def __init__(self, parent, mode=None, entity=None):
'''
:param parent:
:param mode: 添加实体 修改实体
'''
self.entity = entity
super().__init__()
self.top = Toplevel(parent)
# top.attributes('-alpha',0.9) # fun attribute "transparent"
if mode == 'add':
self.top.title("添加实体")
if mode == 'alter':
self.top.title("修改实体")
utils.setCenteredDisply(self.top, 400, 300)
self.ft_1Title = tkFont.Font(family="宋体", size=11, weight=tkFont.BOLD)
self.ft_2Title = tkFont.Font(family="宋体", size=9, weight=tkFont.BOLD)
# l_class 从配置文件中读取
self.conf = Config()
ent_dict = self.conf.getConf('ENTITY')
self.l_class = [ent for ent in ent_dict.keys()]
self.v_class = StringVar()
self.v_class.set(self.l_class)
self.status = IntVar() # 标记复选框是否选中
self.checkEntity = "certain" # 是否对实体确认
self.entity_category = ''
self.ui(self.top)
def val(embed_dim, lstm_dim, optimizer, lr, gpu, gpu_no, text):
if gpu == 'ON':
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
sess = tf.Session(config=config)
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID" # see issue #152
os.environ['CUDA_VISIBLE_DEVICES'] = gpu_no
with open('./model/weights/config.pkl', 'rb') as input_:
(word_vocab, tags) = pickle.load(input_)
TAGS = {}
idx = 0
for i in tags:
TAGS[idx] = i
idx += 1
conf = Config()
ent_dict = conf.getConf('ENTITY')
ent_list = [w for w in ent_dict.keys()]
B_I_dict = {}
for w in ent_list:
B_I_dict['B-' + w.capitalize()] = 'I-' + w.capitalize()
num_classes = len(tags)
num_steps = utils.maxlen
word_dict_size = len(word_vocab)
model = bilstm_crf_model.BiLSTM_CRF_Model(num_classes, num_steps,
word_dict_size, embed_dim,
lstm_dim, optimizer, lr).model()
predict(model, TAGS, word_vocab, text)
get_label_text(B_I_dict)
def __init__(self, parent):
super().__init__()
self.top = Toplevel(parent)
self.parent = parent
self.top.title('Label&ShortcutKey Setting')
utils.setCenteredDisply(self.top, 500, 400)
self.row = 10
self.column = 20
# 可以通过tuple(list) 将列表转为tuple
self.conf = Config()
self.entdict = self.conf.getConf('ENTITY')
if type(self.entdict) == str:
self.entdict = {}
self.ui(self.top)
def sckSet(self):
d = ConfigureEntRel_Dialog(self)
self.wait_window(d.top)
conf = Config()
self.ent_dict = conf.getConf('ENTITY')
for key in self.ent_dict.keys():
self.text.bind("<Control-Alt-" + self.ent_dict[key] + ">",
self.runSckAdaptor(self.runSCK, 'uncertain', key))
self.text.bind("<Control-" + self.ent_dict[key] + ">",
self.runSckAdaptor(self.runSCK, 'certain', key))
self.table.bind(
"<Control-Alt-" + self.ent_dict[key] + ">",
self.runSckAdaptor(self.tableAlter, 'uncertain', key))
self.table.bind(
"<Control-" + self.ent_dict[key] + ">",
self.runSckAdaptor(self.tableAlter, 'certain', key))
#print(self.ent_dict)
self.loadShortKey() # 更新快捷键界面
def load_data():
train, train_word = _parse_data(open('./model/train_all_tag.txt', 'rb'))
word_counts = Counter(row[0].lower() for sample in train_word
for row in sample) #字符 .lower()
word_vocab = [w for w, f in iter(word_counts.items())] # 字典大小
# get the tag
conf = Config()
ent_dict = conf.getConf('ENTITY')
ent_list = [w for w in ent_dict.keys()]
tags = []
tags.append('O')
for w in ent_list:
tags.append('B-' + w.capitalize())
tags.append('I-' + w.capitalize())
# save initial config data 存储配置信息
with open('./model/weights/config.pkl', 'wb') as output:
pickle.dump((word_vocab, tags), output)
train_word, train_y = _process_data(train_word, word_vocab, tags, maxlen)
