def _end_node(self):
# TODO: make error msg more specific
if not self._comp_attrs.get("label"):
raise DaesoError("node must have a 'label' attribute")
if not self._comp_attrs.get("tokens"):
raise DaesoError("node must have a 'tokens' attribute")
self._node = None
self._comp_attrs = None
def _end_graph(self):
if len(self._graph) == 0:
msg = "encountered empty graph (id=%s) while parsing graphbank" % self._graph.id
raise DaesoError(msg)
self._update_tokens()
self._graph = None
def _check_for_unresolved_graphs(self, inf):
unresolved = [ '"{0}"'.format(id)
for (id, graph) in self._id2graph.items()
if graph is None ]
if unresolved:
msg = ( "graphs with the following id's were not found"
' in graphbank file "{0}": {1}'.format(
inf.name,
", ".join(unresolved) ))
raise DaesoError(msg)
def _end_graph(self):
if self._graph_wanted():
if self._graph.tokens is None:
raise DaesoError("graph must have a 'tokens' attribute")
# set the root node, defined as a graph attribute like
# <data key="root">n0</data>
self._graph.root = self._graph.graph.get("graph")
self._graph = None
self._comp_attrs = None
def _get_node_label(self, attrs):
"""
return node label, which is the 'pos', 'cat' or 'index' attribute, in
that order
"""
for key in ("pos", "cat", "index"):
try:
return attrs[key]
except KeyError:
pass
else:
raise DaesoError("Alpino <node> element must have "
"'pos', 'cat' or 'index' attribute")
def _lookup_attr_name(self):
# lookup the name of the component attribute for the current key id in
# self._data_key, which is the value of the "key" attribute on the
# current <data> element
# the tag of the parent of <data>
key_for = self._context[-2]
try:
return self._key_table[key_for][self._data_key]["name"]
except KeyError:
pass
# if the name is not in the key table specifically for this component,
# then it _must_ be in the general property table
try:
return self._key_table["all"][self._data_key]["name"]
except:
raise DaesoError("encountered undefined key " +
repr(self._data_key))
def __init__(self,
function=ff_none,
type="i",
metric="O",
name=None,
pp_graph_hooks=[],
pp_node_hooks=[]):
"""
@keyword function: feature function which is called to compute the
feature's value in a particular instance
@keyword type: Numpy data type
@keyword metric: Timbl feature metric (e.g. "O","N", ...).
@keyword name: alternative feature name; by default the feature name
is derived from the name of the feature function by striping the
initial "ff_" string
@keyword pp_graph_hooks: a list op preprocessing hooks, where each
preprocesing function takes a pair of source and target graphs as
input
@keyword pp_node_hooks: a list op preprocessing hooks, where each
preprocesing function takes a pair of source and target nodes as
input
"""
self.type = type
self.function = function
self.init_metric = self.metric = metric
self.pp_graph_hooks = pp_graph_hooks
self.pp_node_hooks = pp_node_hooks
if name:
self.name = name
elif function.__name__.startswith("ff_"):
self.name = function.__name__[3:]
else:
raise DaesoError("cannot infer feature name from function named " +
function.__name__)