def __init__(self, transformation=LSITransformation):
self.dissimilar_sentences = SentenceGraph()
self.transformation = transformation
class Summarizer:
DISSIMILARITY_THRESHOLD = 0.8
def __init__(self, transformation=LSITransformation):
self.dissimilar_sentences = SentenceGraph()
self.transformation = transformation
def semantic_summary(self, tokenised_sentence_dict):
print "Building Dissimilar Sentences Graph"
self.build_dissimilarity_graph_using_wordnet(tokenised_sentence_dict)
print "Dissimilar Sentences Graph build complete"
cliques = self.dissimilar_sentences.multilevel_communities()
print cliques
return list(cliques)[0]
#TODO Return Sentences
def community_summary(self, tokenised_sentence_dict):
self.build_dissimilarity_graph(tokenised_sentence_dict)
community_levels = self.dissimilar_sentences.multilevel_communities()
best_community = self.best_community(community_levels, tokenised_sentence_dict)
return best_community.vs["name"]
def build_dissimilarity_graph_using_wordnet(self, tokenised_sentence_dict):
sentence_keys = tokenised_sentence_dict.keys()
synonyms = Synonyms(tokenised_sentence_dict.values())
self.dissimilar_sentences.add_nodes(sentence_keys)
connections = ConnectedNodes()
edges = list()
for every_key, tokens in tokenised_sentence_dict.iteritems():
for other_key, other_tokens in tokenised_sentence_dict.iteritems():
if connections.not_connected(every_key, other_key):
score = synonyms.dissimilarity_score(tokens, other_tokens)
if score > self.DISSIMILARITY_THRESHOLD:
edge = (every_key, other_key, {"weight": score})
edges.append(edge)
connections.add((every_key, other_key))
self.dissimilar_sentences.add_edges(edges)
def build_dissimilarity_graph(self, tokenised_sentence_dict):
print "Building Dissimilar Sentences Graph"
sentence_keys = tokenised_sentence_dict.keys()
transformed_sentences = self.transformation(tokenised_sentence_dict)
self.dissimilar_sentences.add_nodes(sentence_keys)
connections = ConnectedNodes()
edges = list()
for every_key, tokens in tokenised_sentence_dict.iteritems():
for other_key, other_tokens in tokenised_sentence_dict.iteritems():
if connections.not_connected(every_key, other_key):
score = transformed_sentences.dissimilarity_score(tokens, other_tokens)
if score > self.DISSIMILARITY_THRESHOLD:
edge = (every_key, other_key, {"weight": score})
edges.append(edge)
connections.add((every_key, other_key))
self.dissimilar_sentences.add_edges(edges)
print "Dissimilar Sentences Graph build complete"
def best_community(self, community_levels, tokenised_sentences_dict):
best_communities = self.dissimilar_sentences.find_best_community_level(community_levels)
communities_subgraphs = best_communities.subgraphs()
best_community_id = 0
best_community_index = 0.0
text_processor = TextProcessor()
for id, community in enumerate(communities_subgraphs):
vertices = community.vs["name"]
sigma_info_index = 0.0
for vertex in vertices:
sentence = tokenised_sentences_dict[vertex]
info_index = text_processor.information_index(sentence)
sigma_info_index += info_index
sigma_info_index /= float(len(vertices))
if best_community_index < sigma_info_index:
best_community_index = sigma_info_index
best_community_id = id
return communities_subgraphs[best_community_id]