return train_word, train_y, word_vocab, tags
def __init__(self, parent_frame):
'''
初始化系统参数
:param parent: frame widget "parent=window"
'''
Frame.__init__(self, parent_frame)
self.Version = "EPAD V1.0"
self.parent_frame = parent_frame
self.currentPath = os.getcwd()
self.fileName = ""
self.textColumn = 7
self.fntSize = 13
self.allKey = "1234567890abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ"
self.pressCommand = {'a': "a1", 'b': "a2", 'c': "a3"}
self.textFontSytle = "Times"
self.textRow = 20
self.no = 0 # 记录实体的序号，并且在treeview中记录插入位置
conf = Config()
self.recommendFlag1 = "ON" # 最大前向匹配
self.recommendFlag2 = "OFF" # 深度学习
self.recoEntity = '' # 设置当前要推荐的词语
self.entity_list = [] # 实体集合
self.ent_dict = conf.getConf('ENTITY') # 获得实体类别
self.pre_ent_list = []
if type(self.ent_dict) == str:
self.ent_dict = {}
else:
self.ent_list = [ent
for ent in self.ent_dict.keys()] # 这个是类别列表
self.selectColor = 'light salmon' # 选中的颜色
self.tagColor = 'lightgreen' # 标记的颜色
self.remColor = 'pink' # 推荐的颜色
self.preColor = 'purple' # 预标注的颜色
self.onlylabel = ''
self.onlyhow = ''
self.initUI()
def __init__(self, parent_frame, dir_name, file_path):
'''
初始化系统参数
:param parent: frame widget "parent=window"
'''
super().__init__()
self.file_path = file_path # 表示要修改的标注文件
self.dir_name = dir_name # 表示要修改的标注者
self.top = Toplevel(parent_frame)
utils.setCenteredDisply(self.top, 1000, 700)
self.Version = "EPAD-V1.0 Modify"
self.currentPath = os.getcwd()
self.fileName = ""
self.textColumn = 7
self.fntSize = 13
self.allKey = "1234567890abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ"
self.pressCommand = {'a': "a1", 'b': "a2", 'c': "a3"}
self.textFontSytle = "Times"
self.textRow = 20
self.no = 0 # 记录实体的序号，并且在treeview中记录插入位置
self.conf = Config()
self.recommendFlag1 = "ON" # 最大前向匹配
self.recommendFlag2 = "OFF" # 深度学习
self.recoEntity = '' # 设置当前要推荐的词语
self.entity_list = [] # entity list # 实体类别
self.ent_dict = self.conf.getConf('ENTITY') # 获得实体类别
self.ent_list = [ent for ent in self.ent_dict.keys()
] # !==[{'entity':,'pos':,},{},{},{}] 这个是类别列表
self.selectColor = 'light salmon' # 选中的颜色
self.tagColor = 'lightgreen' # 标记的颜色
self.remColor = 'pink' # 推荐的颜色
self.preColor = 'purple' # 预标注的颜色
self.onlylabel = ''
self.onlyhow = ''
self.initUI(self.top)
import os
from model.model_utils import *
from configparse import Config
train_new_path='./model/train_data/'
# label_dict need be modified , because the entity is Chinese.
conf = Config()
label_dict = conf.getConf('ENTITY_DESC')
# label_dict = {'身体部位':'Body','症状和体征':'Symptom','检查和检验':'Test','治疗':'Treatment','疾病和诊断':'Disease'}
def BIOtag(data_path):
document_file = os.listdir(data_path)
tag_all_file = open('./model/train_all_tag.txt', 'w', encoding='utf-8')
document_file = [file for file in document_file if 'txt'in file]
for file in document_file:
ori_file = open(data_path + file, 'r', encoding='UTF-8') # ori text
label_file= open(data_path + file[:-3]+'ann', 'r', encoding='UTF-8') # label text
ori_data2=ori_file.readline()
label_list = ['O'] * len(ori_data2)
entity_list = label_file.read().split('\n')
for entity in entity_list:
if entity is not '':
ent_index = entity.split('\t')
ent_s = int(ent_index[1])
ent_e = int(ent_index[2])+1
ent_type=ent_index[3]
###——————————BIOS打标————————————
# if ent_e-ent_s>1: # 多字符实体
# label_list[ent_s]='B-'+ent_type
# for i in range(ent_s+1,ent_e):