initialize t(e|d) uniformly
do until convergence
set count(e|d) to 0 for all e,d
set total(d) to 0 for all d
for all sentence pairs {e_s,d_s}
set total_s(e) =0 for all e
for all words e in e_s
for all words d in d_s
total_s(e) +- t(e|d)
for all words e in e_s
for all words d in d_s
count(e|d) +=t(e|d)/total_s(e)
total(d) +=t(e|d) / total_s(e)
for all d
for all e
t(e|d) - count(e|d)/total(d)