06-09-2020, 07:20 PM
(This post was last modified: 06-09-2020, 11:29 PM by Torsten.
Edit Reason: typo
)
For discussing a hypothesis it would be necessary to look into the details of my argumentation in order to point to concrete mistakes. Only this way it would be possible to correct the mistakes and too improve this way the knowledge about the Voynich manuscript. But René Zandbergen only wants to deny what is unthinkable in his eyes. To make it less obvious that he didn't go into the details of our argumentation René falsely writes that 'the details of the algorithm have not been shown'. Not only was the algorithm published in our paper but also the source code of an implementation of the algorithm is available at You are not allowed to view links. Register or Login to view..
Then René Zandbergen argues that he thinks that a text generated by the self-citation method must look different. For demonstrating his claim he comes up with his own version of the self-citation method. He assumes that someone copying words would solely add 'small modifications'. This assumption does not refer to a statement I had made in one of my papers! I also have pointed out multiple You are not allowed to view links. Register or Login to view. that such an assumption is simply wrong. I can only encourage everybody interested in our argumentation to read our paper(s) and to check yourself what we say and why (see Timm & Schinner 2020, p. 9f and Timm 2015, p. 16).
In René Zandbergens truncated version a word can only be generated by adding 'small modifications of earlier words'. In our paper we present three concrete modification rules: 'Replacing one or more glyphs by similar ones', about 'Adding or removing a prefix' and about 'Combining two source words to a create new word' (Timm & Schinner 2020, p. 9f). There is nothing said about 'small modifications' and there is also nothing said about small edit distances. Renè Zandbergen already wrote in an earlier post that it his idea that only 'small changes' could be applied (You are not allowed to view links. Register or Login to view.). Beside the fact that René Zandbergen is referring to his own idea and this way to his own version of the self-citation method he is also unaware of the possibility to 'combine two source words to create a new word' and to split a word generated this way into its parts (see Timm & Schinner 2020, p. 9f and also Timm 2015, p. 16). This way his results suggests that it would be hard to come up with <daiin> for suggested source words like <chodaiin>, <fchodaiin>, or <paiindaiin>. But since the sequence 'daiin' is already part of these words it would be only necessary to copy a subgroup <daiin>.
René further assumes that someone who was creative enough to invent his own text generation method and has also invented his own script would lack the creativity to come up with more than just 'small modifications'. In my eyes this idea is wrong. Anyway the level of repetition in a text generated with René Zandbergens version would be eye catching. In order to make the level of repetition less obvious it would be necessary to add further modifications. So even someone starting with René Zandbergens idea to apply only 'small modifications' would adjust the number of modifications until the generated text is sufficient in his eyes.
Last but not least it is also impossible to copy a word from a folio if this folio is empty. Each time the scribe was starting to write on a new empty folio he had no other option than to copy at least some initial words from previous folio (see Timm & Schinner 2020, p. 11 and You are not allowed to view links. Register or Login to view., p. 16). René Zandbergen doesn't take this fact into account. His own statistics show that if <daiin> is within the first 10 word tokens only in rare instances a similar word token can be found. René Zandbergen lists 42 folios where <daiin> is one of the first 10 word tokens. The average edit distance computed on EVA for this 42 folios is 3.9 whereas the average for all other 163 folios is 1.7. Therefore these 42 folios where <daiin> is one of the first 10 word tokens shouldn't be counted. Instead of 81 % now 92 % of the instances of <daiin> are only three or less modifications away from a previous word. And instead of 74 % now 84 % of <daiin> are only one or two modifications away from a previous word. (93 folios or 57 % with 1 modification, 42 folios or 26 % with 2 modifications, 15 folios or 9 % with 3 modifications, 11 folios or 7 % with 4 modifications and 2 folios or 1 % with 5 modifications).
