如何将Python中余弦相似度匹配的打印输出转为Pandas DataFrame

你可以通过以下步骤将输出转换为Pandas DataFrame：

1. 导入pandas库
2. 初始化一个空列表用于存储所有匹配结果
3. 替换原有的打印逻辑，将每个query、匹配的corpus文本和相似度分数以字典形式存入列表
4. 最后将列表转换为DataFrame

修改后的完整代码如下：

import pandas as pd
from sentence_transformers import SentenceTransformer, util
import torch

embedder = SentenceTransformer('all-MiniLM-L6-v2')

corpus = ['About us. About Us · Our Coffees · Starbucks Stories & News · Starbucks® Ready to Drink · Foodservice Coffee · Customer Service · Tax Strategy 2022 · Careers.',
          'Costa is the Nation Favourite coffee shop and the largest and fastest growing coffee shop chain in the UK.',
          'Leading UK speciality coffee roaster with a focus on sustainability. B Corp certified. Become a wholesale partner or buy coffee beans online today.',
          'Kick-start your morning with our amazing range of speciality coffee and equipment. World-class coffee, direct from the farmer, delivered free every time.',
          'Coffee Direct - Freshly roasted coffee beans delivered to your door. Origin coffee, coffee blends and flavoured coffee for bean-to-cup',
          'Whether you prefer whole coffee beans or freshly ground coffee, Whittard of Chelsea selection of light, medium and dark roast luxury coffees has something',
          'Coffee beans are the seeds of a fruit called a coffee cherry. Coffee cherries grow on coffee trees from a genus of plants called Coffea.',
          'On these coffee plants, bunches of cherries grow and inside these you will find two coffee beans, Arabica and Robusta coffee.',
          ]
corpus_embeddings = embedder.encode(corpus, convert_to_tensor=True)

# Query sentences:
queries = ['coffee', 'coffee near me', 'coffee bean', 'coffee house', 'coffee jelly','coffee order nyt crossword clue','coffee quotes', 'coffee shops near me']

# 初始化存储结果的列表
results = []

# Find the closest 5 sentences of the corpus for each query sentence based on cosine similarity
top_k = min(5, len(corpus))
for query in queries:
    query_embedding = embedder.encode(query, convert_to_tensor=True)
    
    cos_scores = util.cos_sim(query_embedding, corpus_embeddings)[0]
    top_results = torch.topk(cos_scores, k=top_k)

    # 将结果存入列表
    for score, idx in zip(top_results[0], top_results[1]):
        results.append({
            "query": query,
            "corpus": corpus[idx],
            "score": round(score.item(), 4)
        })

# 转换为DataFrame
similarity_df = pd.DataFrame(results)

# 可以打印查看结果，或者保存为CSV
print(similarity_df.head(10))
# similarity_df.to_csv("coffee_similarity_results.csv", index=False)

关键修改说明： #

• 使用results列表收集所有匹配项，每个元素是包含query、corpus、score的字典
• 通过score.item()将PyTorch张量转换为Python浮点数，并用round()保留4位小数，和原打印格式一致
• 最后用pd.DataFrame()将列表转换为结构化的表格数据

生成的DataFrame示例（部分）： #

内容的提问来源于stack exchange，提问作者Simone De Palma