Comparing compounds: A corpus study of Swedish compound nouns in English-German contrast

This bidirectional corpus study investigates Swedish compound nouns in English and German contrast. On the face of it, there seem to be cross-linguistically similar compounding preferences (dopaminsystem (Sw.); dopamin system (En.); Dopaminsystem (Ge.)). However, our translation corpus data partly indicate considerable differences, as illustrated in (1) and (2):

(1)
    krigsslutet (LEGS; Sw. original)
    war-[s]-end-DEF
    the end of the war (En. translation)
    das Kriegsende (Ge. translation)
    the-DEF war-[s]-end

(2)
    övergångsperiod (LEGS; Sw. original)
    transition-[s]-period
    transitional period (En. translation)
    Übergangszeit (Ge. translation)
    transition-[s]-period

In both (1) and (2), the German translations match the Swedish compound (see, Teleman et al. 1999: II: 42–47), while English in (1) opts for a postmodifying prepositional phrase and a premodifying adjective in (2). Previous studies (Ström Herold & Levin forthcoming; Berg 2017) have indicated about 70% translation correspondence rates between English solid compounds / noun sequences (for this term, see Biber et al. 2021[1999]: 582) and German and Swedish compounds. But it is unclear if these observed rates are “skewed” due to the approach of using English structures as the starting point. To address this issue, the present study instead departs from Swedish and aims to explore: (i) the proportions of English and German compounds or noun sequences as translation correspondences to Swedish compounds, (ii) the distributions of other types of correspondences such as adjectives, and (iii) what these results reveal in terms of language preferences and translation effects. 
    The study is based on the non-fiction Linnaeus University English-German-Swedish corpus (LEGS) (Ström Herold & Levin 2019; forthcoming). LEGS consists of, e.g., popular science, biographies and self-help books amounting to approximately half a million words of each source language with translations into the two other languages. We retrieved words tagged as nouns (and their translations), manually removing all Swedish non-compound nouns. The material comprises 1000 Swedish original instances with their English and German translations, and 1000 Swedish compounds translated each from English and German originals. The results are analyzed quantitatively and qualitatively. 
    Our findings demonstrate that Swedish compounds – in both originals and translations – correspond to either English noun sequences (familjemedlemmar [‘family-members’] > family members) or solid compounds (grundvatten [‘ground-water’] > groundwater) in 50% of the cases. These stable bidirectional distributions suggest that we are dealing with language preferences, rather than translation effects. Interestingly, the Swedish-to-English correspondence rates are much lower than what we found when investigating English noun sequences in Swedish translation (Ström Herold & Levin forthcoming). As for German correspondences, the rate is the highest of all: 80%. This suggests that Swedish and German are more closely aligned with each other than with English (see Carlsson 2004: 75 on German-Swedish similarities). There are two main explanations for the low Swedish-to-English correspondence rates: (i) English resorts to ‘simple’ noun correspondences (dagbok [‘day-book’] > diary) (cf. 10% in De Metsenaere et al.’s (2016) study on German-to-Dutch translation); (ii) English also uses many premodifying adjectives as correspondences, notably those ending in -al (as in (2)).