FineTurnX/qwen_lora/compare_result.py

import json
import re
import traceback
from glob import glob

import pandas as pd
from bs4 import BeautifulSoup
from json_repair import repair_json, json_repair
from get_data import psv_to_dict, psv_to_dict_prefix
from compare_utils import compare_products


def compare_extract_csv():
    df_tenderee = pd.read_csv(r'D:\BIDI_DOC\比地_文档\export_ai_260323_2_tenderee.csv')
    df1 = pd.read_excel(r'D:\BIDI_DOC\比地_文档\export_ai_260323_1.xlsx')
    df2 = pd.read_excel(r'D:\BIDI_DOC\比地_文档\export_ai_260323_3.xlsx')
    # df3 = pd.read_excel(r'D:\BIDI_DOC\比地_文档\export_ai_260323_3_lora_2.xlsx')
    # df3 = pd.read_excel(r'D:\BIDI_DOC\比地_文档\export_ai_260331_lora.xlsx')
    # df3 = pd.read_excel(r'D:\BIDI_DOC\比地_文档\export_ai_260401_lora-r64.xlsx')
    # df3 = pd.read_excel(r'D:\BIDI_DOC\比地_文档\export_ai_260401_lora_2B.xlsx')
    # df3 = pd.read_excel(r'D:\BIDI_DOC\比地_文档\export_ai_260402_lora_2B.xlsx')
    # df3 = pd.read_excel(r'D:\BIDI_DOC\比地_文档\export_ai_260407_0.8B.xlsx')
    # df3 = pd.read_excel(r'D:\BIDI_DOC\比地_文档\export_ai_260408_0.8B.xlsx')
    # df3 = pd.read_excel(r'D:\BIDI_DOC\比地_文档\export_ai_260409_0.8B.xlsx')
    # df3 = pd.read_excel(r'D:\BIDI_DOC\比地_文档\export_ai_260413_0.8B.xlsx')
    df3 = pd.read_excel(r'D:\BIDI_DOC\比地_文档\export_ai_260423_0.8B_prefix_all.xlsx')

    tenderee_data_list = df_tenderee.astype(object).where(pd.notnull(df_tenderee), None).values.tolist()
    data_list1 = df1.astype(object).where(pd.notnull(df1), None).values.tolist()
    data_list2 = df2.astype(object).where(pd.notnull(df2), None).values.tolist()
    data_list3 = df3.astype(object).where(pd.notnull(df3), None).values.tolist()

    tenderee_data_dict = {int(x[0]): x[1] for x in tenderee_data_list}
    data_dict1 = {int(x[0]): x[1] for x in data_list1}
    data_dict2 = {int(x[0]): x[1] for x in data_list2}
    data_dict3 = {int(x[0]): psv_to_dict(x[1]) for x in data_list3}

    for docid, v_dict in data_dict3.items():
        v1 = v_dict.get('招标信息')
        v2 = v_dict.get('招标人联系方式')
        if v1 is None or v2 is None:
            continue
        v1 = v1[0]
        if not v1:
            continue
        v1['招标人联系方式'] = v2
        v_dict['招标信息'] = v1
        data_dict3[docid] = v_dict

    for docid, v_dict in data_dict3.items():
        data_dict3[docid] = json.dumps(v_dict, ensure_ascii=False)

    # print('data_dict3', data_dict3.get(751300501))

    cols = ['招标信息', '中标信息', '产品信息']

    result_list = []
    tenderee_match_cnt = 0
    tenderee_match_empty_cnt = 0
    real_tenderee_match_cnt1 = 0
    real_tenderee_match_cnt2 = 0
    win_tenderer_match_cnt = 0
    agency_match_cnt = 0
    all_product_weight_score = 0
    all_product_complete_score = 0
    all_product_cnt_score = 0
    for docid, data1 in data_dict1.items():
        data2 = data_dict2.get(docid, '{}')
        data3 = data_dict3.get(docid, '{}')

        # if len(re.findall('}]}', data2)) != 2:
        #     data2 = re.sub('}]', '}]}', data2, count=1)
        # if len(re.findall('}]}', data3)) != 2:
        #     data3 = re.sub('}]', '}]}', data3, count=1)

