grass@uiuccsb.UUCP (10/18/83)
#R:rocheste:-339200:uiuccsb:10500007:000:2120 uiuccsb!grass Oct 17 10:40:00 1983 Slavic languages are a mixed bag, as far as verb treatment goes. The south Slavic ones (Serbo-Croatian, Bulgarian, Macedonian, Slovenian) have retained a lot of the old Slavonic forms.. archaic in Serbo-croat, active in Macedonian and Bulgarian.. These last two have all the intricacies of Czech with an additional set of verb forms used especially in the past to distinguish between what was actually WITNESSED vs. that which is known only on a second hand basis. This is in addition to the perfective/ imperfective split. I guess there is one small compensation... Macedonian and Bulgarian have no case declensions for their nouns. (Which is what is generally the most difficult about learning a Slavic language). Re: machine translation.. my favorite example of the problems in that is the machine translation Russian to English of the phrase: "The spirit is willing, but the flesh is weak" It came out: "The liquor is good, but the meat is poor". I have done professional translating in my time. (Mostly various Slavic languages), and it usually is a rather difficult process involving at least 4 steps. 1) Reading the whole thing to get a meaning out of it. 2) A first pass translation from the original with dictionary and whatever else. 3) A pass over the translation without the original to get it into literate, smooth English 4) A verification pass comparing the translation against the original to be sure the meaning has not been altered. This is usually enough for a technical translation (e.g. articles on PVC used as roofing materials from the original Czech (I've actually done that one)), but nowhere near enough for anything that is literary (e.g. papers on Orthodox mysticism as reflected in certain Russian theological works) (another one I've done). And if the translation will be published, there is more editing to be done. In general I am pessimistic about machine translation because of the poor fit of semantics between any two given languages and because I am not convinced anyone has a good enough model of natural language syntax and semantics. (Especially semantics).