So what’s the problem?
	• Transducers and taggers cost a lot – 1-3 person-years of highly-skilled work per language – or more, from a standing start • e.g. without electronic dictionaries and tagged texts • Each language takes a long time – difficult to parallelize the work • Only about 15 languages have been “done” – coverage is uneven for these – natural rate of addition is now slow – only large, wealthy languages are likely candidates
	2/5/01