def fscore_match(self, sem1, sem2):
from jazzparser.formalisms.music_halfspan.harmstruct import \
semantics_to_dependency_graph
from jazzparser.data.dependencies import optimal_node_alignment, \
alignment_to_graph
from jazzparser.formalisms.music_halfspan.semantics import \
EnharmonicCoordinate
if sem1 is None:
max_score1 = 0.0
else:
graph1,timings1 = semantics_to_dependency_graph(sem1)
max_score1 = float(len(graph1))
if sem2 is None:
max_score2 = 0.0
else:
graph2,timings2 = semantics_to_dependency_graph(sem2)
max_score2 = float(len(graph2))
if sem1 is None or sem2 is None:
# Empty input: give zero score to everything
alignment_score = 0.0
alignment = []
transpose = None
else:
graph1,timings1 = semantics_to_dependency_graph(sem1)
graph2,timings2 = semantics_to_dependency_graph(sem2)
graphs = []
# Try all possible transpositions and assume the best
for transx in range(4):
for transy in range(3):
def _label_compare(label1, label2):
if isinstance(label1, EnharmonicCoordinate) and \
isinstance(label2, EnharmonicCoordinate):
coord1 = label1.zero_coord
x2,y2 = label2.zero_coord
return coord1 == ((x2+transx)%4, (y2+transy)%3)
else:
return label1 == label2
# Find the alignment of the nodes that matches most dependencies
alignment = optimal_node_alignment(graph1, graph2, label_compare=_label_compare)
# Get the common dependency graph
graph, node_map1, node_map2 = alignment_to_graph(alignment,
graph1, graph2, label_compare=_label_compare)
graphs.append(graph)
# Score on the basis of the shared dependencies
alignment_score,graph = max([(len(graph),graph) for graph in graphs], key=lambda x:x[0])
return alignment_score,max_score1,max_score2
if options.align_max:
graphs = []
# Try all possible transpositions and assume the best
for transx in range(4):
for transy in range(3):
def _label_compare(label1, label2):
if isinstance(label1, EnharmonicCoordinate) and \
isinstance(label2, EnharmonicCoordinate):
coord1 = label1.zero_coord
x2,y2 = label2.zero_coord
return coord1 == ((x2+transx)%4, (y2+transy)%3)
else:
return label1 == label2
# Find the alignment of the nodes that matches most dependencies
alignment = optimal_node_alignment(gold_graph, graph, label_compare=_label_compare)
# Get the common dependency graph
common_graph, node_map1, node_map2 = alignment_to_graph(alignment,
gold_graph, graph, label_compare=_label_compare)
graphs.append((common_graph, node_map1, node_map2, alignment,transx,transy))
# Score on the basis of the shared dependencies
alignment_score,common_graph,node_map1,node_map2,alignment,transx,transy = \
max([(len(g),g,nm1,nm2,al,tx,ty) for (g,nm1,nm2,al,tx,ty) in graphs], key=lambda x:x[0])
transpose = (transx,transy)
# Produce the list of node alignments we used
# Reverse the mapping from the graphs to the common graph
node_rmap1 = dict([(common,node) for (node,common) in node_map1.items()])
node_rmap2 = dict([(common,node) for (node,common) in node_map2.items()])
# Try all possible transpositions and assume the best
for transx in range(4):
for transy in range(3):
def _label_compare(label1, label2):
if isinstance(label1, EnharmonicCoordinate) and \
isinstance(label2, EnharmonicCoordinate):
coord1 = label1.zero_coord
x2, y2 = label2.zero_coord
return coord1 == ((x2 + transx) % 4,
(y2 + transy) % 3)
else:
return label1 == label2
# Find the alignment of the nodes that matches most dependencies
alignment = optimal_node_alignment(
gold_graph, graph, label_compare=_label_compare)
# Get the common dependency graph
common_graph, node_map1, node_map2 = alignment_to_graph(
alignment,
gold_graph,
graph,
label_compare=_label_compare)
graphs.append((common_graph, node_map1, node_map2,
alignment, transx, transy))
# Score on the basis of the shared dependencies
alignment_score,common_graph,node_map1,node_map2,alignment,transx,transy = \
max([(len(g),g,nm1,nm2,al,tx,ty) for (g,nm1,nm2,al,tx,ty) in graphs], key=lambda x:x[0])
transpose = (transx, transy)
