CORDIS - EU research results

Measuring and Modelling Language Interactions

Final Report Summary - MEMOLI (Measuring and Modelling Language Interactions)

Large-scale migrations result in different languages and cultures getting in contact with each other. It happened so in the past, and continues happening so in the 21st century Europe. Jewish languages, such as Yiddish, Ladino, Judeo-Arabic or Judeo-Italian, represent eminent examples of such linguistic and cultural interactions, and they have served as motivating examples for our project.

The language knowledge ("competence") in a speaker's mind has been modelled by two linguistic frameworks: Optimality Theory and Harmonic Grammar. The two approaches are very close in that both are constraint-based. Yet, these constraints are combined differently into a single grammar. Moreover, a single speaker growing up in a multilingual environment may develop parallel grammars, each corresponding to a separate language, or language variety, in the environment. Language acquisition is never perfect, and the linguistic input from different languages (or language varieties) may also influence the process. Language production is not perfect, either, as speakers tend to make mistakes due to a number of reasons.

The research fellow, earlier in his career, had developed a symbolic simulated annealing model of language production based on Optimality Theory. In the framework of the current project, he had developed this model further in order to be able to understand contact language situations, such as those mentioned above. In particular:

1. He has compared the behaviour of Optimality Theory to Harmonic Grammar when both are implemented using simulated annealing. It turns out that the implementation of Optimality Theory is faster, but more prone to errors.

2. He has developed a learning algorithm to model the acquisition of a language that presupposes the learner to entertain parallel grammars, as it is expected in the case of a multilingual environment.

3. He has applied Enoch O. Aboh’s notion "hybrid grammars" to interpret Yiddish plural morphology, using Luigi Burzio's notion of "representational entailments" implemented computationally.

Beside these computational experiments, he has also carried out two experiments, both focussing on Hungarian vowel harmony. In one of them, he demonstrated that production frequencies may be dependent on speech rate, corroborating his simulated annealing model. In the other experiment, he has collected data for the hypothesis that speakers may take into account a word's origin ("naive etymology") even when no explicit co-phonologies are present in the language.