def run(pipeline):
'''Run a complete pipeline of graph operations.
Parameters
----------
pipeline : dict
The pipeline description.
Returns
-------
The result from running the pipeline with the provided arguments.
'''
# INPUT
if 'graph' in pipeline:
graph = pipeline['graph']
elif 'text' in pipeline:
ts = []
try:
ts.append(pipeline.get('parser', DEF_PARSER) + '_parse')
except KeyError:
pass
try:
ts += pipeline.get('transformers', DEF_TRANSFORMERS)
T = transformers.get_pipeline(ts)
except KeyError:
raise ValueError("Unknown transformer pipeline")
T_args = pipeline.get('transformer_args', DEF_T_ARGS)
graph = CG(transformer=T,
transformer_args=T_args,
text=pipeline['text'])
else:
raise ValueError('Must provide either graph or text')
# OPERATIONS
for operation in pipeline.get('operations', []):
try:
name = operation.pop('op')
except KeyError:
raise ValueError("No name for the operation")
try:
graph = operate(graph, name, **operation)
except TypeError as e:
raise ValueError(e)
operation['op'] = name
# OUTPUT
if 'linearizers' in pipeline:
try:
L = linearizers.get_pipeline(
pipeline.get('linearizers', DEF_LINEARIZERS))
except KeyError:
raise ValueError("Unknown linearizer pipeline")
L_args = pipeline.get('linearizer_args', DEF_L_ARGS)
return graph.linearize(linearizer=L, linearizer_args=L_args)
else:
return graph
def __init__(self, text, superficial=False, autocorrect=True):
if autocorrect:
self.text = correct_phrase(text)
else:
self.text = text
if superficial:
self.parse = nlp(self.text)
else:
self.graph = CG(transformer=semantic_analyzer, text=self.text)
self.parse = self.graph.spacy_parse
def __init__(self, text, superficial=False, autocorrect=True):
# Corrects posible misspelled words in the queries
if autocorrect:
self.text = correct_phrase(text)
else:
self.text = text
if superficial:
# Parses the query without generating the semantic graph
self.parse = nlp(self.text)
else:
# Deeply parses the query generating the semantic graph
self.graph = CG(transformer=semantic_analyzer, text=self.text)
self.parse = self.graph.spacy_parse
def operate(graph, **args):
subgraph, main_entity = spot_domain(graph)
r = CG(graph, subgraph=subgraph)
r.gram['main_entity'] = main_entity
return r
def compose_answer(question, answer_graph):
g = CG(original=question.graph)
graft(g, question.graph.questions[0], answer_graph, answer_graph.roots[0])
return (g.linearize(linearizer=nlg))
#!/usr/bin/env python3 from grafeno import Graph as CG, transformers as tr, linearizers as ln T = tr.get_pipeline(['pos_extract', 'thematic', 'phrasal', 'wordnet']) L = ln.get_pipeline(['simple_nlg']) g1 = CG(transformer=T, text="A man picked up an apple.") g2 = CG(transformer=T, text="The woman lifted some pears.") import grafeno.operations.generalize as gen g3 = gen.generalize(g1, g2, node_generalize=gen.wordnet_generalize) print(g3.linearize(linearizer=L))
arg_parser = argparse.ArgumentParser(
description='Example topic detection script')
arg_parser.add_argument('file', type=argparse.FileType('r'))
arg_parser.add_argument('-l',
'--lang',
help='language of the text',
default='en')
args = arg_parser.parse_args()
T = tr.get_pipeline(
['spacy_parse', 'pos_extract', 'thematic', 'phrasal', 'wordnet'])
g = CG(transformer=T,
transformer_args={
'lang': args.lang,
'sempos': {
'noun': 'n'
},
'unique_gram': {
'hyper': [True]
},
'extended_sentence_edges': ['HYP']
},
text=args.file.read())
g = operate(g, 'cluster', hubratio=0.2)
for hvs in g.gram['HVS']:
keywords = set(g.node[n]['concept'] for n in hvs)
print('Topic: {}'.format(', '.join(keywords)))