        # if len(data1) >= 1000:
        #     data1 = '{}'
        # if len(data2) >= 1000:
        #     data2 = '{}'
        # if len(data3) >= 1000:
        #     data3 = '{}'

        # if len(data1) >= 1000:
        #     extract_json1 = '{}'
        # else:
        extract_json1 = repair_json(data1)
        if len(data2) >= 1000:
            extract_json2 = '{}'
        else:
            extract_json2 = repair_json(data2)
        # if len(data3) >= 1000:
        #     extract_json3 = '{}'
        # else:
        #     extract_json3 = repair_json(data3)
        # print('data2', data2)
        extract_json3 = repair_json(data3)

        extract_json1 = json_repair.loads(extract_json1)
        extract_json2 = json_repair.loads(extract_json2)
        extract_json3 = json_repair.loads(extract_json3)
        # print('extract_json2', extract_json2)

        if type(extract_json1) != dict:
            extract_json1 = {}
        if type(extract_json2) != dict:
            extract_json2 = {}
        if type(extract_json3) != dict:
            extract_json3 = {}

        col_match_list = []
        for col in cols:
            str1 = json.dumps(extract_json1.get(col, ""), ensure_ascii=False)
            str2 = json.dumps(extract_json2.get(col, ""), ensure_ascii=False)
            str3 = json.dumps(extract_json3.get(col, ""), ensure_ascii=False)

            if str1 != str3:
                col_match_list += [str1, str2, str3, 0]
            else:
                col_match_list += [str1, str2, str3, 1]

        tenderee_real = tenderee_data_dict.get(docid)
        tenderee1 = extract_json1.get('招标信息', {}).get('招标人名称', "")
        try:
            tenderee3 = extract_json3.get('招标信息', {}).get('招标人名称', "")
        except:
            print('no 招标信息 招标人名称 extract_json3', extract_json3)
            # raise
            tenderee3 = ''
        if re.sub(' ', '', tenderee1) == re.sub(' ', '', tenderee3):
            tenderee_match_cnt += 1
            if not tenderee1:
                tenderee_match_empty_cnt += 1

        if tenderee1 == tenderee_real:
            real_tenderee_match_cnt1 += 1
        if tenderee3 == tenderee_real:
            real_tenderee_match_cnt2 += 1

        win_tenderer1 = [x.get('中标人名称', "") for x in extract_json1.get('中标信息', [])]
        win_tenderer1 = list(set(win_tenderer1))
        win_tenderer1.sort(key=lambda x: x)
        win_tenderer1 = ','.join(win_tenderer1)
        try:
            win_tenderer3 = [x.get('中标人名称', "") for x in extract_json3.get('中标信息', [])]
            win_tenderer3 = list(set(win_tenderer3))
            win_tenderer3.sort(key=lambda x: x)
            win_tenderer3 = ','.join(win_tenderer3)
        except:
            print('no 中标信息 中标人名称 extract_json3', extract_json3)
            # raise
            win_tenderer3 = ''

        if win_tenderer1 == win_tenderer3:
            print('win_tenderer1 == win_tenderer3', win_tenderer1, win_tenderer3)
            win_tenderer_match_cnt += 1

        agency1 = extract_json1.get('招标信息', {}).get('代理人名称', "")
        try:
            agency3 = extract_json3.get('招标信息', {}).get('代理人名称', "")
        except:
            print('no 招标信息 中标人名称 extract_json3', extract_json3)
            # raise
            agency3 = ''

        if agency1 == agency3:
            agency_match_cnt += 1

        # 计算products匹配率
        products1 = extract_json1.get('产品信息')
        products3 = extract_json3.get('产品信息')
        # print('products3', products3)
        product_weight_score, product_complete_score, product_cnt_score = compare_products(products1, products3)
        all_product_weight_score += product_weight_score
        all_product_complete_score += product_complete_score
        all_product_cnt_score += product_cnt_score

        result_list.append(
            [docid, data1, data2, data3] +
            col_match_list +
            [tenderee1, tenderee3, tenderee_real] +
            [str(tenderee1==tenderee3), str(tenderee1==tenderee_real), str(tenderee3==tenderee_real)] +
            [win_tenderer1, win_tenderer3, str(win_tenderer1==win_tenderer3)] +
            [agency1, agency3, str(agency1==agency3)]
        )

