基于CRF的中文地址要素解析

import jieba
import sklearn_crfsuite
from sklearn_crfsuite import metrics

def parse_address(address):
    # 分词
    words = jieba.cut(address)
    # 初始化地址要素
    province = ''
    city = ''
    district = ''
    street = ''
    houseno = ''
    cellno = ''
    floorno = ''
    devzone = ''
    community = ''
    
    # 初始化标签
    tag_list = []
    # 遍历分词结果
    for word in words:
        # 判断是否为省份
        if word.endswith('省') or word.endswith('自治区'):
            for i, char in enumerate(word):
                if i == 0:
                    tag_list.append('B-prov')
                elif i == len(word) - 1:
                    tag_list.append('E-prov')
                else:
                    tag_list.append('I-prov')
            province = word
        # 判断是否为直辖市
        elif word in ['北京', '上海', '天津', '重庆']:
            for i, char in enumerate(word):
                if i == 0:
                    tag_list.append('B-city')
                elif i == len(word) - 1:
                    tag_list.append('E-city')
                else:
                    tag_list.append('I-city')
            province = word + '市'
            city = word + '市'
        # 判断是否为城市
        elif word.endswith('市'):
            for i, char in enumerate(word):
                if i == 0:
                    tag_list.append('B-city')
                elif i == len(word) - 1:
                    tag_list.append('E-city')
                else:
                    tag_list.append('I-city')
            city = word
        # 判断是否为区县
        elif word.endswith('区') or word.endswith('社区')or word.endswith('庄')or word.endswith('塘')or word.endswith('演'):
            for i, char in enumerate(word):
                if i == 0:
                    tag_list.append('B-district')
                elif i == len(word) - 1:
                    tag_list.append('E-district')
                else:
                    tag_list.append('I-district')
            community = word
        # 判断是否为村
        elif word.endswith('村') or word.endswith('县'):
            for i, char in enumerate(word):
                if i == 0:
                    tag_list.append('B-district')
                elif i == len(word) - 1:
                    tag_list.append('E-district')
                else:
                    tag_list.append('I-district')
            district = word
        # 判断是否为城
        elif word.endswith('城'):
            for i, char in enumerate(word):
                if i == 0:
                    tag_list.append('B-devzone')
                elif i == len(word) - 1:
                    tag_list.append('E-devzone')
                else:
                    tag_list.append('I-devzone')
            devzone = word
        # 判断是否为街道
        elif word.endswith('街') or word.endswith('路') or word.endswith('巷') or word.endswith('道'):
            for i, char in enumerate(word):
                if i == 0:
                    tag_list.append('B-town')
                elif i == len(word) - 1:
                    tag_list.append('E-town')
                else:
                    tag_list.append('I-town')
            street = word
        # 判断是否为门牌号
        elif word.isdigit():
            for i, char in enumerate(word):
                if i == 0:
                    tag_list.append('B-roadno')
                elif i == len(word) - 1:
                    tag_list.append('E-roadno')
                else:
                    tag_list.append('I-roadno')
        # 判断是否为楼号
        elif word.endswith('栋') or word.endswith('幢') or word.endswith('座'):
            for i, char in enumerate(word):
                if i == 0:
                    tag_list.append('B-houseno')
                elif i == len(word) - 1:
                    tag_list.append('E-houseno')
                else:
                    tag_list.append('I-houseno')
                houseno = word
        # 判断是否为单元
        elif word.endswith('单元') :
            for i, char in enumerate(word):
                if i == 0:
                    tag_list.append('B-cellno')
                elif i == len(word) - 1:
                    tag_list.append('E-cellno')
                else:
                    tag_list.append('I-cellno')
                cellno = word
        # 判断是否为楼层
        elif word.endswith('楼') :
            for i, char in enumerate(word):
                if i == 0:
                    tag_list.append('B-floorno')
                elif i == len(word) - 1:
                    tag_list.append('E-floorno')
                else:
                    tag_list.append('I-floorno')
                floorno = word
        # 判断是否为POI
        else:
            for i, char in enumerate(word):
                if i == 0:
                    tag_list.append('B-poi')
                elif i == len(word) - 1:
                    tag_list.append('E-poi')
                else:
                    tag_list.append('I-poi')
    # 返回地址要素和标签
    return province, city, district, street, community, devzone, houseno, cellno, floorno, tag_list

# 读取训练数据
with open('train.conll', 'r', encoding='utf-8') as f:
    # 初始化训练数据
    train_data = []
    # 遍历每一行
    for line in f:
        # 跳过空行
        if line.strip() == '':
            continue
        # 获取字符和标签
        char, tag = line.strip().split(' ')
        train_data.append((char, tag))

# 将训练数据转换为CRF模型所需的格式
X_train = [[{'char': char} for char, tag in sentence] for sentence in train_data]

y_train = [[tag for char, tag in sentence] for sentence in train_data]

# 训练CRF模型
crf = sklearn_crfsuite.CRF(
    algorithm='lbfgs',
    c1=0.1,
    c2=0.1,
    max_iterations=100,
    all_possible_transitions=True
)
crf.fit(X_train, y_train)

# 读取测试数据
with open('1.txt', 'r', encoding='utf-8') as f:
    with open('output.txt', 'w', encoding='utf-8') as f_out:
        # 遍历每一行
        for line in f:
            # 获取地址信息
            id_, info = line.strip().split('
')
            # 对地址信息进行分词
            chars = list(info)
            # 将分词结果转换为CRF模型所需的格式
            X_test = [{'char': char} for char in chars]
            # 预测地址要素标签
            y_pred = crf.predict_single(X_test)
            # 将预测结果写入文件
            f_out.write(id_ + '
' + info + '
' + ' '.join(y_pred) + '
')