def __init__(self, initial_tagger, templates, trace=0,
**property_names):
self._initial_tagger = initial_tagger
self._templates = templates
self._trace = trace
self._property_names = property_names
PropertyIndirectionMixIn.__init__(self, **property_names)
def __init__(self, preterminal_tags=False, **property_names):
"""
@param preterminal_tags: If true, then treat preterminal
nodes as tags.
@type preterminal_tags: C{boolean}
"""
PropertyIndirectionMixIn.__init__(self, **property_names)
self._preterminal_tags = preterminal_tags
self._source = None # <- not thread-safe.
def __init__(self, encoders, **property_names):
"""
Create a new merged feature encoder.
@param encoders: The basic feature encoders whose output
should be combined to form this encoder's output.
"""
PropertyIndirectionMixIn.__init__(self, **property_names)
self._encoders = encoders
def __init__(self, initial_tagger, rules, **property_names):
"""
@param initial_tagger: The initial tagger
@type initial_tagger: L{TaggerI}
@param rules: An ordered list of transformation rules that
should be used to correct the initial tagging.
@type rules: C{list} of L{BrillRuleI}
"""
self._initial_tagger = initial_tagger
self._rules = rules
PropertyIndirectionMixIn.__init__(self, **property_names)
def __init__(self, top_node='S', chunk_node='CHUNK', **property_names):
"""
@include: AbstractTokenizer.__init__
@type chunk_node: C{string}
@param chunk_node: The node label that should be used for
chunk subtrees. This is typically a short string
describing the type of information contained by the chunk,
such as C{"NP"} for base noun phrases.
"""
PropertyIndirectionMixIn.__init__(self, **property_names)
self._chunk_node = chunk_node
self._top_node = top_node
def __init__(self, **property_names):
"""
Construct a new tokenizer.
@type property_names: C{dict}
@param property_names: A dictionary that can be used to override
the default property names. Each entry maps from a
default property name to a new property name.
"""
if self.__class__ == AbstractTokenizer:
raise AssertionError, "Abstract classes can't be instantiated"
PropertyIndirectionMixIn.__init__(self, **property_names)
def __init__(self, reverse=False, **property_names):
"""
Construct a new sequential tagger.
@param reverse: If true, then assign tags to subtokens in
reverse sequential order (i.e., from right to left).
@type property_names: C{dict}
@param property_names: A dictionary that can be used to override
the default property names. Each entry maps from a
default property name to a new property name.
"""
self._reverse = reverse
PropertyIndirectionMixIn.__init__(self, **property_names)
def __init__(self, **property_names):
"""
Create a new stemmer.
@type property_names: C{dict}
@param property_names: A dictionary that can be used to override
the default property names. Each entry maps from a
default property name to a new property name.
"""
# Make sure we're not directly instantiated:
if self.__class__ == AbstractStemmer:
raise AssertionError, "Abstract classes can't be instantiated"
PropertyIndirectionMixIn.__init__(self, **property_names)
def __init__(self, chunk_types = ['LOCATION', 'ORGANIZATION', 'PERSON',
'DURATION', 'DATE', 'CARDINAL', 'PERCENT', 'MONEY', 'MEASURE'],
**property_names):
"""
Create a new C{IeerChunkedTokenizer}.
@type chunk_types: C{string}
@param chunk_types: A list of the node types to be extracted
from the input. Possible node types are
C{'LOCATION'}, C{'ORGANIZATION'}, C{'PERSON'},
C{'DURATION'}, C{'DATE'}, C{'CARDINAL'}, C{'PERCENT'},
C{'MONEY'}, C{'MEASURE'}
"""
PropertyIndirectionMixIn.__init__(self, **property_names)
self._chunk_types = chunk_types
def __init__(self, base_encoder, C=None, **property_names):
"""
@param C: The correction constant for this encoder. This
value must be at least as great as the highest sum of
feature vectors that could be returned by C{base_encoder}.
If no value is given, a default of C{len(base_encoder)} is
used. While this value is safe (for boolean feature
vectors), it is highly conservative, and usually leads to
poor performance.
@type C: C{int}
"""
PropertyIndirectionMixIn.__init__(self, **property_names)
self._encoder = base_encoder
if C is None:
self._C = encoder.num_features()
else:
self._C = C
def __init__(self, classes, weights, **property_names):
"""
Construct a new conditional exponential classifier model.
Typically, new classifier models are created by
C{ClassifierTrainer}s.
@type classes: C{list}
@param classes: A list of the classes that can be generated by
this classifier. The order of these classes must
correspond to the order of the weights.
@type weights: C{list} of C{float}
@param weights: The feature weight vector for this classifier.
Weight M{w[i,j]} is encoded by C{weights[i+j*N]}, where
C{N} is the length of the feature vector.
"""
PropertyIndirectionMixIn.__init__(self, **property_names)
self._classes = classes # <- order matters here!
self._weights = weights
def __init__(self, feature_name, values, **property_names):
"""
Create a new feature encoder that encodes the feature with the
given name.
@type feature_name: C{string}
@param feature_name: The name of the feature to encode.
@type values: C{list}
@param values: A list of the feature values of subvalues that
the feature is known to take. A feature vector index
will also be created for unseen values.
"""
PropertyIndirectionMixIn.__init__(self, **property_names)
self._feature_name = feature_name
# Initialize the mappings between basic values and feature
# vector indices. Reserve index 0 for unseen feature values.
self._index_to_val = ['<unknown>']+list(values)
self._val_to_index = dict([(v,i+1) for (i,v) in enumerate(values)])
def __init__(self, **property_names):
PropertyIndirectionMixIn.__init__(self, **property_names)
# A token reader for processing sentences.
self._sent_reader = ChunkedTaggedTokenReader(
top_node='S', chunk_node='NP', **property_names)
def __init__(self, **property_names):
PropertyIndirectionMixIn.__init__(self, **property_names)
def __init__(self, chunk_types=None, **property_names):
PropertyIndirectionMixIn.__init__(self, **property_names)
self._chunk_types = chunk_types