对python中句子相似度的结果进行排序

Question

我试图找到句子标记化文档和得到结果保存在列表中的句子之间的相似性。我想根据相似性评分对结果进行排序。当我尝试根据相似度评分对输出进行排序时，我得到了一个错误?

results=[]

#embedding all the documents and find the similarity between search text and all the tokenize sentences
for docs_sent_token in docs_sent_tokens:
sentence_embeddings = model.encode(docs_sent_token)
sim_score1 = cosine_sim(search_sentence_embeddings, sentence_embeddings)
if sim_score1 > 0:
results.append({
sim_score1,
docs_sent_token,
})
results.sort(key=lambda k : k['sim_score1'] , reverse=True)
print(results)

这是我得到的错误。

TypeError: 'set' object is not subscriptable

这个问题可以使用字典来解决。

if sim_score1 > 0:
results.append({
'Score':sim_score1,
'Token':docs_sent_token,
})
results.sort(key=lambda k : k['Score'] , reverse=True)
print(results)

但是有没有可能使用列表来完成排序呢?我想要得到这种格式的结果。

[{0.91, 'Sentence 1'}, {0.87, 'Sentence 2'}, {0.33, 'Sentence 3'}, {0.30, 'Sentence 4'},

Answer 1

set没有索引或键来指示要排序的值。您可以创建tuples或dicts的列表，对其进行排序并稍后将其转换为sets

results.append((
sim_score1,
docs_sent_token
))
# results = [(0.91, 'Sentence 1'), (0.33, 'Sentence 3'), (0.87, 'Sentence 2'), (0.30, 'Sentence 4')]
results.sort(key=lambda k: k[0], reverse=True)
results = [set(t) for t in results]
# or
results.append({
'Score': sim_score1,
'Token': docs_sent_token
})
# results = [{'Score': 0.91, 'Token': 'Sentence 1'}, {'Score': 0.33, 'Token': 'Sentence 3'}, {'Score': 0.87, 'Token': 'Sentence 2'}, {'Score': 0.30, 'Token': 'Sentence 4'}]
results.sort(key=lambda k: k['Score'], reverse=True)
results = [set(d.values()) for d in results]
print(results)

输出

[{0.91, 'Sentence 1'}, {0.87, 'Sentence 2'}, {0.33, 'Sentence 3'}, {0.3, 'Sentence 4'}]