    df_new = pd.DataFrame(result_list)
    df_new.columns = ['docid',
                      'doubao', 'qwen', 'qwen-lora',
                      '招标信息1', '招标信息2', '招标信息3', '招标信息相同',
                      '中标信息1', '中标信息2', '中标信息3', '中标信息相同',
                      '产品信息1', '产品信息2', '产品信息3', '产品信息相同',
                      'tenderee_doubao', 'tenderee_qwen', '线上_tenderee',
                      'doubao_qwen_相同', 'doubao_线上_相同', 'qwen_线上相同',
                      'win_doubao', 'win_qwen', 'doubao_qwen_相同',
                      'agency_doubao', 'agency_qwen', 'doubao_qwen_相同',
                      ]

    df_new.to_excel(r'D:\BIDI_DOC\比地_文档\export_ai_260423_compare_doubao_0.8B_prefix.xlsx', index=False)
    all_cnt = len(data_dict1)
    print('tenderee_match_cnt', tenderee_match_cnt, tenderee_match_cnt/all_cnt, 'tenderee_match_empty_cnt', tenderee_match_empty_cnt)
    print('real_tenderee_match_cnt1', real_tenderee_match_cnt1, real_tenderee_match_cnt1/all_cnt)
    print('real_tenderee_match_cnt2', real_tenderee_match_cnt2, real_tenderee_match_cnt2/all_cnt)
    print('win_tenderer_match_cnt', win_tenderer_match_cnt, win_tenderer_match_cnt/all_cnt)
    print('agency_match_cnt', agency_match_cnt, agency_match_cnt/all_cnt)
    print('product_weight_score', all_product_weight_score / all_cnt)
    print('product_complete_score', all_product_complete_score / all_cnt)
    print('product_cnt_score', all_product_cnt_score / all_cnt)


def compare_extract_csv_prefix():
    # df_tenderee = pd.read_csv(r'D:\BIDI_DOC\比地_文档\export_ai_260323_2_tenderee.csv')

    df_tenderee1 = pd.read_excel(r'D:\BIDI_DOC\比地_文档\招标人标注_260414.xlsx', sheet_name=0)
    df_tenderee2 = pd.read_excel(r'D:\BIDI_DOC\比地_文档\招标人标注_260414.xlsx', sheet_name=1)
    df_tenderee3 = pd.read_excel(r'D:\BIDI_DOC\比地_文档\招标人标注_260414.xlsx', sheet_name=2)
    df_tenderee = pd.concat([df_tenderee1, df_tenderee2, df_tenderee3], ignore_index=True)

    df1 = pd.read_excel(r'D:\BIDI_DOC\比地_文档\export_ai_260323_1.xlsx')
    df2 = pd.read_excel(r'D:\BIDI_DOC\比地_文档\export_ai_260323_3.xlsx')
    # df3 = pd.read_excel(r'D:\BIDI_DOC\比地_文档\export_ai_260413_0.8B_prefix_only_tenderee.xlsx')
    df3 = pd.read_excel(r'D:\BIDI_DOC\比地_文档\export_ai_260413_0.8B_prefix_only_tenderee_all.xlsx')

    tenderee_data_list = df_tenderee.astype(object).where(pd.notnull(df_tenderee), None).values.tolist()
    data_list1 = df1.astype(object).where(pd.notnull(df1), None).values.tolist()
    data_list2 = df2.astype(object).where(pd.notnull(df2), None).values.tolist()
    data_list3 = df3.astype(object).where(pd.notnull(df3), None).values.tolist()

    tenderee_data_dict = {int(x[0]): x[1] for x in tenderee_data_list}

    data_dict1 = {int(x[0]): x[1] for x in data_list1}
    data_dict2 = {int(x[0]): x[1] for x in data_list2}
    data_dict3 = {int(x[0]): psv_to_dict_prefix(x[1]) for x in data_list3}

    for docid, v_dict in data_dict3.items():
        data_dict3[docid] = json.dumps(v_dict, ensure_ascii=False)

    cols = ['招标信息', '中标信息', '产品信息']

