No abstract

Kao globalni jezik modernoga doba engleski je postao dominantan jezik davatelj. Danas se smatra da hrvatski jezik najviše posuđuje upravo iz engleskoga. Utjecaj engleskoga jezika na hrvatski vidljiv je u različitim funkcionalnim stilovima te na gotovo svim jezičnim razinama, no najizraženiji je na leksičkoj razini. U novije vrijeme, posebice u medijima i na društvenim mrežama, sve se češće javljaju neprilagođene engleske riječi, tj. riječi koje su zadržale izvorni oblik, a kojima se po potrebi dodaju hrvatski afiksi. Za sada još uvijek ne postoje konkretni podaci o takvim riječima u hrvatskome jeziku. U cilju pronalaženja engleskih riječi, u drugim su se jezicima koristile različite metode, od ručnih klasifikacija i korištenja postojećih jezičnih resursa do razvoja novih alata i/ili resursa. Međutim, jezične tehnologije za hrvatski jezik još uvijek su nedostatno razvijene. Stoga je cilj ovoga rada ispitati mogućnosti nekih od postojećih alata i resursa za crpljenje engleskih riječi i stvaranje baze engleskih riječi. U tu svrhu pretraživan je mrežni hrvatskog jezika hrWaC pomoću platforme Sketch Engine. Ovom metodom dobiven je popis od 1217 engleskih riječi. Rezultati su pokazali da se pomoću dostupnih alata i resursa za hrvatski jezik može izraditi popis engleskih riječi i njihovih frekvencija, ali i da postoje brojni problemi zbog kojih se rezultati ne mogu smatrati u potpunosti pouzdanima. Isto tako, sam se postupak i dalje mora kombinirati s ručnim metodama i klasifikacijama. Zaključujemo da je za izradu cjelovite baze engleskih riječi u hrvatskome potrebno razviti nove alate i resurse koji bi omogućili automatsko crpljenje engleskih riječi iz korpusa hrvatskoga jezika.

Lexical bundles are recurring frequent word combinations. Research has shown that lexical bundles vary in genre and register (Biber 2006; Biber, Conrad and Cortes 2004; Hyland 2008a, 2008b; Scott and Tribble 2006). However, the degree to which they vary by discipline remains inconclusive. The main aim of this paper is to establish whether lexical bundles are discipline specific, i.e., whether each discipline draws on a specialized lexical repertoire or whether there is a core vocabulary shared across various disciplines. For that purpose, maritime texts covering the subdomains marine engineering, navigation, maritime law and shipping have been collected so as to investigate the structure and function of lexical bundles and to find out how they shape meaning in specialized discourse. For the purposes of the study, a 7.4 M corpus consisting of two monolingual subcorpora and one bilingual subcorpus was compiled. This corpus can be used as a basis for further studies in the field. Furthermore, the paper discusses problems encountered while extracting N-grams from a corpus, as well as classification criteria for the identification of lexical bundles. The results show that lexical bundles identified in maritime texts are phrasal rather than clausal. The results also indicate that lexical bundles are discipline specific. Teaching these specialized features that shape discourse can improve students’ language production and should thus be the focus of instruction in ESP.

Translation research focuses mainly on parallel and comparable corpora, whereby it is constantly faced with issues of representativeness, balance and comparability as its main constraints. This research aims to introduce the concept of genre as a way of observing linguistic features under controlled conditions. The study analyses the application of external and internal criteria with particular focus on the genre criterion in selecting texts for the compilation of a highly-specialized bilingual maritime legal corpus, consisting of source texts in English and their translations into Croatian. The main advantages and constraints of genre as a criterion are discussed. The main benefits of such an approach are found in its application in translator training and practice. In addition, genre-based approaches to corpus analysis may raise awareness of generic features specific to a target language, ultimately improving the quality of translation.