BERT的中文问答系统33

weixin_543662862024-11-22 15:44

我们在现有的代码基础上增加网络搜索的功能。我们使用 requests 和 BeautifulSoup 来从百度搜索结果中提取信息。以下是完整的代码，包括项目结构、README.md 文件以及所有必要的代码。

项目结构

bash 复制代码 
xihe241117/
├── data/
│   └── train_data.jsonl
├── logs/
├── models/
│   └── xihua_model.pth
├── requirements.txt
├── README.md
└── xihe_chatbot.py

README.md

bash 复制代码 
# 羲和聊天机器人

## 项目介绍
羲和聊天机器人是一个基于BERT的中文问答系统，支持用户提问并获取回答。如果模型提供的回答不满意，用户可以选择"不正确"，机器人将自动从百度搜索相关信息并提供更详细的答案。

## 目录结构
xihe241117/
├── data/
│ └── train_data.jsonl
├── logs/
├── models/
│ └── xihua_model.pth
├── requirements.txt
├── README.md
└── xihe_chatbot.py

## 安装依赖

pip install -r requirements.txt

运行项目

python xihe_chatbot.py

功能
用户提问
模型提供回答
用户评价回答（正确/不正确）
如果回答不正确，自动从百度搜索相关信息
查看历史记录
保存历史记录
训练模型
重新训练模型
评估模型（暂未实现）
联系我们
如有任何问题或建议，请联系 [554687453@qq.com]。

requirements.txt

txt 复制代码 
torch
transformers
jsonlines
tkinter
requests
beautifulsoup4

xihe_chatbot.py

python 复制代码 
import os
import json
import jsonlines
import torch
import torch.optim as optim
from torch.utils.data import Dataset, DataLoader
from transformers import BertModel, BertTokenizer
import tkinter as tk
from tkinter import filedialog, messagebox, ttk
import logging
from difflib import SequenceMatcher
from datetime import datetime
import requests
from bs4 import BeautifulSoup

# 获取项目根目录
PROJECT_ROOT = os.path.dirname(os.path.abspath(__file__))

# 配置日志
LOGS_DIR = os.path.join(PROJECT_ROOT, 'logs')
os.makedirs(LOGS_DIR, exist_ok=True)

def setup_logging():
    log_file = os.path.join(LOGS_DIR, datetime.now().strftime('%Y-%m-%d_%H-%M-%S_羲和.txt'))
    logging.basicConfig(
        level=logging.INFO,
        format='%(asctime)s - %(levelname)s - %(message)s',
        handlers=[
            logging.FileHandler(log_file),
            logging.StreamHandler()
        ]
    )

setup_logging()

# 数据集类
class XihuaDataset(Dataset):
    def __init__(self, file_path, tokenizer, max_length=128):
        self.tokenizer = tokenizer
        self.max_length = max_length
        self.data = self.load_data(file_path)

    def load_data(self, file_path):
        data = []
        if file_path.endswith('.jsonl'):
            with jsonlines.open(file_path) as reader:
                for i, item in enumerate(reader):
                    try:
                        data.append(item)
                    except jsonlines.jsonlines.InvalidLineError as e:
                        logging.warning(f"跳过无效行 {i + 1}: {e}")
        elif file_path.endswith('.json'):
            with open(file_path, 'r') as f:
                try:
                    data = json.load(f)
                except json.JSONDecodeError as e:
                    logging.warning(f"跳过无效文件 {file_path}: {e}")
        return data

    def __len__(self):
        return len(self.data)

    def __getitem__(self, idx):
        item = self.data[idx]
        question = item['question']
        human_answer = item['human_answers'][0]
        chatgpt_answer = item['chatgpt_answers'][0]

        try:
            inputs = self.tokenizer(question, return_tensors='pt', padding='max_length', truncation=True, max_length=self.max_length)
            human_inputs = self.tokenizer(human_answer, return_tensors='pt', padding='max_length', truncation=True, max_length=self.max_length)
            chatgpt_inputs = self.tokenizer(chatgpt_answer, return_tensors='pt', padding='max_length', truncation=True, max_length=self.max_length)
        except Exception as e:
            logging.warning(f"跳过无效项 {idx}: {e}")
            return self.__getitem__((idx + 1) % len(self.data))

        return {
            'input_ids': inputs['input_ids'].squeeze(),
            'attention_mask': inputs['attention_mask'].squeeze(),
            'human_input_ids': human_inputs['input_ids'].squeeze(),
            'human_attention_mask': human_inputs['attention_mask'].squeeze(),
            'chatgpt_input_ids': chatgpt_inputs['input_ids'].squeeze(),
            'chatgpt_attention_mask': chatgpt_inputs['attention_mask'].squeeze(),
            'human_answer': human_answer,
            'chatgpt_answer': chatgpt_answer
        }

