157 lines
6.0 KiB
Python
157 lines
6.0 KiB
Python
# -*- coding: UTF-8 -*-
|
||
# !/usr/bin/python
|
||
# @time :2019/4/9 22:49
|
||
# @author :Mo
|
||
# @function :get main_part by stanfordcorenlp
|
||
|
||
|
||
from conf.path_config import stanford_corenlp_full_path
|
||
from stanfordcorenlp import StanfordCoreNLP
|
||
|
||
|
||
# stanford-corenlp-full-2018-10-05需要预先下载，启动较慢
|
||
nlp = StanfordCoreNLP(stanford_corenlp_full_path, lang='zh')
|
||
|
||
|
||
def stanford_parse(sentence):
|
||
tokenize = nlp.word_tokenize(sentence)
|
||
pos_tag = nlp.pos_tag(sentence)
|
||
name_entity = nlp.ner(sentence)
|
||
syntax_tree = nlp.parse(sentence)
|
||
dependence = nlp.dependency_parse(sentence)
|
||
|
||
result_dict = {}
|
||
result_dict['tokenize'] = tokenize
|
||
result_dict['dependence'] = dependence
|
||
result_dict['parse'] = syntax_tree
|
||
return result_dict
|
||
|
||
|
||
def combine_nn(tokenize, dependence, target):
|
||
"""
|
||
合并名词短语等
|
||
:param dependence: dict, enhancedPlusPlusDependencies
|
||
:param target: str, subject or object
|
||
:return: str, nn
|
||
"""
|
||
if not target:
|
||
return target
|
||
else:
|
||
for dependence_one in dependence:
|
||
if target == tokenize[dependence_one[1]-1] if dependence_one[1]!=0 else "root" and dependence_one[0] == "nn":
|
||
target = tokenize[dependence_one[2]-1] + target
|
||
return target
|
||
return target
|
||
|
||
|
||
def get_main_part_by_stanfordcorenlp(text):
|
||
"""
|
||
根据依存句法生成句子
|
||
:param text: str, 输入
|
||
:return: str, result of syn sentence
|
||
"""
|
||
# standcoreNLP 分词
|
||
result_dict = stanford_parse(text)
|
||
tokenize = result_dict['tokenize']
|
||
dependence = result_dict['dependence']
|
||
syntax_tree = result_dict['parse']
|
||
# 提取主谓宾
|
||
part_main = {"主": "", "谓": "", "宾": ""}
|
||
if len(syntax_tree) >= 2:
|
||
if "NP" in syntax_tree[1] or "ROOT" not in str(dependence): # 名词短语 或者是没有谓语
|
||
count = 0
|
||
for syntax_tree_single in syntax_tree:
|
||
if "NP" in syntax_tree_single and "(" in syntax_tree_single and ")" in syntax_tree_single:
|
||
token_np = syntax_tree_single.split(" ")[-1]
|
||
token_np = token_np.replace("'", "").replace(")", "").strip()
|
||
part_main["主"] = token_np if count == 0 else part_main["主"] + token_np
|
||
count += 1
|
||
return part_main["主"] + part_main["谓"] + part_main["宾"]
|
||
else:
|
||
for dependence_one in dependence:
|
||
dep = dependence_one[0]
|
||
dep_dep_gloss = tokenize[dependence_one[2]-1]
|
||
if dep == "ROOT": # ROOT作谓语
|
||
part_main["谓"] = dep_dep_gloss
|
||
elif dep == "cop": # 主系结构
|
||
part_main["谓"] = dep_dep_gloss + part_main["谓"]
|
||
else: # 主语和宾语
|
||
if dep == "nsubjpass" or dep == "dobj" or dep == "attr":
|
||
part_main["宾"] = dep_dep_gloss
|
||
elif dep == "nsubj" or dep == "top":
|
||
part_main["主"] = dep_dep_gloss
|
||
|
||
part_main["主"] = combine_nn(tokenize, dependence, part_main["主"])
|
||
part_main["宾"] = combine_nn(tokenize, dependence, part_main["宾"])
|
||
return part_main["主"] + part_main["谓"] + part_main["宾"]
|
||
|
||
|
||
if __name__ == "__main__":
|
||
sentence_list = [
|
||
"大漠帝国确实很喜欢JY",
|
||
"JY也喜欢大漠帝国哦！",
|
||
"这个工程的作者是momo",
|
||
"momo是一个无门无派的浪人",
|
||
"只有自信的程序员才能把握未来",
|
||
"主干识别可以提高检索系统的智能",
|
||
"打更的住在这里",
|
||
"人民的名义",
|
||
"名词短语",
|
||
"我一直很喜欢你",
|
||
"你被我喜欢",
|
||
"美丽又善良的你被卑微的我深深的喜欢着……",
|
||
|
||
"搜索momo可以找到我的博客",
|
||
"静安区体育局2013年部门决算情况说明",
|
||
"红旗飘",
|
||
"柳丝长",
|
||
"乐队奏国歌",
|
||
"红扑扑的朝霞露出了笑脸",
|
||
"初升的太阳照耀着峻峭的群山",
|
||
"一个农人在路上看见一条冻僵了的蛇",
|
||
"我打量了他一眼", ]
|
||
sentence_type = ["陈述句与否定句",
|
||
"秦耕真是一个聪明的孩子",
|
||
"衣服洗得不干净",
|
||
"他没有做完作业",
|
||
"他不敢不来",
|
||
"没有一个人不怕他",
|
||
"我非把这本书读完不可",
|
||
"同学们无不欢欣鼓舞",
|
||
"他妈妈不让他去,无非是怕他吃亏",
|
||
"想起一个人的旅途,不无寂寥之感",
|
||
"你未必不知道",
|
||
"各种问句",
|
||
"你可以那到100分, 是吗?",
|
||
"刚才接你的人是谁?",
|
||
"什么叫函数?",
|
||
"你爸爸怎么样了?",
|
||
"你每天几点休息?",
|
||
"你爸爸在哪儿?",
|
||
"我们是从广州走, 还是从成都走?",
|
||
"他是不是又迟到了?",
|
||
"难道他已经跑了?",
|
||
"我怎么能负这个责任呢?",
|
||
"你是来帮助我们的, 还是来拆我们的台的?",
|
||
"这些人甘愿当走狗, 你说可恨不可恨?",
|
||
"祈使句",
|
||
"快去捞饭!米烂了!",
|
||
"给我喝水, 我渴!",
|
||
"走哇, 妈妈!",
|
||
"不许动!",
|
||
"太好啦",
|
||
]
|
||
for sen_one in sentence_list:
|
||
subject_object = get_main_part_by_stanfordcorenlp(sen_one)
|
||
print(sen_one + " " + subject_object)
|
||
|
||
while True:
|
||
print("请输入sentence： ")
|
||
sen_test = input()
|
||
# syn_sentence_test = syn_by_syntactic_analys==(test_test)
|
||
syn_sentence_test = get_main_part_by_stanfordcorenlp(sen_test)
|
||
print(syn_sentence_test)
|
||
|
||
# Do not forget to close! The backend server will consume a lot memery
|
||
nlp.close()
|