Hi, Do the following sections from Koehn's pseudocode align to the E and M steps of the EM algorithm?
E-Step set count(e|f) to 0 for all e,f set total(f) to 0 for all f for all sentence pairs (e_s,f_s) for all words e in e_s total_s(e) = 0 for all words f in f_s total_s(e) += t(e|f) for all words e in e_s for all words f in f_s count(e|f) += t(e|f) / total_s(e) total(f) += t(e|f) / total_s(e) M-Step for all f for all e t(e|f) = count(e|f) / total(f) The reason I ask is that while I can see the parameters being updated in the M-Step it really doesn't look like any kind of maximisation logic is being applied. The section I labelled E-Step looks more like Maximisation logic. Also in the section I label E-step there doesn't seem to any kind of alignment logic going on. Or is that what the continual update of counts and totals amounts to? Is this implicitly aligning words on a per sentence basis? Thanks guys. I'd really appreciate an answer. It's the last week to dissertation hand in and I'm writing up and don't want to make any silly mistakes. James -- The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. _______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
