tag:blogger.com,1999:blog-6748877443699290050.post46420943272594395..comments2024-03-29T00:21:17.976-07:00Comments on eMpTy Pages: Rule-based MT vs. Statistical MT: Does it Matter?Kirti Vasheehttp://www.blogger.com/profile/16795076802721564830noreply@blogger.comBlogger1125tag:blogger.com,1999:blog-6748877443699290050.post-85130821709875462452010-03-08T14:17:54.557-08:002010-03-08T14:17:54.557-08:00This debate is apparently a hot current issue as t...This debate is apparently a hot current issue as the latest Multilingual magazine covers it some detail with an article from Lori Thicke (who is now on the board of the magazine) that covers this in some detail. <br /><br />This is a quote from the article:<br />Says Yanishevsky, “clearly, hybridization will be the development of the future for both SMT and RBMT engines. However, fundamentally, we believe that it is faster and more efficient to hybridize with rule-based underpinnings than with SMT underpinnings since it is easier to graft statistics onto rules rather than vice versa. The hybridization of our MT engine will linguistically smooth an already robust and quality output."<br /><br />I think that the enterprise decision will be made on how much control each approach gives to skilled and un-skilled users, and the ability to show ongoing improvements. I believe that the evidence already suggests that SMT will win even for Russian and German in time. I have already seen it done for Japanese.<br /><br />I will perhaps provide some counter opinions in a future blog entry but it is good to see this growing debate.<br /><br />Perhaps we are indeed heading into an MT "perfect storm".Kirti Vasheehttps://www.blogger.com/profile/16795076802721564830noreply@blogger.com