Single Positional Indexer越界：对话AI中Question转Regresponse报错

问题描述 #

我正在开发一款对话AI，作为Python新手，代码和调试方法大多来自网络。目前除了Question转Regresponse的场景外，其他功能都正常运行，但反复出现Single Positional Indexer out-of-bounds错误，尝试多种调试方式均无效。

代码实现 #

import pandas as pd
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.naive_bayes import MultinomialNB
import random 

# Load the data from the CSV file
data = pd.read_csv("conversational_english.csv")

# Split the data into training and testing sets
training_data = data[:int(0.8 * len(data))]
testing_data = data[int(0.8 * len(data)):]

# Convert the text into numerical feature vectors using CountVectorizer
vectorizer = CountVectorizer()
text_features = vectorizer.fit_transform(training_data['text'])

# Train a Naive Bayes classifier on the training data
classifier = MultinomialNB().fit(text_features, training_data['label'])

# Continuously get user input and generate a response based on the label
while True:
    user_input = input("Enter a conversational text: ")
    if user_input == "exit":
        break
    user_input_features = vectorizer.transform([user_input])
    predicted_label = classifier.predict(user_input_features)[0]
    if predicted_label == 'feeling_question':
        predicted_label = 'feeling_response'
    if predicted_label == 'question':
        predicted_label = 'regresponse'
    if predicted_label == 'joke_request':
        predicted_label = 'joke'
    print(predicted_label)
    response = data.loc[data['label'] == predicted_label, 'text'].iloc[0]
    print("Response:", response)
    correct = input("Is this response correct? (yes/no): ")
    if correct == "no":
        new_label = input("Enter the correct label for the user input: ")
        if user_input in data['text'].values:
            continue
        else:
            new_data = pd.DataFrame({'text': [user_input], 'label': [new_label]})
            data = data.append(new_data, ignore_index=True)
            data.to_csv("conversational_english.csv", index=False)
    else: 
        if user_input in data['text'].values:
            continue
        else:
            new_data = pd.DataFrame({'text': [user_input], 'label': [predicted_label]})
            data = data.append(new_data, ignore_index=True)
            data.to_csv("conversational_english.csv", index=False)

CSV数据 #

text,label,
Can you tell me a joke?,joke_request,
I'm not sure because I am an AI, regresponse,
Hi how are you doing today?,greeting,
Why did the chicken cross the road?,joke,
Where is the bus stop?,question,
how are you?,feeling_question,
What are you doing?,feeling_question,
I am an AI. How should I know?, regresponse,
How are you?,feeling_question,
Goodbye,farewell,
Bye,farewell,
"Two whales walk into a bar. One says ""oOooOoooOOh"". The other says ""What the hell Jim""",joke,
Hi!,greeting,
I'm not sure. I am an AI, regresponse,
Hello!,greeting,
Greetings!,greeting,
what's up?,question,
I'm doing good!,feeling_response,
What's up?,question,
Tell me a joke,joke_request,
Tell me something funny,joke_request,
I'm doing great!,feeling_reponse,
I am an AI. I don't know., regresponse,
Not bad!,feeling_response,
Hi,greeting,
Hello,greeting,
how's it going?,question,
My name is Zach,name,
how's it goin,question,
what are you up to?,question,
I don't know I am and AI, regresponse,
How are you,feeling_question,
How's it goin,question,
How do you do,feeling_question,
"A horse walks into a bar and the bartender says ""why the long face""",joke,
goodbye,farewell,
hello,greeting,
Good evening,greeting,
I'm not sure., regresponse,
hello!,greeting,
What's up,question,
Heyo,greeting,
hi,greeting,
I am an AI. I can't help you., regresponse,
How's it going?,question,
I don't know because I am an AI., regresponse,
What's up!,question,

报错信息 #

Traceback (most recent call last):
  File "C:\Users\Kyn\OneDrive\Documents\test-rep\Python AI\response.py", line 35, in <module>
    response = data.loc[data['label'] == predicted_label, 'text'].iloc[0]
  File "C:\Users\Kyn\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.10_qbz5n2kfra8p0\LocalCache\local-packages\Python310\site-packages\pandas\core\indexing.py", line 1073, in __getitem__
    return self._getitem_axis(maybe_callable, axis=axis)
  File "C:\Users\Kyn\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.10_qbz5n2kfra8p0\LocalCache\local-packages\Python310\site-packages\pandas\core\indexing.py", line 1625, in _getitem_axis
    self._validate_integer(key, axis)
  File "C:\Users\Kyn\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.10_qbz5n2kfra8p0\LocalCache\local-packages\Python310\site-packages\pandas\core\indexing.py", line 1557, in _validate_integer
    raise IndexError("single positional indexer is out-of-bounds")

错误原因 #

核心问题出在CSV数据的标签格式不统一：

1. 所有regresponse标签前面都有一个多余的空格（比如 regresponse），但代码中转换后的标签是regresponse（无空格），导致data.loc[data['label'] == 'regresponse']找不到任何匹配行，空的Series调用.iloc[0]就会触发索引越界错误。
2. CSV中存在拼写错误：feeling_reponse应为feeling_response，这会导致后续匹配该标签时同样可能出现无结果的情况。

修复方案 #

1. 清理CSV标签数据 #

打开conversational_english.csv，做两处修改：

• 把所有 regresponse（带前置空格）替换为regresponse（无空格）
• 把feeling_reponse替换为feeling_response

2. 代码层面增加容错与优化 #

（1）加载数据时自动清理标签

在读取CSV后，添加一行代码自动去除标签的前后空格，避免后续再出现空格问题：

data = pd.read_csv("conversational_english.csv")
# 清理label列的前后空白字符
data['label'] = data['label'].str.strip()

（2）增加空结果判断，避免索引越界

替换原来获取响应的代码，先检查是否有匹配行，没有则返回默认提示，同时随机选择响应让对话更自然：

# 替换原response = ...那一行
matching_responses = data.loc[data['label'] == predicted_label, 'text']
if not matching_responses.empty:
    # 随机选一个匹配的响应
    response = random.choice(matching_responses.tolist())
else:
    response = "Sorry, I don't have a suitable response right now."

（3）简化标签转换逻辑

用字典替代多个if判断，代码更简洁易维护：

# 替换原三个if标签转换的代码
label_mapping = {
    'feeling_question': 'feeling_response',
    'question': 'regresponse',
    'joke_request': 'joke'
}
predicted_label = label_mapping.get(predicted_label, predicted_label)

内容的提问来源于stack exchange，提问作者Zach