    result_list = []
    tenderee_match_cnt = 0
    tenderee_match_empty_cnt = 0
    real_tenderee_match_cnt1 = 0
    real_tenderee_match_cnt2 = 0
    for docid, tenderee_data in tenderee_data_dict.items():
        data1 = data_dict1.get(docid, '{}')
        data2 = data_dict2.get(docid, '{}')
        data3 = data_dict3.get(docid, '{}')

        extract_json1 = repair_json(data1)
        if len(data2) >= 1000:
            extract_json2 = '{}'
        else:
            extract_json2 = repair_json(data2)
        extract_json3 = repair_json(data3)

        extract_json1 = json_repair.loads(extract_json1)
        extract_json2 = json_repair.loads(extract_json2)
        extract_json3 = json_repair.loads(extract_json3)

        if type(extract_json1) != dict:
            extract_json1 = {}
        if type(extract_json2) != dict:
            extract_json2 = {}
        if type(extract_json3) != dict:
            extract_json3 = {}

        col_match_list = []
        for col in cols:
            str1 = json.dumps(extract_json1.get(col, ""), ensure_ascii=False)
            str2 = json.dumps(extract_json2.get(col, ""), ensure_ascii=False)
            str3 = json.dumps(extract_json3.get(col, ""), ensure_ascii=False)

            if str1 != str3:
                col_match_list += [str1, str2, str3, 0]
            else:
                col_match_list += [str1, str2, str3, 1]

        tenderee_real = tenderee_data_dict.get(docid)
        if tenderee_real in ['文中无招标人', '-']:
            tenderee_real = ''
        tenderee1 = extract_json1.get('招标信息', {}).get('招标人名称', "")
        try:
            tenderee3 = extract_json3.get('招标信息', {}).get('招标人名称', "")
        except:
            print('no 招标信息 招标人名称 extract_json3', extract_json3)
            # raise
            tenderee3 = ''
        if tenderee1 in ['文中无招标人', '-']:
            tenderee1 = ''
        if tenderee3 in ['文中无招标人', '-']:
            tenderee3 = ''

        if re.sub(' ', '', tenderee1) == re.sub(' ', '', tenderee3):
            tenderee_match_cnt += 1
            if not tenderee1:
                tenderee_match_empty_cnt += 1

        if tenderee1 == tenderee_real:
            real_tenderee_match_cnt1 += 1
        if tenderee3 == tenderee_real:
            real_tenderee_match_cnt2 += 1

        result_list.append(
            [docid, data1, data2, data3] +
            col_match_list +
            [tenderee1, tenderee3, tenderee_real] +
            [str(tenderee1==tenderee3), str(tenderee1==tenderee_real), str(tenderee3==tenderee_real)]
        )

    df_new = pd.DataFrame(result_list)
    df_new.columns = ['docid', 'doubao', 'qwen', 'qwen-lora',
                      '招标信息1', '招标信息2', '招标信息3', '招标信息相同',
                      '中标信息1', '中标信息2', '中标信息3', '中标信息相同',
                      '产品信息1', '产品信息2', '产品信息3', '产品信息相同',
                      'tenderee_doubao', 'tenderee_qwen', '线上_tenderee',
                      'doubao_qwen_相同', 'doubao_线上_相同', 'qwen_线上相同',
                      ]

    df_new.to_excel(r'D:\BIDI_DOC\比地_文档\export_ai_260414_compare_only_tenderee_labeled.xlsx', index=False)
    print('tenderee_match_cnt', tenderee_match_cnt, 'tenderee_match_empty_cnt', tenderee_match_empty_cnt)
    print('real_tenderee_match_cnt1', real_tenderee_match_cnt1, real_tenderee_match_cnt1/len(tenderee_data_dict))
    print('real_tenderee_match_cnt2', real_tenderee_match_cnt2, real_tenderee_match_cnt2/len(tenderee_data_dict))
    print('len(tenderee_data_dict)', len(tenderee_data_dict))


