Usporedna evaluacija automatskih prijevoda primjenom tehnologije

Svrha ovog rada je predstavljanje rezultata usporedbe dva besplatna online programa za automatsko prevođenje, alata Google Translate i Yandex.Translate, za jezične parove engleski-hrvatski i ruski-hrvatski. Rad se sastoji od dvije cjeline. Prvi dio je teoretski i donosi pregled raznih aspekata automatskog prevođenja, kao što je povijesni razvoj i pregled različitih pristupa problemu automatskog prevođenja. Drugi dio donosi analizu istraživanja koje je provedeno s ciljem da se ocijeni kvaliteta navedenih alata, ali i samog automatskog prevođenja. Samo istraživanje je podijeljeno u dva dijela; u prvom dijelu su ispitanici ocjenjivali kvalitetu dobivenih prijevoda prema kriterijima točnosti (koliko dobro je značenje izvornog teksta zadržano u prijevodu) i tečnosti (koliko dobro se prevedene rečenice uklapaju u duh jezika). Drugi dio donosi rezultate analize pogrešaka provedene na odabranim rečenicama. Svrha te analize je bio pronalazak korelacija između tipova pogrešaka i ocjena iz ankete.The purpose of this paper is to present the comparison of the two of the free online programs for machine translation, Google Translate and Yandex.Translate. Language pairs used in the comparison were English-Croatian and Russian-Croatian. The paper consists of two parts. The first part is theoretical and deals with various aspects of machine translation in general, such as the historical development of machine translation and explanations of different methods used in machine translation. The second part is the analysis of a survey, which was conducted to assess both the quality of the programs in question and machine translation in general. The survey itself was divided into two parts; in the first part the task of the participants was to rate the sentences produced by machine translation according to the categories of adequacy (how much information was transferred between the original and the translation) and fluency (how well was the translation formed in terms of style). The second part brigs the error analysis of selected translated sentences. The purpose of that analysis was to find correlations between error types and general ratings from the survey