Predicting Translation Equivalents and Norm Formulation: A Study on Some EU Legislative Features

Abstract

This study explores translation patterns of EU legislative discourse from a probabilistic point of view. The main focus is on prescriptive sentences, i.e. deontic norms and performatives and their linguistic representation in a multilingual parallel corpus of EU Secondary Legislation. Although the intended message is expected to be the same for all four languages, linguistic patterns vary consistently according to the type of legislative statement and to the EU legal instrument. Information theory is applied to measure the degree of randomness occurring in the translation of particular legislative sentences and to draw possible conclusions on the standardization of EU instruments.

Notes

¹For parallel corpus, we mean here an original text in language A and its translated version in language B, C, D etc (Baker, 1993, p. 248; 1995, p. 230; Bowker & Pearson, 2002, pp. 92–93; McEnery & Wilson, 1996, p. 57).

²There is no agreement as to whether to opt for a full text document or for a text extract. A full text would broadly allow for a better representativeness because linguistic patterns can be located everywhere. On the other hand, texts can have different size, and depending on the feature to be investigated, results can be equally distorted.

³This process of incorporating and implementing the directive into the national law is called “transposition”.

⁴According to Austin (1962, p. 5) there are utterances where the “uttering of the sentence is, or is part of the doing of an action”, because it fulfils some felicitous conditions and performs the action in the same moment in which it is uttered. In other words, saying “I do” (e.g. in a wedding context) has precise legal consequences than uttering the same sentence in a non-legal setting.

⁵Retrieved September 8, 2010, from http://ec.europa.eu/translation/writing/style_guides/english/style_guide_en.pdf

⁶The whole study includes the four EU secondary legislation instruments. For reasons of space, we are presenting only data concerning regulations, which are also the most binding texts out of the other three. The final entropy discussion will, however, shed light on all the four legal instruments.

⁷Prohibitions are often considered as an obligation “not to act” and over generalizing are part of the same category of norms.

⁸It is expressed by the bit, e.g. 0/1, yes/no, or the information contained in two equiprobable options. According to Shannon (1948), a stream of data in ordinary language is less than random, each new bit carries somewhat less than a bit's worth of information. It follows from this, that the more random the data stream, the more information would be conveyed by each new bit.

⁹ p_x represents p _ind, p _mv, p _me, p _vp and p _ell of this study.

¹⁰In order to be consistent with the previous tables, we decided to show only figures and graphs for the regulation instrument, which is also the most binding.

¹¹The 1.34 in the English recommendations is due to the fact that this language has been used as a term of comparison, but the value is still higher in comparisons with the scores in the other three English instruments.