def compare_extract_csv_prefix2():
    file_path1 = r'C:\Users\Administrator\Downloads\招标人_中标人_标注_260421 (1).xlsx'
    file_path2 = r'D:\BIDI_DOC\比地_文档\export_ai_260506_doubao.xlsx'
    file_path3 = r'D:\BIDI_DOC\比地_文档\export_ai_260506_0.8B_prefix_all.xlsx'
    file_path4 = r'D:\BIDI_DOC\比地_文档\招标人_中标人_标注_260421_tenderee_win_2.csv'

    data_list1 = []
    for sheet_no in range(0, 7):
        df1 = pd.read_excel(file_path1, sheet_name=sheet_no)
        data_list1 += df1.astype(object).where(pd.notnull(df1), None).values.tolist()

    print('len(data_list1)', len(data_list1), data_list1[0])

    df2 = pd.read_excel(file_path2)
    df3 = pd.read_excel(file_path3)
    df4 = pd.read_csv(file_path4)

    data_list2 = df2.astype(object).where(pd.notnull(df2), None).values.tolist()
    data_list3 = df3.astype(object).where(pd.notnull(df3), None).values.tolist()
    data_list4 = df4.astype(object).where(pd.notnull(df4), None).values.tolist()

    data_dict_label = {int(x[0]): {'招标信息': {'招标人名称': x[3]}, '中标信息': [{'中标人名称': x[4]}]} for x in data_list1}
    data_dict_doubao = {int(x[0]): x[2] for x in data_list2}
    data_dict_qwen = {int(x[0]): psv_to_dict(x[1]) for x in data_list3}
    data_dict_online = {int(x[0]): {'招标信息': {'招标人名称': x[1]}, '中标信息': [{'中标人名称': x[2]}]} for x in data_list4}

    for docid, v_dict in data_dict_qwen.items():
        data_dict_qwen[docid] = json.dumps(v_dict, ensure_ascii=False)
    for docid, v_dict in data_dict_label.items():
        data_dict_label[docid] = json.dumps(v_dict, ensure_ascii=False)
    for docid, v_dict in data_dict_online.items():
        data_dict_online[docid] = json.dumps(v_dict, ensure_ascii=False)

    cols = ['招标信息', '中标信息', '产品信息']

    result_list = []
    tenderee_match_cnt = 0
    tenderee_match_empty_cnt = 0
    real_tenderee_match_cnt1 = 0
    real_tenderee_match_cnt2 = 0
    real_tenderee_match_cnt3 = 0
    win_match_cnt = 0
    win_match_empty_cnt = 0
    real_win_match_cnt1 = 0
    real_win_match_cnt2 = 0
    real_win_match_cnt3 = 0
    for docid, label_data in data_dict_label.items():
        data1 = data_dict_label.get(docid, '{}')
        data2 = data_dict_doubao.get(docid, '{}')
        data3 = data_dict_qwen.get(docid, '{}')
        data4 = data_dict_online.get(docid, '{}')

        extract_json1 = repair_json(data1)
        if len(data2) >= 1000:
            extract_json2 = '{}'
        else:
            extract_json2 = repair_json(data2)
        extract_json3 = repair_json(data3)
        extract_json4 = repair_json(data4)

        extract_json1 = json_repair.loads(extract_json1)
        extract_json2 = json_repair.loads(extract_json2)
        extract_json3 = json_repair.loads(extract_json3)
        extract_json4 = json_repair.loads(extract_json4)

        if type(extract_json1) != dict:
            extract_json1 = {}
        if type(extract_json2) != dict:
            extract_json2 = {}
        if type(extract_json3) != dict:
            extract_json3 = {}
        if type(extract_json4) != dict:
            extract_json4 = {}

        col_match_list = []
        for col in cols:
            str1 = json.dumps(extract_json1.get(col, ""), ensure_ascii=False)
            str2 = json.dumps(extract_json2.get(col, ""), ensure_ascii=False)
            str3 = json.dumps(extract_json3.get(col, ""), ensure_ascii=False)

            if str1 != str3:
                col_match_list += [str1, str2, str3, 0]
            else:
                col_match_list += [str1, str2, str3, 1]

