My interest in NLP/MT is multifaceted. My background is in software engineering. I am currently a student in the Linguistics Department at UMD, and working in the NLP group. In addition, I studied for two years with future translators at SIL International, and spent a year as a analyst in their software development group.
Many SIL linguists believe it is a waste of time to study NLP, since translation requires too much adaptability and cultural understanding for a computer to do well. At one level, I agree with them: fully automated MT will never replace a human translator, except in limited domains. However, I still believe that there are many areas in which NLP can take some of the workload off the translator, freeing the translator to focus on the more difficult problems. I therefore see my goal as being Computer-Aided Translation (CAT) rather than fully automated MT.
Computers have already had a huge impact on translation fieldwork. Word-processing was a quantum leap forward from the shoebox and filecards of the old days. Data storage and analysis tools have provided more recent advances. A task that used to take a pair of linguists 20-25 years can now be completed in half the time.
I expect that the next generation of tools for translation fieldwork will be linguistically informed, and capable of inferring linguistic concepts through supervised learning. The linguist will have more aids for analysis, data storage, and generation. I really don't know how far computers can go in aiding the translator's task. My career goal is to find out.
Bible translation offers several unusual technical challenges:
Several different approaches can help address these issues:
Written: September 2004.