Distinguishing translations from non-translations and identifying (in)direct translations’ source languages

Abstract

<p>The scope of this study is threefold. First, machine learning will be applied to<br />distinguish translated from non-translated Finnish texts. Then, it will attempt to<br />identify the source languages of the translated Finnish texts. Finally, the source<br />language identification will be tested with indirect translations, that is, with<br />translations made from translations. The three underlying research questions are: 1)<br />Can translated Finnish be distinguished from non-translated Finnish? 2) Can the<br />source languages of Finnish translations be identified? 3) If the answer to question<br />2 is yes, then what happens when the method is applied to indirect translations; will<br />the analysis identify the ultimate source language, the mediating language, or<br />neither?</p><p>This study is based on the hypothesis that translated language contains traces<br />of the source language (Toury 1995). The corpus of the study consists of nontranslated<br />Finnish prose, Finnish prose literature translations made from English,<br />German, French, Modern Greek, and Swedish, as well as indirect translations from<br />Modern Greek into Finnish via English, German, French, and Swedish. The<br />analyses are based on cluster analysis and support vector machines using the<br />frequencies of the most frequent lemmatized words.<br />Results show that translated and non-translated Finnish can be distinguished<br />by using machine learning techniques. Support vector machine-based source<br />language identification, however, was only partially successful, while a cluster<br />analysis suggested that there is coherence within a group of texts translated from<br />the same source language and variation between the groups of texts with different<br />source languages. Clustering was further tested with indirect translations, and the<br />results were mixed: six of the thirteen tested indirect translations clustered with<br />direct translations from the ultimate source language, two with translations from<br />their mediating languages, and five with neither.<br /></p>