        tenderee_label = extract_json1.get('招标信息', []).get('招标人名称', "")
        try:
            tenderee_doubao = extract_json2.get('招标信息', {}).get('招标人名称', "")
        except:
            print('no 招标信息 招标人名称 extract_json2', extract_json2)
            tenderee_doubao = ''
        try:
            tenderee_qwen = extract_json3.get('招标信息', [])
            if tenderee_qwen:
                tenderee_qwen = tenderee_qwen[0].get('招标人名称', "")
            else:
                tenderee_qwen = ''
        except:
            print('no 招标信息 招标人名称 extract_json3', type(extract_json3), extract_json3)
            traceback.print_exc()
            tenderee_qwen = ''
        tenderee_online = extract_json4.get('招标信息', []).get('招标人名称', "")

        if tenderee_label in ['文中无招标人', '-', '无', None]:
            tenderee_label = ''
        if tenderee_doubao in ['文中无招标人', '-', '无', None] or type(tenderee_doubao) != str:
            tenderee_doubao = ''
        if tenderee_qwen in ['文中无招标人', '-', '无', None] or type(tenderee_qwen) != str:
            tenderee_qwen = ''
        if tenderee_online in ['文中无招标人', '-', '无', None] or type(tenderee_online) != str:
            tenderee_online = ''

        # print('tenderee_doubao', tenderee_doubao)
        if re.sub(' ', '', tenderee_doubao) == re.sub(' ', '', tenderee_qwen):
            tenderee_match_cnt += 1
            if not tenderee_doubao:
                tenderee_match_empty_cnt += 1

        if tenderee_doubao == tenderee_label:
            real_tenderee_match_cnt1 += 1
        if tenderee_qwen == tenderee_label:
            real_tenderee_match_cnt2 += 1
        if tenderee_online == tenderee_label:
            real_tenderee_match_cnt3 += 1

        win_label = extract_json1.get('中标信息', [])
        if win_label:
            win_label = win_label[0].get('中标人名称', "")
        else:
            win_label = ''
        win_doubao = extract_json2.get('中标信息', [])
        if win_doubao:
            win_doubao = win_doubao[0].get('中标人名称', "")
        else:
            win_doubao = ''
        try:
            win_qwen = extract_json3.get('中标信息', [])
            if win_qwen:
                win_qwen = win_qwen[0].get('中标人名称', "")
            else:
                win_qwen = ''

        except:
            print('no 中标人名称 extract_json3', extract_json3)
            win_qwen = ''

        win_online = extract_json4.get('中标信息', [])
        if win_online:
            win_online = win_online[0].get('中标人名称', "")
        else:
            win_online = ''

        if win_label in ['文中无招标人', '-', '无', None]:
            win_label = ''
        if win_doubao in ['文中无招标人', '-', '无', None]:
            win_doubao = ''
        if win_qwen in ['文中无招标人', '-', '无', None]:
            win_qwen = ''
        if win_online in ['文中无招标人', '-', '无', None]:
            win_online = ''

        # print('win_doubao', win_doubao, win_qwen)
        if re.sub(' ', '', win_doubao) == re.sub(' ', '', win_qwen):
            win_match_cnt += 1
            if not win_doubao:
                win_match_empty_cnt += 1

        if win_doubao == win_label:
            real_win_match_cnt1 += 1
        if win_qwen == win_label:
            real_win_match_cnt2 += 1
        if win_online == win_label:
            real_win_match_cnt3 += 1

        result_list.append(
            [docid, data1, data2, data3] +
            col_match_list +
            [tenderee_doubao, tenderee_qwen, tenderee_online, tenderee_label] +
            [str(tenderee_doubao==tenderee_qwen), str(tenderee_doubao==tenderee_label), str(tenderee_online==tenderee_label), str(tenderee_qwen==tenderee_label)] +
            [win_doubao, win_qwen, win_online, win_label] +
            [str(win_doubao==win_qwen), str(win_doubao==win_label), str(win_online==win_label), str(win_qwen==win_label)]
        )

    df_new = pd.DataFrame(result_list)
    df_new.columns = ['docid', 'doubao', 'qwen', 'qwen-lora',
                      '招标信息1', '招标信息2', '招标信息3', '招标信息相同',
                      '中标信息1', '中标信息2', '中标信息3', '中标信息相同',
                      '产品信息1', '产品信息2', '产品信息3', '产品信息相同',
                      'tenderee_doubao', 'tenderee_qwen', 'tenderee_online', 'tenderee_label',
                      'doubao_qwen_相同', 'doubao_标注_相同', '线上_标注_相同', 'qwen_标注_相同',
                      'win_doubao', 'win_qwen', 'win_online', 'win_label',
                      'doubao_qwen_相同', 'doubao_标注_相同', '线上_标注_相同', 'qwen_标注_相同',
                      ]

