential rate and it's difficult to get the gist of an entry at a glance. So instead, you decide to enter your observations into a simple spreadsheet table.
The table has 7 columns. Something like this:
Month Week Humidity Temperature Wet? Windy? Summary
Month holds 12 values, Week holds 4 values, Humidity holds 3 value ranges, Temperature holds 4 value ranges, Wet and Windy 2 values each and the summary 2 values.
The month is split onto its own row rather than being repeated in each record, so a month's entries might look like this:
3
1, LOW, HOT, NO, NO, GOOD
2, MEDIUM, HOT, YES, YES, BAD
3, MEDIUM, HOT, NO, NO, GOOD
4, HIGH, MEDIUM, YES, YES, BAD
Note:
I cannot easily show the structure of the table here, so the month row only includes the month and all the other columns in that row are left empty. The remaining rows contain no month, so the first column is left empty.
Now, after a while your spreadsheet will have hundreds of entries but by scanning the numbers, you can get a feel for good and bad days - you get used to their signature.
To make the results easier to use, you decide to print them out and stick them to the inside of your greenhouse. Every time you go in there, you can take a look and be reminded about the last week or last month at a glance.
After a few months things are looking good, but you have a problem. A nosy neighbour has started to show an interest in your hard work. This wouldn't matter normally, but he too is a keen gardener and you and he are quite competitive.
Your solution - you replace the numbers and text in your chart with letters. Moreover, so he doesn't know that it is a chart, you write each line of the chart out as a single 'word', created by concatenating the column values together. A month's results might look like this:
c tandda fbneeb wbndda kcmeeb
The text above is an example of faceted notation written as running text. The notation provides a rigid and meaningful way of capturing information. The process for creating a faceted record is simple - it's dictated by the columns and their possible values. The conversion of the values to letters and their subsequent commitment as running text provides a level of obfuscation while presenting a repeating signature meaningful to the author. Scanning the words gives the gist, just like scanning the entries of a log file gives the reader a 'feel' without looking at every entry of the log in detail.
When creating a record, it may be necessary to leave a column blank because you forgot to record the result. If this happens, you have two choices. When you convert the entry into a 'word', you include a placeholder for the missing value. Alternatively, you can just leave the missing value out. If you do leave missing values out, that's not a problem if each column uses unique letters for its values - the columns are disjoint. However, if you share letters between columns, you need to introduce some rules, otherwise it becomes impossible to distinguish columns.
Introducing a 13-facet Notation
Taking the standard EVA transliteration mnemonics, let's suppose that the Voynich running text was created using a 13-facet notation like this:
FRONT P1 - qo ch y sh q
PRE-CORE - e s ee r eee i
FRONT P2 - d
FRONT P3 - a
FRONT P4 - o
FRONT P5 - l
GALLOWS - k t p f ckh cth cph cfh
CORE 1 - l r lr rl
CORE 2 - e ch ee che sh she eee chee … (27 in all)
CORE 3 - o d od s os do sd so … (13 in all)
CORE 4 - a
CORE 5 - i ii
CLOSER - y r in l m g n
Now, as you can see, the facet glyph pools are not disjoint, so we will need three simple rules to read back words 'encoded' using this notation:
1) The gallows divides the word. Everything before it belongs to the front group; everything after it is the core.
2) Where there is no gallows, nothing marks the division, so the reader asks one question instead: the leading signs are the front if core material follows them, and are themselves core if it does not.
3) FRONT P2, P3 and P4 are used only where a gallows divides the word, or where there is no gallows but there are core values.
So let's try this out on some words found in the Voynich manuscript.
