Our idea was to write a program which would allow us to calculate the diachronical evolution of romance languages. Those languages present the following characteristics:
  1. both the mother and the daughter language are known (latin vs italian/french/spanish/portuguese etc.)
  2. the evolution process is also known (intermediate steps of the evolution are documented)

In a very simple way, the evolution process of such a language can be described by a set of ordered rules which is applied to a set of words. The calculation starts with the first rule (r1), which transformes the set (L1 > L2). This new set of words (L2) becomes the input for the next rule (r2). After the last rule has been aplied we obtain the set Ln which represents the language in its actual form:

This is of course a very simple view of the problem: it is evident that there can be words in L1 which do not exist any more in Ln (those are words that have been lost in the evolution process). On the other hand, there might be words in Ln which didn't exist in L1 (e.g. neologisms). Nevertheless, this model can be useful for the calculation of many words and so we choose it for our program.


We knew from the beginning that the problem of language evolution is very complex. That's why we decided to limit the project: only phonetical phenomena would be calculated by the program, that is we would neglect gramatical, analogical or other phenomena of the evolution. (*)


From the beginning we wanted to separate the program from the linguistical rules. We conceived the program as an interpreter of linguistical rules. This interpreter needs two types of information: 1) definitions of the features and sounds that are going to be used, 2) the rules, that describe the evolution of a language. By modifying the sound definitions and the set of rules, it should then be possible to calculate the evolution of any language (the only restriction is that it must be possible to describe the sounds of the language by distinctive attributes)


For the calculation, every word is translated into a list of distinctive attributes. The rules are also translated into an internal structure which makes it easier for the program to interpret them. During the calculaction, the rules operate modifications on the disctinctive attributes of the word. In order to present the results of the calculation in an easier way, the word is finally retranslated into a chain of phonemas.


In order to test our program, we decided to define a set of rules valid for the evolution from latin to spanish. We choose that language because in comparison with other romance languages (especially french) it has undergone a relatively modest number of modifications. A part from that, the fact that such a set of rules has already been proposed by C. Lleal (see bibliography) would considerably facilitate our work.

(*) although such information can also be treated by the program up to a certain degree

Last updated: October 2000 Back to main page