def print_map(s: int,
z: int,
a: 'np.array',
pz: float,
pa: 'np.array',
e_corpus: Corpus,
f_corpus: Corpus,
ostream=sys.stdout):
e_snt = e_corpus.sentence(s)
f_snt = f_corpus.sentence(s)
tokens = []
for j, (i, p) in enumerate(zip(a, pa)):
tokens.append('%d:%s|%d:%s|%.2f' % (j + 1, f_corpus.translate(
f_snt[j]), i, e_corpus.translate(e_snt[i]), p))
print('%d|%.2f ||| %s' % (z, pz, ' '.join(tokens)), file=ostream)
def print_lola_format(sid, alignments, posterior, e_corpus: Corpus, f_corpus: Corpus, ostream):
"""
Print alignment in a human readable format.
:param e_corpus: data we condition on
:param f_corpus: data we generate
:param sid: sentence id
:param alignments: alignments (sequence of a_j values for each j)
:param posterior: posterior p(a_j|f,e)
:param ostream: where to write alignments to
:return:
"""
e_snt = e_corpus.sentence(sid)
f_snt = f_corpus.sentence(sid)
# in printing we make the French sentence 1-based by convention
# we keep the English sentence 0-based because of the NULL token
print(' '.join(['{0}:{1}|{2}:{3}|{4:.2f}'.format(j + 1,
f_corpus.translate(f_snt[j]),
i,
e_corpus.translate(e_snt[i]),
p)
for j, (i, p) in enumerate(zip(alignments, posterior))]),
file=ostream)
def print_lola_format(sid, alignments, posterior, e_corpus: Corpus,
f_corpus: Corpus, ostream):
"""
Print alignment in a human readable format.
:param e_corpus: data we condition on
:param f_corpus: data we generate
:param sid: sentence id
:param alignments: alignments (sequence of a_j values for each j)
:param posterior: posterior p(a_j|f,e)
:param ostream: where to write alignments to
:return:
"""
e_snt = e_corpus.sentence(sid)
f_snt = f_corpus.sentence(sid)
# in printing we make the French sentence 1-based by convention
# we keep the English sentence 0-based because of the NULL token
print(' '.join([
'{0}:{1}|{2}:{3}|{4:.2f}'.format(j + 1, f_corpus.translate(f_snt[j]),
i, e_corpus.translate(e_snt[i]), p)
for j, (i, p) in enumerate(zip(alignments, posterior))
]),
file=ostream)