The statistics for <daiin> presented by René Zandbergen doesn't contradict anything we write in our paper. On the contrary also Renés calculations demonstrate that <daiin> occurs together with similar tokens on the same folios. We describe this observation as context-dependent self-similarity (Timm & Schinner 2020, p. 3). There are ten folios (f1r, f7r, f10v, f14r, f32r, f32v, f35v, f37v, f38v, and f45r) where the words <daiin> and <dain> are two out of three among the most frequent tokens. But there are also folios where <daiin> is used together with <aiin> (f41v, f46r, f55v, f89v2, v105v and f114r) or <saiin> (f2r, f16r, and f90r2). There is also a systematic difference between the two Currier languages: in Currier A the words <daiin> and <dain> are used frequently together on the same folios whereas in Currier B <daiin> is more frequently used together with <aiin> (Timm & Schinner 2020, p. 3f). It is noteworthy that René Zandbergen refuses to discuss the observations we present in our paper for the word <daiin> or to refer to any other observation we describe in one of our papers (see You are not allowed to view links. Register or Login to view., You are not allowed to view links. Register or Login to view., 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.).
In our paper we describe the Voynich manuscript as context-dependent and self-similar (see Timm & Schinner 2020, p. 3ff). This means that it is possible to describe local variations of repetition. Most times repetition and patterns of local variations are seen as distinct entities of the text. But it is possible to describe glyphs typical for a certain position within a word, line, or within a paragraph as repetitive elements. In the same way it is possible to describe glyph sequences typical for Currier A as well as glyph sequences typical for Currier B (see Timm & Schinner 2020, p. 6). Word tokens does not only occur multiple times they also occur together similar word tokens. Sometimes this is more than obvious like in 'kol chol chol kor chol sho chol' on folio f1r. The general tendency for the Voynich manuscript is that words occur together with similar ones: "when we look at the three most frequent words on each page, for more than half of the pages two of three will differ in only one detail" (Timm & Schinner 2020, p. 3).
Local variations and repetition are not only connected to each other they depend on each other. Words sharing the same glyph sequences does not only prefer a certain position within a line, they also have something in common with word tokens sharing the same line, the same paragraph, the same folio, and on a larger scale with word tokens in Currier A and B. With other words the observed patterns are part of a larger system. Typical for this system is that patterns observed on a smaller scale does also repeat on a wider scale. There are glyphs typical for the start/end of a word and there are also glyphs typical for the start/end of a line. There are also glyphs typical for the first line of a paragraph. In the same way it is possible to describe glyph sequences typical for a certain paragraph, folio, quire as well as for Currier A and B. Based on this observations we argue "These results not only cement the hypothesis that the VMS text is an artificial structure rather than natural language, but also provide constraints for the generating algorithm" (Timm & Schinner 2020, p. 8).
Then René Zandbergen argues that he thinks that a text generated by the self-citation method must look different. For demonstrating his claim he comes up with his own version of the self-citation method. He assumes that someone copying words would solely add 'small modifications'. This assumption does not refer to a statement I had made in one of my papers! I also have pointed out multiple You are not allowed to view links. Register or Login to view. that such an assumption is simply wrong. I can only encourage everybody interested in our argumentation to read our paper(s) and to check yourself what we say and why (see Timm & Schinner 2020, p. 9f and Timm 2015, p. 16).
In René Zandbergens truncated version a word can only be generated by adding 'small modifications of earlier words'. In our paper we present three concrete modification rules: 'Replacing one or more glyphs by similar ones', about 'Adding or removing a prefix' and about 'Combining two source words to a create new word' (Timm & Schinner 2020, p. 9f). There is nothing said about 'small modifications' and there is also nothing said about small edit distances. Renè Zandbergen already wrote in an earlier post that it his idea that only 'small changes' could be applied (You are not allowed to view links. Register or Login to view.). Beside the fact that René Zandbergen is referring to his own idea and this way to his own version of the self-citation method he is also unaware of the possibility to 'combine two source words to create a new word' and to split a word generated this way into its parts (see Timm & Schinner 2020, p. 9f and also Timm 2015, p. 16). This way his results suggests that it would be hard to come up with <daiin> for suggested source words like <chodaiin>, <fchodaiin>, or <paiindaiin>. But since the sequence 'daiin' is already part of these words it would be only necessary to copy a subgroup <daiin>.