    df_new.to_excel(r'D:\BIDI_DOC\比地_文档\export_ai_260506_compare_labeled.xlsx', index=False)
    print('tenderee_match_cnt', tenderee_match_cnt, 'tenderee_match_empty_cnt', tenderee_match_empty_cnt)
    print('real_tenderee_match_cnt1', real_tenderee_match_cnt1, real_tenderee_match_cnt1/len(data_dict_label))
    print('real_tenderee_match_cnt2', real_tenderee_match_cnt2, real_tenderee_match_cnt2/len(data_dict_label))
    print('real_tenderee_match_cnt3', real_tenderee_match_cnt3, real_tenderee_match_cnt3/len(data_dict_label))

    # print('len(tenderee_data_dict)', len(tenderee_data_dict))
    print('win_match_cnt', win_match_cnt, 'win_match_empty_cnt', win_match_empty_cnt)
    print('real_win_match_cnt1', real_win_match_cnt1, real_win_match_cnt1/len(data_dict_label))
    print('real_win_match_cnt2', real_win_match_cnt2, real_win_match_cnt2/len(data_dict_label))
    print('real_win_match_cnt3', real_win_match_cnt3, real_win_match_cnt3/len(data_dict_label))


def compare_extract_entity():
    df1 = pd.read_excel('df_train.xlsx')
    df2 = pd.read_excel('df_train_doubao.xlsx')
    # df3 = pd.read_excel(r'D:\BIDI_DOC\比地_文档\export_entity_260414_0.8B_prefix_only_tenderee.xlsx')
    df3 = pd.read_excel(r'D:\BIDI_DOC\比地_文档\export_entity_260414_0.8B_prefix_only_tenderee_all.xlsx')


    df1 = df1[:200]

    data_list1 = df1.astype(object).where(pd.notnull(df1), None).values.tolist()
    data_list2 = df2.astype(object).where(pd.notnull(df2), None).values.tolist()
    data_list3 = df3.astype(object).where(pd.notnull(df3), None).values.tolist()

    data_dict1 = {x[2]: x for x in data_list1}
    data_dict2 = {x[0]: x[2] for x in data_list2}
    data_dict3 = {x[0]: psv_to_dict_prefix(x[1]) for x in data_list3}

    match_cnt_2 = 0
    match_cnt_3 = 0
    all_cnt = 0
    for docid, line in data_dict1.items():
        content = line[5] + line[1] + line[8]
        label = line[4]
        entity = line[1]

        if label == '其他角色':
            continue

        if line[4] != '招标人':
            tenderee1 = '-'
            continue
        else:
            tenderee1 = line[1]

        result_dict2 = data_dict2.get(docid)
        result_dict2 = repair_json(result_dict2)
        result_dict2 = json_repair.loads(result_dict2)
        tenderee2 = result_dict2.get('招标信息', {}).get('招标人名称')
        if not tenderee2:
            tenderee2 = '-'
        if tenderee1 == tenderee2:
            match_cnt_2 += 1
        else:
            print('not match2', tenderee1, tenderee2)
            print('content', content)
            print('label', label, entity)

        result_dict3 = data_dict3.get(docid)
        tenderee3 = result_dict3.get('招标信息', {}).get('招标人名称')
        if not tenderee3:
            tenderee3 = '-'

        if tenderee1 == tenderee3:
            match_cnt_3 += 1
        else:
            print('not match3', tenderee1, tenderee3)

        all_cnt += 1

    print('match_cnt_2', match_cnt_2, all_cnt, match_cnt_2 / all_cnt)
    print('match_cnt_3', match_cnt_3, all_cnt, match_cnt_3 / all_cnt)


if __name__ == '__main__':
    # compare_extract_csv()

    # compare_extract_csv_prefix()

    # compare_extract_entity()

    compare_extract_csv_prefix2()