QOKEEDY, which occurs 307 times in the manuscript:
FRONT P1 qo
GALLOWS qok
CORE2 qokee
CORE3 qokeed
CLOSER qokeedy
Read backwards, the word contains the gallows k so:
qo is FRONT P1
k is the GALLOWS
ee is CORE2
d is CORE3
y is the CLOSER
CHEDY, which occurs 506 times in the manuscript:
CORE2 che
CORE3 ched
CLOSER chedy
Read backwards, the word contains no gallows and no front so:
che is CORE2
d is CORE3
y is the closer
OLAIIN, which occurs 48 times in the manuscript:
FRONT P4 o
CORE1 ol
CORE4 ola
CORE5 olai
CLOSER olaiin
Read backwards, the word contains no gallows and a front so:
o is FRONT P4
l is CORE1
a is CORE4
i is CORE5
in is the closer
CHEKY, which occurs 65 times in the manuscript:
FRONT P1 ch
PRE-CORE che
GALLOWS chek
CLOSER cheky
Read backwards, the word contains a gallows and a PRE-CORE entry:
ch is FRONT P1
e is PRE-CORE
k is GALLOWS
y is the CLOSER
Now taking all the running text in the Voynich manuscript, this notation will 'decode' approximately 90% of the words. Allowing for duplication of words in the manuscript, each deduplicated word decodes to a unique record.
Of course, you could argue that any mechanism free enough could parse that number of words. But if you retain all the same words in the same order but simply shuffle their signs, that percentage drops to approximately 33% (however many times you do the shuffle) - so the approximate 57% gap in decoding is down to understanding the structure of the Voynich words and not to do with the flexibility of the notation and its rules.
The notation inverts to produce consistent unique records with values supplied by the Voynich words. The notation manages form, but not interpretation. Interpretation is down to the meaning attributed to facet values. And this is where the strength of a notation lies. Once mastered by the author, they can 'encode' any type of data within the notation structure - the Voynich three subjects being an example.
And of course, inversion is not really what a notation is about. Given the restricted glyph inventory and shapes of words, a proficient author would quickly master getting the gist of a record's content simply from the notational word shape.
Let's say that you are a keen gardener. The weather probably plays a big part in the decisions that you make about what you plant and when.
Frequently, you find that the general weather forecasts are inaccurate, so you decide to track the weather for yourself.
At first, you decide to keep a diary of events, but after a few weeks you realise that the diary is growing at an exponential rate and it's difficult to get the gist of an entry at a glance. So instead, you decide to enter your observations into a simple spreadsheet table.
The table has 7 columns. Something like this:
Month Week Humidity Temperature Wet? Windy? Summary
Month holds 12 values, Week holds 4 values, Humidity holds 3 value ranges, Temperature holds 4 value ranges, Wet and Windy 2 values each and the summary 2 values.
The month is split onto its own row rather than being repeated in each record, so a month's entries might look like this:
3
1, LOW, HOT, NO, NO, GOOD
2, MEDIUM, HOT, YES, YES, BAD
3, MEDIUM, HOT, NO, NO, GOOD
4, HIGH, MEDIUM, YES, YES, BAD
Note:
I cannot easily show the structure of the table here, so the month row only includes the month and all the other columns in that row are left empty. The remaining rows contain no month, so the first column is left empty.
Now, after a while your spreadsheet will have hundreds of entries but by scanning the numbers, you can get a feel for good and bad days - you get used to their signature.
To make the results easier to use, you decide to print them out and stick them to the inside of your greenhouse. Every time you go in there, you can take a look and be reminded about the last week or last month at a glance.
After a few months things are looking good, but you have a problem. A nosy neighbour has started to show an interest in your hard work. This wouldn't matter normally, but he too is a keen gardener and you and he are quite competitive.
Your solution - you replace the numbers and text in your chart with letters. Moreover, so he doesn't know that it is a chart, you write each line of the chart out as a single 'word', created by concatenating the column values together. A month's results might look like this:
c tandda fbneeb wbndda kcmeeb
The text above is an example of faceted notation written as running text. The notation provides a rigid and meaningful way of capturing information. The process for creating a faceted record is simple - it's dictated by the columns and their possible values. The conversion of the values to letters and their subsequent commitment as running text provides a level of obfuscation while presenting a repeating signature meaningful to the author. Scanning the words gives the gist, just like scanning the entries of a log file gives the reader a 'feel' without looking at every entry of the log in detail.
