def extract_common_features(item): #提取共同特征
title_features = transform_feature(
string_utils.clean_sentence(item["title"], stemming=True).lower(),
"title") #将title项提取特征
keywords_features = []
keywords = item.get("keywords") #提取 keywords 项
if keywords:
keywords_features = transform_feature(
[string_utils.clean_name(k) for k in keywords],
'keyword') #提取keyword中的单词， 去除连接符
fields_features = []
fields = item.get('fields')
if fields:
fields_features = transform_feature(
[string_utils.clean_name(k) for k in fields], 'fields')
venue_features = [] #提取机构venue项
venue_name = item.get('venue', '')
if venue_name:
if len(venue_name) > 2:
venue_features = transform_feature(
string_utils.clean_sentence(venue_name.lower()),
"venue") #小写化 去除分界符 变换特征
abst_features = []
abst = item.get('abst')
if abst:
abst_features = transform_feature(
string_utils.clean_sentence(abst.lower()), "abst")
return title_features, keywords_features, venue_features, abst_features, fields_features
def extract_common_features(item):
title_features = string_utils.clean_sentence(item["title"], stemming=True)
keywords_features = []
keywords = item.get("keywords")
if keywords:
for k in keywords:
keywords_features.extend(
string_utils.clean_sentence(k, stemming=True))
venue_features = []
venue_name = item.get('venue', '')
if len(venue_name) > 2:
venue_features = string_utils.clean_sentence(venue_name.lower(),
stemming=True)
return title_features, keywords_features, venue_features
def extract_common_features(item):
title_features = transform_feature(
string_utils.clean_sentence(item["title"], stemming=True).lower(),
"title")
keywords_features = []
keywords = item.get("keywords")
if keywords:
keywords_features = transform_feature(
[string_utils.clean_name(k) for k in keywords], 'keyword')
venue_features = []
venue_name = item.get('venue', '')
if len(venue_name) > 2:
venue_features = transform_feature(
string_utils.clean_sentence(venue_name.lower()), "venue")
return title_features, keywords_features, venue_features
def extract_author_features(item, order=None):
title_features, keywords_features, venue_features = extract_common_features(
item)
author_features = []
for i, author in enumerate(item["authors"]):
if order is not None and i != order:
continue
name_feature = []
org_features = []
org_name = string_utils.clean_name(author.get("org", ""))
if len(org_name) > 2:
org_features.extend(transform_feature(org_name, "org"))
for j, coauthor in enumerate(item["authors"]):
if i == j:
continue
coauthor_name = coauthor.get("name", "")
coauthor_org = string_utils.clean_name(coauthor.get("org", ""))
if len(coauthor_name) > 2:
name_feature.extend(
transform_feature([string_utils.clean_name(coauthor_name)],
"name"))
if len(coauthor_org) > 2:
org_features.extend(
transform_feature(
string_utils.clean_sentence(coauthor_org.lower()),
"org"))
author_features.append(name_feature + org_features + title_features +
keywords_features + venue_features)
author_features = list(chain.from_iterable(author_features))
return author_features
def extract_author_features(item):
title_features, keywords_features, venue_features = extract_common_features(
item)
name_features = []
org_features = []
for i, author in enumerate(item['authors']):
name = string_utils.clean_name(author.get('name', ''))
name = transform_feature(name, 'name')
org = string_utils.clean_sentence(author.get('org', ''))
org = org + ' ' + org.replace(' ', '_')
org = transform_feature(org, 'org')
if name:
name_features.extend(name)
if org:
org_features.extend(org)
title_features = ' '.join(title_features) + ' '
keywords_features = ' '.join(keywords_features) + ' '
venue_features = ' '.join(venue_features) + ' '
name_features = ' '.join(name_features) + ' '
org_features = ' '.join(org_features) + ' '
author_features = name_features + org_features + title_features + keywords_features + venue_features
return author_features
def extract_author_features(item, order=None):
title_features, keywords_features, venue_features = extract_common_features(
item)
word_features = title_features + keywords_features + venue_features
author_features = []
for i, author in enumerate(item["authors"]):
if order is not None and i != order:
continue
org_name = string_utils.clean_sentence(author.get("org", ""),
stemming=True)
if len(org_name) > 2:
word_features += org_name
for j, coauthor in enumerate(item["authors"]):
if i == j:
continue
coauthor_name = coauthor.get("name", "")
if (coauthor_name == None):
continue
if (len(coauthor_name.strip()) > 0):
if len(coauthor_name.strip()) > 2:
author_features.append(
string_utils.clean_name(coauthor_name))
else:
author_features.append(coauthor_name.lower())
return author_features, word_features
def pre_data(author_dict,pub_dict): # 预处理字符串数据，去掉字符串中的停用词，将名字统一成小写
author_new = dict() # 对作者名字进行预处理过的数据
for name in author_dict:
new_name = string_utils.clean_name_author(name)
author_new[new_name] = author_dict[name]
#
# # 把pub中的字符串都进行预处理
for id in pub_dict:
for item in pub_string_item:
pub_dict[id][item] = string_utils.clean_sentence(pub_dict[id][item])
authors=[]
for author in pub_dict[id]["authors"]:
authors.append(
{'name':string_utils.clean_name(author['name']),
'org':string_utils.clean_sentence(author['org'])}
)
pub_dict[id]["authors"] =authors
dump_json(author_new, wfpath='../data/', wfname='train_author_new.json', indent=4)
dump_json(pub_dict, wfpath='../data/', wfname='train_pub_new.json', indent=4)
return author_new,pub_dict
def extract_author_features(item, order=None): #提取作者特征 item中的order
title_features, keywords_features, venue_features, abst_features, fields_features = extract_common_features(
item) #提取共同特征 标题， 关键字， 收录机构
author_features = []
for i, author in enumerate(item["authors"]): #枚举第i个作者， author
if order is not None and i != order: #找到所要的 第order个作者
continue
name_feature = [] #姓名特征
org_features = [] #机构特征
org_name = string_utils.clean_name(author.get(
"org", "")) #格式化机构名 按".", "-", " "分割 小写化
if len(org_name) > 2:
org_features.extend(transform_feature(org_name, "org")) #列表加列表
for j, coauthor in enumerate(item["authors"]): #枚举 合作者
if i == j:
continue
coauthor_name = coauthor.get("name", "") #获得名字
coauthor_org = string_utils.clean_name(coauthor.get("org",
"")) #获得格式化机构名
if len(coauthor_name) > 2:
name_feature.extend(
transform_feature([string_utils.clean_name(coauthor_name)],
"name") #格式化 与 特征变换
) #将合作者名特征加入 名字特征中
if len(coauthor_org) > 2:
org_features.extend(
transform_feature(
string_utils.clean_sentence(coauthor_org.lower()),
"org") #格式化 与 特征变换
) #将合作者机构特征加入 机构特征中
author_features.append(name_feature + org_features + title_features +
keywords_features + venue_features +
abst_features +
fields_features) #将以上特征 都 整合 到 作者特征 中
author_features = list(chain.from_iterable(author_features)) #创建 为 迭代器 列表
return author_features # 到这里 就是把 各个特征 对应的单词列表 合并到一个列表里了
