Introduction

Is structural difference in a language pair associated with difficulty in translating or interpreting complex sentences in that pair? If so, are there differences in that association between the three modes of language transfer considered – legal translation, subtitle translation and simultaneous interpretation? Those are the questions this study aims to help answer. To do so, it first proposes a method for analyzing the semantic structure of a sentence and measuring changes in that structure in translation or interpretation. Then, based on research establishing such changes as indicators of cognitive difficulty, the study seeks associations between their frequency and structural difference in the language pair of translation or interpretation.

Many factors can contribute to the difficulty of a translation or interpretation task. Some of those factors can have to do with the text or speech – like subject, terms, idioms, register, style or (for interpretation) speed. Others can have to do with the translator or interpreter – like experience, background knowledge, personal beliefs, or physical or emotional state. The task can be made more difficult by cultural differences – like history, politics, popular references or norms of politeness. The same is true of linguistic features – like writing systems, morphology, irregularity, grammatical ambiguity or homophony.

But there’s another major factor of difficulty which can often be underestimated: structural difference in the language pair of translation or interpretation – especially large-scale typological difference in the branching direction of subordinate clauses. This study seeks associations between that typological difference and three identified indicators of production difficulty in translation and interpretation. It counts the average rate for each indicator in a corpus of more than 1000 English sentences, each translated or interpreted into five languages from different families: Russian, Hungarian, Turkish, Mandarin and Japanese. It then seeks associations between those rates and the structural difference of each language pair.

Each indicator of difficulty involves relations between propositions. A proposition is the semantic relation underlying a syntactic clause. The proposition is a good unit for cross-linguistic comparison, because an event or situation can be described in different syntactic forms in different languages. This study uses the proposition as a unit of analysis to count rates for three features of translation or interpretation identified as indicators of difficulty: reordering, nesting changes and changes in semantic relations.

The study doesn’t claim to directly measure the level of difficulty in a given translation or interpretation task. It doesn’t compare the levels of difficulty reflected, say, in creating or eliminating two different nested structures. Nor does it compare the levels of difficulty reflected, say, in a nesting change and a change in semantic relations. But it does assume that, if reordering, nesting changes and changes in semantic relations are accepted as indicators of some degree of production difficulty, then, in a given sentence, a higher count for any of those indicators in one language pair suggests a greater degree of difficulty than a lower count for the same indicator in another language pair: two place shifts, two nesting changes or two changes in semantic relations reflect a higher degree of difficulty than one.

The sentences examined are from three modes of language transfer: legal translation, subtitle translation and simultaneous interpretation. For legal translation, I analyze translations of the Universal Declaration of Human Rights, the Paris Agreement on climate change and the US Foreign Corrupt Practices Act. For subtitle translation, I analyze translated subtitles for the five most popular TED talks to date at the time of writing. For simultaneous interpretation, I analyze recorded interpretation of President Obama’s 2015 speech to the UN General Assembly.

The statistical analysis uses generalized linear mixed models, applied by an expert from the statistical service of the University of Louvain. The analysis seeks associations between three independent variables – structural difference, mode of transfer and sentence complexity – and three dependent variables – the indicators of difficulty mentioned above. Strong associations between those variables are revealed by statistical analysis of the corpus data.

The study is in a website format, to make it reader-friendly and easy to browse. It consists of a Foreword and five chapters. Chapter 1 – Introduction outlines the topic and content. Chapter 2 – Literature review gives an overview of relevant research. Chapter 3 – Method and data summarizes the method used to assess the preservation of semantic relations between the original version of a sentence and the same sentence in translation or interpretation. It then presents the chosen corpus of sentences and discusses the variables counted. Chapter 4 – Analysis and results explains how the data was collected and analyzed, describes associations and tendencies found, and presents the results of a reliability check. Chapter 5 – Discussion summarizes the findings, discusses tactics for interpretation and draws conclusions on implications for the profession. There’s also an Epilogue on Other structural challenges not included in the statistical analysis. Annex I details the Semantic parsing method used. Annex II contains all Data from the sentences analyzed and data for the reliability check. A short version of the thesis is also available at: https://structural-difficulty-in-translation-and-interpretation.com/short-version.

Taken together, the findings of the study suggest that, the more a language pair differs in structure, the more difficult it may be to translate or interpret a complex sentence in that pair, and the more the meaning may be changed – with some major differences between the three modes of transfer considered.