# 获取数据加载器
def get_data_loader(file_path, tokenizer, batch_size=8, max_length=128):
    dataset = XihuaDataset(file_path, tokenizer, max_length)
    return DataLoader(dataset, batch_size=batch_size, shuffle=True)

# 模型定义
class XihuaModel(torch.nn.Module):
    def __init__(self, pretrained_model_name='F:/models/bert-base-chinese'):
        super(XihuaModel, self).__init__()
        self.bert = BertModel.from_pretrained(pretrained_model_name)
        self.classifier = torch.nn.Linear(self.bert.config.hidden_size, 1)

    def forward(self, input_ids, attention_mask):
        outputs = self.bert(input_ids=input_ids, attention_mask=attention_mask)
        pooled_output = outputs.pooler_output
        logits = self.classifier(pooled_output)
        return logits

# 训练函数
def train(model, data_loader, optimizer, criterion, device, progress_var=None):
    model.train()
    total_loss = 0.0
    num_batches = len(data_loader)
    for batch_idx, batch in enumerate(data_loader):
        try:
            input_ids = batch['input_ids'].to(device)
            attention_mask = batch['attention_mask'].to(device)
            human_input_ids = batch['human_input_ids'].to(device)
            human_attention_mask = batch['human_attention_mask'].to(device)
            chatgpt_input_ids = batch['chatgpt_input_ids'].to(device)
            chatgpt_attention_mask = batch['chatgpt_attention_mask'].to(device)

            optimizer.zero_grad()
            human_logits = model(human_input_ids, human_attention_mask)
            chatgpt_logits = model(chatgpt_input_ids, chatgpt_attention_mask)

            human_labels = torch.ones(human_logits.size(0), 1).to(device)
            chatgpt_labels = torch.zeros(chatgpt_logits.size(0), 1).to(device)

            loss = criterion(human_logits, human_labels) + criterion(chatgpt_logits, chatgpt_labels)
            loss.backward()
            optimizer.step()

            total_loss += loss.item()
            if progress_var:
                progress_var.set((batch_idx + 1) / num_batches * 100)
        except Exception as e:
            logging.warning(f"跳过无效批次: {e}")

    return total_loss / len(data_loader)

# 主训练函数
def main_train(retrain=False):
    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
    logging.info(f'Using device: {device}')

    tokenizer = BertTokenizer.from_pretrained('F:/models/bert-base-chinese')
    model = XihuaModel(pretrained_model_name='F:/models/bert-base-chinese').to(device)

    if retrain:
        model_path = os.path.join(PROJECT_ROOT, 'models/xihua_model.pth')
        if os.path.exists(model_path):
            model.load_state_dict(torch.load(model_path, map_location=device))
            logging.info("加载现有模型")
        else:
            logging.info("没有找到现有模型，将使用预训练模型")

    optimizer = optim.Adam(model.parameters(), lr=1e-5)
    criterion = torch.nn.BCEWithLogitsLoss()

    train_data_loader = get_data_loader(os.path.join(PROJECT_ROOT, 'data/train_data.jsonl'), tokenizer, batch_size=8, max_length=128)

    num_epochs = 30
    for epoch in range(num_epochs):
        train_loss = train(model, train_data_loader, optimizer, criterion, device)
        logging.info(f'Epoch [{epoch+1}/{num_epochs}], Loss: {train_loss:.8f}')

    torch.save(model.state_dict(), os.path.join(PROJECT_ROOT, 'models/xihua_model.pth'))
    logging.info("模型训练完成并保存")

# 网络搜索函数
def search_baidu(query):
    url = f"https://www.baidu.com/s?wd={query}"
    headers = {
        'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/58.0.3029.110 Safari/537.3'
    }
    response = requests.get(url, headers=headers)
    soup = BeautifulSoup(response.text, 'html.parser')
    results = soup.find_all('div', class_='c-abstract')
    if results:
        return results[0].get_text().strip()
    return "没有找到相关信息"