René further assumes that someone who was creative enough to invent his own text generation method and has also invented his own script would lack the creativity to come up with more than just 'small modifications'. In my eyes this idea is wrong. Anyway the level of repetition in a text generated with René Zandbergens version would be eye catching. In order to make the level of repetition less obvious it would be necessary to add further modifications. So even someone starting with René Zandbergens idea to apply only 'small modifications' would adjust the number of modifications until the generated text is sufficient in his eyes.
Last but not least it is also impossible to copy a word from a folio if this folio is empty. Each time the scribe was starting to write on a new empty folio he had no other option than to copy at least some initial words from previous folio (see Timm & Schinner 2020, p. 11 and You are not allowed to view links. Register or Login to view., p. 16). René Zandbergen doesn't take this fact into account. His own statistics show that if <daiin> is within the first 10 word tokens only in rare instances a similar word token can be found. René Zandbergen lists 42 folios where <daiin> is one of the first 10 word tokens. The average edit distance computed on EVA for this 42 folios is 3.9 whereas the average for all other 163 folios is 1.7. Therefore these 42 folios where <daiin> is one of the first 10 word tokens shouldn't be counted. Instead of 81 % now 92 % of the instances of <daiin> are only three or less modifications away from a previous word. And instead of 74 % now 84 % of <daiin> are only one or two modifications away from a previous word. (93 folios or 57 % with 1 modification, 42 folios or 26 % with 2 modifications, 15 folios or 9 % with 3 modifications, 11 folios or 7 % with 4 modifications and 2 folios or 1 % with 5 modifications).
The statistics for <daiin> presented by René Zandbergen doesn't contradict anything we write in our paper. On the contrary also Renés calculations demonstrate that <daiin> occurs together with similar tokens on the same folios. We describe this observation as context-dependent self-similarity (Timm & Schinner 2020, p. 3). There are ten folios (f1r, f7r, f10v, f14r, f32r, f32v, f35v, f37v, f38v, and f45r) where the words <daiin> and <dain> are two out of three among the most frequent tokens. But there are also folios where <daiin> is used together with <aiin> (f41v, f46r, f55v, f89v2, v105v and f114r) or <saiin> (f2r, f16r, and f90r2). There is also a systematic difference between the two Currier languages: in Currier A the words <daiin> and <dain> are used frequently together on the same folios whereas in Currier B <daiin> is more frequently used together with <aiin> (Timm & Schinner 2020, p. 3f). It is noteworthy that René Zandbergen refuses to discuss the observations we present in our paper for the word <daiin> or to refer to any other observation we describe in one of our papers (see You are not allowed to view links. Register or Login to view., You are not allowed to view links. Register or Login to view., 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.).
In our paper we describe the Voynich manuscript as context-dependent and self-similar (see Timm & Schinner 2020, p. 3ff). This means that it is possible to describe local variations of repetition. Most times repetition and patterns of local variations are seen as distinct entities of the text. But it is possible to describe glyphs typical for a certain position within a word, line, or within a paragraph as repetitive elements. In the same way it is possible to describe glyph sequences typical for Currier A as well as glyph sequences typical for Currier B (see Timm & Schinner 2020, p. 6). Word tokens does not only occur multiple times they also occur together similar word tokens. Sometimes this is more than obvious like in 'kol chol chol kor chol sho chol' on folio f1r. The general tendency for the Voynich manuscript is that words occur together with similar ones: "when we look at the three most frequent words on each page, for more than half of the pages two of three will differ in only one detail" (Timm & Schinner 2020, p. 3).
Local variations and repetition are not only connected to each other they depend on each other. Words sharing the same glyph sequences does not only prefer a certain position within a line, they also have something in common with word tokens sharing the same line, the same paragraph, the same folio, and on a larger scale with word tokens in Currier A and B. With other words the observed patterns are part of a larger system. Typical for this system is that patterns observed on a smaller scale does also repeat on a wider scale. There are glyphs typical for the start/end of a word and there are also glyphs typical for the start/end of a line. There are also glyphs typical for the first line of a paragraph. In the same way it is possible to describe glyph sequences typical for a certain paragraph, folio, quire as well as for Currier A and B. Based on this observations we argue "These results not only cement the hypothesis that the VMS text is an artificial structure rather than natural language, but also provide constraints for the generating algorithm" (Timm & Schinner 2020, p. 8).