When creating a record, it may be necessary to leave a column blank because you forgot to record the result. If this happens, you have two choices. When you convert the entry into a 'word', you include a placeholder for the missing value. Alternatively, you can just leave the missing value out. If you do leave missing values out, that's not a problem if each column uses unique letters for its values - the columns are disjoint. However, if you share letters between columns, you need to introduce some rules, otherwise it becomes impossible to distinguish columns.
Introducing a 13-facet Notation
Taking the standard EVA transliteration mnemonics, let's suppose that the Voynich running text was created using a 13-facet notation like this:
FRONT P1 - qo ch y sh q
PRE-CORE - e s ee r eee i
FRONT P2 - d
FRONT P3 - a
FRONT P4 - o
FRONT P5 - l
GALLOWS - k t p f ckh cth cph cfh
CORE 1 - l r lr rl
CORE 2 - e ch ee che sh she eee chee … (27 in all)
CORE 3 - o d od s os do sd so … (13 in all)
CORE 4 - a
CORE 5 - i ii
CLOSER - y r in l m g n
Now, as you can see, the facet glyph pools are not disjoint, so we will need three simple rules to read back words 'encoded' using this notation:
1) The gallows divides the word. Everything before it belongs to the front group; everything after it is the core.
2) Where there is no gallows, nothing marks the division, so the reader asks one question instead: the leading signs are the front if core material follows them, and are themselves core if it does not.
3) FRONT P2, P3 and P4 are used only where a gallows divides the word, or where there is no gallows but there are core values.
So let's try this out on some words found in the Voynich manuscript.
QOKEEDY, which occurs 307 times in the manuscript:
FRONT P1 qo
GALLOWS qok
CORE2 qokee
CORE3 qokeed
CLOSER qokeedy
Read backwards, the word contains the gallows k so:
qo is FRONT P1
k is the GALLOWS
ee is CORE2
d is CORE3
y is the CLOSER
CHEDY, which occurs 506 times in the manuscript:
CORE2 che
CORE3 ched
CLOSER chedy
Read backwards, the word contains no gallows and no front so:
che is CORE2
d is CORE3
y is the closer
OLAIIN, which occurs 48 times in the manuscript:
FRONT P4 o
CORE1 ol
CORE4 ola
CORE5 olai
CLOSER olaiin
Read backwards, the word contains no gallows and a front so:
o is FRONT P4
l is CORE1
a is CORE4
i is CORE5
in is the closer
CHEKY, which occurs 65 times in the manuscript:
FRONT P1 ch
PRE-CORE che
GALLOWS chek
CLOSER cheky
Read backwards, the word contains a gallows and a PRE-CORE entry:
ch is FRONT P1
e is PRE-CORE
k is GALLOWS
y is the CLOSER
Now taking all the running text in the Voynich manuscript, this notation will 'decode' approximately 90% of the words. Allowing for duplication of words in the manuscript, each deduplicated word decodes to a unique record.
Of course, you could argue that any mechanism free enough could parse that number of words. But if you retain all the same words in the same order but simply shuffle their signs, that percentage drops to approximately 33% (however many times you do the shuffle) - so the approximate 57% gap in decoding is down to understanding the structure of the Voynich words and not to do with the flexibility of the notation and its rules.
The notation inverts to produce consistent unique records with values supplied by the Voynich words. The notation manages form, but not interpretation. Interpretation is down to the meaning attributed to facet values. And this is where the strength of a notation lies. Once mastered by the author, they can 'encode' any type of data within the notation structure - the Voynich three subjects being an example.
And of course, inversion is not really what a notation is about. Given the restricted glyph inventory and shapes of words, a proficient author would quickly master getting the gist of a record's content simply from the notational word shape.