# GUI界面
class XihuaChatbotGUI:
    def __init__(self, root):
        self.root = root
        self.root.title("羲和聊天机器人")

        self.tokenizer = BertTokenizer.from_pretrained('F:/models/bert-base-chinese')
        self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
        self.model = XihuaModel(pretrained_model_name='F:/models/bert-base-chinese').to(self.device)
        self.load_model()
        self.model.eval()

        # 加载训练数据集以便在获取答案时使用
        self.data = self.load_data(os.path.join(PROJECT_ROOT, 'data/train_data.jsonl'))

        # 历史记录
        self.history = []

        self.create_widgets()

    def create_widgets(self):
        # 顶部框架
        top_frame = tk.Frame(self.root)
        top_frame.pack(pady=10)

        self.question_label = tk.Label(top_frame, text="问题:", font=("Arial", 12))
        self.question_label.grid(row=0, column=0, padx=10)

        self.question_entry = tk.Entry(top_frame, width=50, font=("Arial", 12))
        self.question_entry.grid(row=0, column=1, padx=10)

        self.answer_button = tk.Button(top_frame, text="获取回答", command=self.get_answer, font=("Arial", 12))
        self.answer_button.grid(row=0, column=2, padx=10)

        # 中部框架
        middle_frame = tk.Frame(self.root)
        middle_frame.pack(pady=10)

        self.answer_label = tk.Label(middle_frame, text="回答:", font=("Arial", 12))
        self.answer_label.grid(row=0, column=0, padx=10)

        self.answer_text = tk.Text(middle_frame, height=10, width=70, font=("Arial", 12))
        self.answer_text.grid(row=1, column=0, padx=10)

        # 底部框架
        bottom_frame = tk.Frame(self.root)
        bottom_frame.pack(pady=10)

        self.correct_button = tk.Button(bottom_frame, text="准确", command=self.mark_correct, font=("Arial", 12))
        self.correct_button.grid(row=0, column=0, padx=10)

        self.incorrect_button = tk.Button(bottom_frame, text="不准确", command=self.mark_incorrect, font=("Arial", 12))
        self.incorrect_button.grid(row=0, column=1, padx=10)

        self.train_button = tk.Button(bottom_frame, text="训练模型", command=self.train_model, font=("Arial", 12))
        self.train_button.grid(row=0, column=2, padx=10)

        self.retrain_button = tk.Button(bottom_frame, text="重新训练模型", command=lambda: self.train_model(retrain=True), font=("Arial", 12))
        self.retrain_button.grid(row=0, column=3, padx=10)

        self.progress_var = tk.DoubleVar()
        self.progress_bar = ttk.Progressbar(bottom_frame, variable=self.progress_var, maximum=100, length=200)
        self.progress_bar.grid(row=1, column=0, columnspan=4, pady=10)

        self.log_text = tk.Text(bottom_frame, height=10, width=70, font=("Arial", 12))
        self.log_text.grid(row=2, column=0, columnspan=4, pady=10)

        self.evaluate_button = tk.Button(bottom_frame, text="评估模型", command=self.evaluate_model, font=("Arial", 12))
        self.evaluate_button.grid(row=3, column=0, padx=10, pady=10)

        self.history_button = tk.Button(bottom_frame, text="查看历史记录", command=self.view_history, font=("Arial", 12))
        self.history_button.grid(row=3, column=1, padx=10, pady=10)

        self.save_history_button = tk.Button(bottom_frame, text="保存历史记录", command=self.save_history, font=("Arial", 12))
        self.save_history_button.grid(row=3, column=2, padx=10, pady=10)

    def get_answer(self):
        question = self.question_entry.get()
        if not question:
            messagebox.showwarning("输入错误", "请输入问题")
            return

        inputs = self.tokenizer(question, return_tensors='pt', padding='max_length', truncation=True, max_length=128)
        with torch.no_grad():
            input_ids = inputs['input_ids'].to(self.device)
            attention_mask = inputs['attention_mask'].to(self.device)
            logits = self.model(input_ids, attention_mask)
        
        if logits.item() > 0:
            answer_type = "羲和回答"
        else:
            answer_type = "零回答"

        specific_answer = self.get_specific_answer(question, answer_type)

        self.answer_text.delete(1.0, tk.END)
        self.answer_text.insert(tk.END, f"{answer_type}\n{specific_answer}")

