04-02-2017, 10:15 PM
(This post was last modified: 04-02-2017, 11:41 PM by Torsten.
Edit Reason: some remarks added
)
Hello Emma,
With "Therefore you can't use this text to determine if some of this texts are language like or not." I mean that there is a difference if you search a feature typical for languages in the VMS or if you search for a feature typical for the VMS in languages. My test shows that the compared texts in Latin, English, German and Arabic are not like the text of the VMS. This way I have compared a feature typical for the VMS. In my eyes this is more then nothing.
At least it is more then Montemurro and Zanette did. They used features typical for language: "We first apply methods from information theory that identify content-bearing words without any prior knowledge of the underlying language of the text under analysis. Then, we consider putative semantic relationships between the most informative words by analysing their patterns of co-occurrence along the text" [You are not allowed to view links. Register or Login to view.]. This way they started with the assumption that the VMS contains language. While analyzing the "100 most frequently used words within the VMS" Montemurro and Zanette also come to the conclusion that "Words that are related by their semantic contents tend to co-occur along the text." [You are not allowed to view links. Register or Login to view.]. But even for there 100 word types they found something like a network for the VMS:
![[Image: journal.pone.0066344.g002]](http://journals.plos.org/plosone/article/figure/image?size=inline&id=info:doi/10.1371/journal.pone.0066344.g002)
Also the shift from Currier A to Currier B can be found in there paper:
![[Image: journal.pone.0066344.g004]](http://journals.plos.org/plosone/article/figure/image?size=inline&id=info:doi/10.1371/journal.pone.0066344.g004)
Since they did not distinguish between herbal pages in Currier A and B there is an additional link from the astronomical section to the herbal section.
Since the goal of Montemurro and Zanette was to demonstrate that the VMS is language like they interpreted there result as confirmation of there starting hypothesis. They didn't check if the co-occurence pattern also holds for all words in the VMS and they didn't check if it is common for a language that more then the most frequently used words are related by their semantic contents: "The strongest links between the different sections as determined by the co-occurrence of the most informative words." [Montemurro 2013]. I would expect that the most frequently used words in a language are words like "and" or "the" and that this words are equally distributed within a text. In my eyes it is there starting hypotheses that prevents them seeing that there data demonstrate that the VMS is different from language.
Quote:Well, great, so your assertion that the pattern is not language like is based on nothing?
With "Therefore you can't use this text to determine if some of this texts are language like or not." I mean that there is a difference if you search a feature typical for languages in the VMS or if you search for a feature typical for the VMS in languages. My test shows that the compared texts in Latin, English, German and Arabic are not like the text of the VMS. This way I have compared a feature typical for the VMS. In my eyes this is more then nothing.
At least it is more then Montemurro and Zanette did. They used features typical for language: "We first apply methods from information theory that identify content-bearing words without any prior knowledge of the underlying language of the text under analysis. Then, we consider putative semantic relationships between the most informative words by analysing their patterns of co-occurrence along the text" [You are not allowed to view links. Register or Login to view.]. This way they started with the assumption that the VMS contains language. While analyzing the "100 most frequently used words within the VMS" Montemurro and Zanette also come to the conclusion that "Words that are related by their semantic contents tend to co-occur along the text." [You are not allowed to view links. Register or Login to view.]. But even for there 100 word types they found something like a network for the VMS:
Also the shift from Currier A to Currier B can be found in there paper:
Since they did not distinguish between herbal pages in Currier A and B there is an additional link from the astronomical section to the herbal section.
Since the goal of Montemurro and Zanette was to demonstrate that the VMS is language like they interpreted there result as confirmation of there starting hypothesis. They didn't check if the co-occurence pattern also holds for all words in the VMS and they didn't check if it is common for a language that more then the most frequently used words are related by their semantic contents: "The strongest links between the different sections as determined by the co-occurrence of the most informative words." [Montemurro 2013]. I would expect that the most frequently used words in a language are words like "and" or "the" and that this words are equally distributed within a text. In my eyes it is there starting hypotheses that prevents them seeing that there data demonstrate that the VMS is different from language.