有没有在 spaCy 中打印出最相似句子的功能？答案

【问题标题】：Is there a function to print out the most similar sentence in spaCy?有没有在 spaCy 中打印出最相似句子的功能？
【发布时间】：2020-11-28 15:10:58
【问题描述】：

我有一个包含 10 个电影概要的 txt 文件。我有一个单独的 Hulk 电影的概要，它作为字符串存储在一个变量中。我需要将这 10 个概要与绿巨人的概要进行比较，以找到最相似的电影来推荐。我的代码如下：

import spacy

nlp = spacy.load('en_core_web_lg')

hulk_description = """Will he save their world or destroy it? When the Hulk becomes too dangerous for the
Earth, the Illuminati trick Hulk into a shuttle and launch him into space to a
planet where the Hulk can live in peace. Unfortunately, Hulk land on the
planet Sakaar where he is sold into slavery and trained as a gladiator."""

hulk = nlp(hulk_description)

movies = []

with open('movies.txt', 'r') as f_in:
    for line in map(str.strip, f_in):
        if not line:
            continue
        tmp = line.split()
        movies.append(line)

for token in movies:
    token = nlp(token)
    print(token.similarity(hulk))

所以这行得通，它会打印出以下内容：

0.9299734027118595
0.9045154830561336
0.9248706809139479
0.6760996697288897
0.8521583959686228
0.9340271750528514
0.9251483541429658
0.8806094116148976
0.8709798309015676
0.8489256857995392

我可以看到第 6 部电影简介最相似的是 0.9340271750528514。但我的问题是； spaCy 中是否有一个功能可以让我在完成比较后只打印出最相似的句子？即我基本上想比较所有这些，然后通过显示其概要来推荐最相似的电影。

【问题讨论】：

max(similarities) 不工作吗？

标签： python nlp spacy sentence-similarity

【解决方案1】：

尝试使用这个：

max((nlp(token).similarity(hulk), token) for token in movies)

【讨论】：

谢谢，在我做了一些调整后，效果很好

【解决方案2】：

您可以尝试通过 nlp.pipe 生成器传递电影来计算字符串和电影之间的相似度：

import spacy

nlp = spacy.load('en_core_web_md')

hulk_description = """Will he save their world or destroy it? When the Hulk becomes too dangerous for the
Earth, the Illuminati trick Hulk into a shuttle and launch him into space to a
planet where the Hulk can live in peace. Unfortunately, Hulk land on the
planet Sakaar where he is sold into slavery and trained as a gladiator."""

hulk = nlp(hulk_description)

movies = ["this is a movie", "this is another movie about Hulk"]

sims = []
for movie in nlp.pipe(movies):
    sims.append(hulk.similarity(movie))
    
id_max = np.argmax(sims)
print(movies[id_max])
# this is another movie about Hulk

【讨论】：