|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
• |
Well-understood
FST technology
|
|
|
|
– |
broad coverage,
accurate, efficient
|
|
|
|
– |
1-2 person-years
per language
|
|
|
|
• |
native
speaker/linguist/programmer combination
|
|
|
|
– |
application
adaptation is also understood
|
|
|
• |
More informal
approaches
|
|
|
|
– |
often less
demanding applications such as stemmers
|
|
|
– |
sui generis
algorithms developed for single language
|
|
|
|
– |
typically more
human effort
|
|