From Babylon to Bergen: On the usefulness of aligned texts

Signe Oksefjell Ebeling,Jarle Ebeling

doi:10.15845/bells.v3i1.359

Abstract

After outlining a short and select history of (the usefulness of) parallel texts and alignment, this paper presents a case study where the point of departure is a Norwegian text extract aligned against its translations into seven different target languages, using the Translation Corpus Aligner, originally developed by Knut Hofland. Our main concern is cases where there is not a one-to-one correspondence at sentence level between original (source) and translation (target) text. We seek to answer questions such as why a translator, translating into a specific language has chosen to split, or merge, a sentence in the source texts, while translators, translating into other languages have chosen not to do so. The study shows that a multitude of contributing factors seem to be involved , including author and translator style, target language constraints and preferences and perhaps even country- or language-specific translation guidelines. Keywords : alignment; parallel texts; contrastive analysis; corpora; translation strategies

Highlights

In this article we wish to pay tribute to Knut Hofland and his contribution to the field of corpus linguistics by focusing on his collaboration with the late Stig Johansson, in connection with the development of the English-‐Norwegian Parallel Corpus (ENPC) and the Translation Corpus Aligner (TCA).In 1977, Stig Johansson went to Bergen to attend an introductory course in humanities computing; here he met the young computational engineer Knut Hofland
Our point of departure in this part of the paper will be two articles which both report on some basic statistics between original and translated texts aligned with the Translation Corpus Aligner
In Johansson & Hofland (2000) a small multilingual corpus of six English source texts and their translations into German and Norwegian are studied, while Johansson (2011) performs studies of a multiple translation corpus consisting of two short English texts and their ten different translations into Norwegian

Summary

Introduction

In this article we wish to pay tribute to Knut Hofland and his contribution to the field of corpus linguistics by focusing on his collaboration with the late Stig Johansson, in connection with the development of the English-‐Norwegian Parallel Corpus (ENPC) and the Translation Corpus Aligner (TCA). The text extract below is from a clay tablet from Assyria from the 2nd millennium BCE (see the Appendix for a drawing of the original tablet) It contains a parallel, aligned text of transliterated cuneiform signs, where the source text (in bold) is in Sumerian and the target text (in italics) is in Akkadian. Having stored electronically combined bi-‐texts in the form of segments of source and target text, Harris imagines a database of hyper-‐bi-‐texts, i.e. a hypertext base or translation memory, where the translator can search his or her own previous work Such a hyper-‐bi-‐text system should include a search engine “programmed in such a way that when it finds an occurrence of the word [one is looking for], it retrieves and displays the whole translation unit in which it occurs” (ibid.: 9). In the following we explore how aligned texts can be used to highlight differences between languages and between source and target text

Background

Case study

Basic statistics

Conclusion