History Online - Chronology
[an error occurred while processing this directive]Ancient and Medieval Chronology
New methods for the discovery of dependences between texts
A.T.Fomenko - G.V.Nosovsky
The problem of recognizing dependences (and dependent texts) arises in many branches of applied statistics, linguistics, physics, genetics, and so forth. For example, as applied to source research, the discovery of dependent texts with a common primary source or original (which may not have survived) is of considerable interest. On the other hand, it is useful to have an idea which texts may be called independent or are based on substantially different primary sources and archival data. Meanwhile, the concept of text itself can be treated extremely differently. We can consider a sequence of symbols, signals, codes (of various kinds; for example, genetic codes in DNA chains) as a text, where the general problem in the search for 'dependent texts' consists of find-omg 'similar' portions in a given long signal sequence, i.e., textual fragments 'duplicating' each other.
Today, there are many methods for finding dependences of this sort. We suggest certain new empirico-statistical procedures which can prove useful both in analyzing narrative texts (such as annals and chronicles) and in studying biological codes to find so-called homologous fragments, and so on.
The problem of discovering statistically covert dependences and dependent texts is solved, to which purpose a number of new statistical models (or hypotheses) are formulated. They are then checked against sufficiently extensive experimental data consisting of concrete narrative texts like annals or chronicles. It turns out that the suggested models can be confirmed.
For the reader's convenience, we divide the contents into several 'topics', which may be helpful in getting oriented in the material and in separating reliable statistical evidence from hypotheses. This devision is arbitrary in the sense that the highlights listed in the following are intimately related to each other. Therefore, it would be more correct to speak of the book's 'fibres', rather than of its parts. The book's chapters receive different emphasis, and I will briefly describe this accentuation here. I hope that the reader will be able to relate each fragment of the book to some particular 'fibre', and, in particular, to make out the author's attitude toward each fibre.
The first fibre.
In other words, we managed to discover interesting statistical regularities controlling the chroniclers' process of creating long narrative texts. The discovery of these laws is one of the principal results of our work. And on their basis, the methods for dating the events described are offered, for which the texts under investigation are statistically compared with those whose dating is undisputed. The methods are then verified against sufficiently extensive concrete material. We see that their application to texts describing the events from the 13th to the 20th century supports the efficiency of the method. Namely, the statistical datings obtained are consistent with those that had been known previously and were established by traditional methods. In particular, textual pairs originating from common primary sources, and known a priori as dependent between the 13th and the 20th century, also turn out to be dependent from the point of view of our methods; and pairs of texts known as positively independent prove to be independent from the standpoint of our methods as well.
The discovery of the laws that govern the distribution of information in large historical texts, with the establishment and experimental verification (based on these laws) of new dating methods (there being eight of them at present), is the first basic result of our work. Certainly, the dates we obtained cannot be regarded as absolute and final. Therefore, we will speak in the following only of 'statistical datings', although, for brevity, we will sometimes omit the term 'statistical', which is always implied. We thereby regard the obtained empirico-statistical dates only as a formal result of the statistical experiments carried out with narrative texts and do not believe that they are undisputed. Meanwhile, the consistency of these dates with those known earlier and obtained by the classical methods points to the objective character of our results.