        # 添加到历史记录
        self.history.append({
            'question': question,
            'answer_type': answer_type,
            'specific_answer': specific_answer,
            'accuracy': None  # 初始状态为未评价
        })

    def get_specific_answer(self, question, answer_type):
        # 使用模糊匹配查找最相似的问题
        best_match = None
        best_ratio = 0.0
        for item in self.data:
            ratio = SequenceMatcher(None, question, item['question']).ratio()
            if ratio > best_ratio:
                best_ratio = ratio
                best_match = item

        if best_match:
            if answer_type == "羲和回答":
                return best_match['human_answers'][0]
            else:
                return best_match['chatgpt_answers'][0]
        return "这个我也不清楚，你问问零吧"

    def load_data(self, file_path):
        data = []
        if file_path.endswith('.jsonl'):
            with jsonlines.open(file_path) as reader:
                for i, item in enumerate(reader):
                    try:
                        data.append(item)
                    except jsonlines.jsonlines.InvalidLineError as e:
                        logging.warning(f"跳过无效行 {i + 1}: {e}")
        elif file_path.endswith('.json'):
            with open(file_path, 'r') as f:
                try:
                    data = json.load(f)
                except json.JSONDecodeError as e:
                    logging.warning(f"跳过无效文件 {file_path}: {e}")
        return data

    def load_model(self):
        model_path = os.path.join(PROJECT_ROOT, 'models/xihua_model.pth')
        if os.path.exists(model_path):
            self.model.load_state_dict(torch.load(model_path, map_location=self.device))
            logging.info("加载现有模型")
        else:
            logging.info("没有找到现有模型，将使用预训练模型")

    def train_model(self, retrain=False):
        file_path = filedialog.askopenfilename(filetypes=[("JSONL files", "*.jsonl"), ("JSON files", "*.json")])
        if not file_path:
            messagebox.showwarning("文件选择错误", "请选择一个有效的数据文件")
            return

        try:
            dataset = XihuaDataset(file_path, self.tokenizer)
            data_loader = DataLoader(dataset, batch_size=8, shuffle=True)
            
            # 加载已训练的模型权重
            if retrain:
                self.model.load_state_dict(torch.load(os.path.join(PROJECT_ROOT, 'models/xihua_model.pth'), map_location=self.device))
                self.model.to(self.device)
                self.model.train()

            optimizer = torch.optim.Adam(self.model.parameters(), lr=1e-5)
            criterion = torch.nn.BCEWithLogitsLoss()
            num_epochs = 30
            for epoch in range(num_epochs):
                train_loss = train(self.model, data_loader, optimizer, criterion, self.device, self.progress_var)
                logging.info(f'Epoch [{epoch+1}/{num_epochs}], Loss: {train_loss:.4f}')
                self.log_text.insert(tk.END, f'Epoch [{epoch+1}/{num_epochs}], Loss: {train_loss:.4f}\n')
                self.log_text.see(tk.END)
            torch.save(self.model.state_dict(), os.path.join(PROJECT_ROOT, 'models/xihua_model.pth'))
            logging.info("模型训练完成并保存")
            self.log_text.insert(tk.END, "模型训练完成并保存\n")
            self.log_text.see(tk.END)
            messagebox.showinfo("训练完成", "模型训练完成并保存")
        except Exception as e:
            logging.error(f"模型训练失败: {e}")
            self.log_text.insert(tk.END, f"模型训练失败: {e}\n")
            self.log_text.see(tk.END)
            messagebox.showerror("训练失败", f"模型训练失败: {e}")

    def evaluate_model(self):
        # 这里可以添加模型评估的逻辑
        messagebox.showinfo("评估结果", "模型评估功能暂未实现")

    def mark_correct(self):
        if self.history:
            self.history[-1]['accuracy'] = True
            messagebox.showinfo("评价成功", "您认为这次回答是准确的")

    def mark_incorrect(self):
        if self.history:
            self.history[-1]['accuracy'] = False
            question = self.history[-1]['question']
            answer = search_baidu(question)
            self.answer_text.delete(1.0, tk.END)
            self.answer_text.insert(tk.END, f"搜索引擎结果:\n{answer}")
            messagebox.showinfo("评价成功", "您认为这次回答是不准确的")

    def view_history(self):
        history_window = tk.Toplevel(self.root)
        history_window.title("历史记录")

