class CorpusGraph:
def __init__(self):
self.corpus = nx.DiGraph()
self.reversed_corpus_cache = None
self.corpus_io = CorpusIO()
# 需要mongodb
def build_corpus(self):
edges_gen = self.corpus_io.read_from_mongo(limit=None)
for edge in edges_gen:
self.corpus.add_edge(edge[0], edge[1], weight=edge[2])
# 将语料库的networkx实例转为json
def to_json(self):
json_obj = nx.to_dict_of_dicts(self.corpus)
return json_obj
# 将语料库的networkx实例存入硬盘，以json文件的形式
def save_as_json(self, path='./data/corpus.json'):
json_obj = self.to_json()
self.corpus_io.save_as_json(json_obj, path)
# 从json文件读取一个networkx的语料库实例
def load_from_json(self, path='./data/corpus_in_use.json'):
print("loading corpus json file: " + str(path))
json_obj = self.corpus_io.load_as_json(path)
print("loaded")
self.corpus = nx.from_dict_of_dicts(json_obj, create_using=self.corpus)
def get_edge_weight(self, start, end):
weight = 0
try:
weight = self.corpus[start][end]['weight']
except KeyError:
pass
return weight
def reverse(self):
if self.reversed_corpus_cache is None:
self.reversed_corpus_cache = self.corpus.reverse()
tmp = self.corpus
self.corpus = self.reversed_corpus_cache
self.reversed_corpus_cache = tmp
# 对于给定的字（key），取前K个最大的后接字
def get_sorted_neighbour(self, key, exclude=None, K=6):
corpus = self.corpus
# if reverse:
# corpus = self.corpus.reverse()
if key not in corpus.adj:
return []
nbr = corpus.adj[key]
rs = []
# print(nbr)
# ########### 只需要获得前K个最大值，这里的排序可以优化(堆排序/K次冒泡排序...) ####################
sorted_nbr = sorted(nbr.items(), key=lambda item: item[1]['weight'], reverse=True)
j = 0
for i in range(K - 1):
if j >= len(sorted_nbr):
break
# 循环K次，如果相邻字正好是下一个字，则跳过这个相邻字
if sorted_nbr[j][0] == exclude:
j += 1
if j >= len(sorted_nbr):
break
rs.append((sorted_nbr[j][0], sorted_nbr[j][1]['weight']))
j += 1
remain_cnt = 0
remain_weight = 0
for i in range(K - 1, len(sorted_nbr)):
if sorted_nbr[i][0] == exclude:
continue
remain_cnt += 1
remain_weight += sorted_nbr[i][1]['weight']
rs.append(("+" + str(remain_cnt), remain_weight))
return rs
class CorpusGraph:
def __init__(self):
self.corpus = nx.DiGraph()
self.corpus_io = CorpusIO()
def build_corpus(self):
edges_gen = self.corpus_io.read_from_mongo(limit=None)
for edge in edges_gen:
self.corpus.add_edge(edge[0], edge[1], weight=edge[2])
def draw(self):
nx.draw_networkx(self.corpus, font_family='SimHei', node_color='white')
plt.show()
def to_json(self):
json = nx.to_dict_of_dicts(self.corpus)
return json
def save_as_json(self, path='./data/corpus.json'):
json = self.to_json()
self.corpus_io.save_as_json(json, path)
def load_from_json(self, path='./data/corpus.json'):
json = self.corpus_io.load_as_json(path)
self.corpus = nx.from_dict_of_dicts(json, create_using=self.corpus)
def get_edge_weight(self, start, end):
weight = 0
try:
weight = self.corpus[start][end]['weight']
except KeyError:
pass
return weight
def get_sorted_neighbour(self, key, exclude=None, K=6):
if key not in self.corpus.adj:
return []
nbr = self.corpus.adj[key]
rs = []
# print(nbr)
# ###########只需要获得前K个最大值，这里的排序可以优化####################
sorted_nbr = sorted(nbr.items(),
key=lambda item: item[1]['weight'],
reverse=True)
j = 0
for i in range(K - 1):
if sorted_nbr[j][0] == exclude:
j += 1
rs.append((sorted_nbr[j][0], sorted_nbr[j][1]['weight']))
j += 1
remain_cnt = 0
remain_weight = 0
for i in range(K - 1, len(sorted_nbr)):
if sorted_nbr[i][0] == exclude:
continue
remain_cnt += 1
remain_weight += sorted_nbr[i][1]['weight']
rs.append(("+" + str(remain_cnt), remain_weight))
return rs