For normal beam search, a traceback is needed to recover the
word string after the Viterbi search. For this reason, the
memory must contain information which word was selected for
each frame. This requires memory cells. A
capacity of
is however sufficient, if we store likelihoods and word strings as
together, where
is the number of
words in a sentence. In this case, no traceback is needed.
Usually,
, the number of words in a sentence, is
smaller than
, the number of input frames, so the
memory is reduced but the calculation cost is slightly
increased.