        history_text = tk.Text(history_window, height=20, width=80, font=("Arial", 12))
        history_text.pack(padx=10, pady=10)

        for entry in self.history:
            history_text.insert(tk.END, f"问题: {entry['question']}\n")
            history_text.insert(tk.END, f"回答类型: {entry['answer_type']}\n")
            history_text.insert(tk.END, f"具体回答: {entry['specific_answer']}\n")
            if entry['accuracy'] is None:
                history_text.insert(tk.END, "评价: 未评价\n")
            elif entry['accuracy']:
                history_text.insert(tk.END, "评价: 准确\n")
            else:
                history_text.insert(tk.END, "评价: 不准确\n")
            history_text.insert(tk.END, "-" * 50 + "\n")

    def save_history(self):
        file_path = filedialog.asksaveasfilename(defaultextension=".json", filetypes=[("JSON files", "*.json")])
        if not file_path:
            return

        with open(file_path, 'w') as f:
            json.dump(self.history, f, ensure_ascii=False, indent=4)

        messagebox.showinfo("保存成功", "历史记录已保存到文件")

# 主函数
if __name__ == "__main__":
    # 启动GUI
    root = tk.Tk()
    app = XihuaChatbotGUI(root)
    root.mainloop()

说明

网络搜索功能：增加了 search_baidu 函数，用于从百度搜索相关信息。

标记不准确回答：在 mark_incorrect 方法中，如果用户标记回答为不准确，将调用 search_baidu 函数获取更详细的信息并显示在文本框中。

项目结构：确保项目结构与 README.md 中描述的一致。

这样，您的聊天机器人不仅可以通过模型提供回答，还可以在网络搜索中获取更详细的信息，提高用户体验。

相关推荐
做萤石二次开发的哈哈5 小时前
连锁门店、园区、交通枢纽的视频如何统一汇集？应用模板的接入与分权管理详解
人工智能·物联网·音视频·二次开发·视频监控·蓝海aiot一站式工作台·萤石云开放平台
老金带你玩AI5 小时前
全程拿着手机，我让dot准备了一场AI分享
人工智能
nvd115 小时前
Flink 批处理驱动的湖仓 AI 财务研报流水线：从双流广播汇聚到 Agent 智能决策与原生 Iceberg 落盘全景实战
大数据·人工智能·flink
snhna5 小时前
向量数据库深度实战：从 HNSW 到 pgvector、Milvus 选型与 RAG 落地
数据库·人工智能·python·postgresql·milvus
架构技术专栏5 小时前
Agent 工具调用：从工具声明到执行反馈
前端·人工智能·后端
点云-激光雷达-Slam-三维牙齿6 小时前
100,000 Records CSV File - India Traffic Accident Analysis Dataset.7z
人工智能·数据集
TechWayfarer6 小时前
精度极限的突破：AI与IPv6如何重塑高精度IP查询
人工智能·网络协议·tcp/ip·transformer
七十二时9566 小时前
视频掉帧怎么用AI补帧
人工智能
可达智灵6 小时前
信创适配｜织灵 Coda Loom 2.0 完成银河麒麟服务器操作系统 V10、V11 适配认证
人工智能·可达智灵
Wx-bishekaifayuan6 小时前
express皖美特色农产品网售系统53118-计算机课程设计、毕业设计
人工智能·spring boot·后端·python·课程设计·express·旅游
热门推荐
01国内可直接用、免费额度/永久免费的大模型API清单（含 SiliconFlow、火山、阿里、智谱、百度、Kimi、DeepSeek、DMXAPI 等）02历年考研数学一、数学二、数学三真题试卷及答案PDF03GitHub 镜像站点04「全语录+省流版」邵艾伦 × 孙宇晨：聊风口、聊个人IP、聊年轻人如何抓住 AI 时代的机会05如何新建文件夹？ 电脑新建文件夹的4种方法06蜘蛛磁力 搜索引擎大全，如何使用蜘蛛磁力查找磁力链接07轻量化AI渗透武器库｜ARTEX自主攻防系统、资产梳理、漏洞挖掘、链路分析全流程智能化落地08Windows 10/11 如何退出微软账户并改用本地账户登录（含找不到选项的解决方法）09Telegram APK 下载安装完整指南 — 2026年最新10微信历史版本含下载地址（ Windows PC | 安卓 | MAC ）及设置微信不更新