def _single_run(self, graph):
# called for optional debug logging
GraphProcess._single_run(self, graph)
seg_counts = self.counts.next()
for u in graph.source_nodes_iter():
self._add_normalized_scores(
*self._score_translations(graph, u, seg_counts))
def _single_run(self, graph):
# TODO: handle hypernodes
GraphProcess._single_run(self, graph)
# skip normalization, because base scores are already normalized
for u in graph.source_nodes_iter():
# first pass to figure out which score attr is present
base_score_attr = self._find_base_score_attr(graph, u)
# second pass to add the best scores
for u, v, data in graph.trans_edges_iter(u):
data[self.score_attr] = data.get(base_score_attr, 0.0)
def _single_run(self, graph):
# called for optional debug logging
GraphProcess._single_run(self, graph)
seg_counts = self.counts.next()
# skip if source lempos is not in ambiguity map or is filtered
for u in graph.source_nodes_iter():
try:
source_lempos = u" ".join(graph.node[u]["lex_lempos"])
except:
# source lempos was not in lexicon (e.g. punctuation)
continue
if source_lempos not in self.ambig_map:
continue
if self.filter(graph, u):
continue
self._add_normalized_scores(
*self._score_translations(graph, u, seg_counts))
def __init__(self, score_attr="best_score", unknown=None):
GraphProcess.__init__(self)
self.score_attr = score_attr
self.unknown = unknown