我正在尝试找到一种检查两个句子之间相似度的方法。
以下这些库大多数都可以用于语义相似度比较。您可以通过使用这些库中预训练模型生成单词或句子向量来跳过直接单词比较。
Spacy
进行句子相似度比较首先必须加载所需的模型。
要使用 en_core_web_md
,请使用 python -m spacy download en_core_web_md
进行下载。要使用 en_core_web_lg
,请使用 python -m spacy download en_core_web_lg
进行下载。
由于较大的模型大小约为830MB且速度相当慢,因此选择中等大小的模型可能是一个不错的选择。
https://spacy.io/usage/vectors-similarity/
代码:
import spacy
nlp = spacy.load("en_core_web_lg")
#nlp = spacy.load("en_core_web_md")
doc1 = nlp(u'the person wear red T-shirt')
doc2 = nlp(u'this person is walking')
doc3 = nlp(u'the boy wear red T-shirt')
print(doc1.similarity(doc2))
print(doc1.similarity(doc3))
print(doc2.similarity(doc3))
输出:
0.7003971105290047
0.9671912343259517
0.6121211244876517
https://github.com/UKPLab/sentence-transformers
https://www.sbert.net/docs/usage/semantic_textual_similarity.html
使用pip install -U sentence-transformers
进行安装,该库可生成句子嵌入。
代码:
from sentence_transformers import SentenceTransformer
model = SentenceTransformer('distilbert-base-nli-mean-tokens')
sentences = [
'the person wear red T-shirt',
'this person is walking',
'the boy wear red T-shirt'
]
sentence_embeddings = model.encode(sentences)
for sentence, embedding in zip(sentences, sentence_embeddings):
print("Sentence:", sentence)
print("Embedding:", embedding)
print("")
输出:
Sentence: the person wear red T-shirt
Embedding: [ 1.31643847e-01 -4.20616418e-01 ... 8.13076794e-01 -4.64620918e-01]
Sentence: this person is walking
Embedding: [-3.52878094e-01 -5.04286848e-02 ... -2.36091137e-01 -6.77282438e-02]
Sentence: the boy wear red T-shirt
Embedding: [-2.36365378e-01 -8.49713564e-01 ... 1.06414437e+00 -2.70157874e-01]
现在可以使用嵌入向量来计算各种相似性度量。
代码:
from sentence_transformers import SentenceTransformer, util
print(util.pytorch_cos_sim(sentence_embeddings[0], sentence_embeddings[1]))
print(util.pytorch_cos_sim(sentence_embeddings[0], sentence_embeddings[2]))
print(util.pytorch_cos_sim(sentence_embeddings[1], sentence_embeddings[2]))
输出:
tensor([[0.4644]])
tensor([[0.9070]])
tensor([[0.3276]])
同样的情况也适用于scipy
和pytorch
,
代码:
from scipy.spatial import distance
print(1 - distance.cosine(sentence_embeddings[0], sentence_embeddings[1]))
print(1 - distance.cosine(sentence_embeddings[0], sentence_embeddings[2]))
print(1 - distance.cosine(sentence_embeddings[1], sentence_embeddings[2]))
输出:
0.4643629193305969
0.9069876074790955
0.3275738060474396
代码:
import torch.nn
cos = torch.nn.CosineSimilarity(dim=0, eps=1e-6)
b = torch.from_numpy(sentence_embeddings)
print(cos(b[0], b[1]))
print(cos(b[0], b[2]))
print(cos(b[1], b[2]))
输出:
tensor(0.4644)
tensor(0.9070)
tensor(0.3276)
TFHub Universal Sentence Encoder
进行句子相似度比较https://tfhub.dev/google/universal-sentence-encoder/4
这个模型非常大,约为1GB,并且似乎比其他模型慢。它还为句子生成嵌入。
代码:
import tensorflow_hub as hub
embed = hub.load("https://tfhub.dev/google/universal-sentence-encoder/4")
embeddings = embed([
"the person wear red T-shirt",
"this person is walking",
"the boy wear red T-shirt"
])
print(embeddings)
输出:
tf.Tensor(
[[ 0.063188 0.07063895 -0.05998802 ... -0.01409875 0.01863449
0.01505797]
[-0.06786212 0.01993554 0.03236153 ... 0.05772103 0.01787272
0.01740014]
[ 0.05379306 0.07613157 -0.05256693 ... -0.01256405 0.0213196
-0.00262441]], shape=(3, 512), dtype=float32)
代码:
from scipy.spatial import distance
print(1 - distance.cosine(embeddings[0], embeddings[1]))
print(1 - distance.cosine(embeddings[0], embeddings[2]))
print(1 - distance.cosine(embeddings[1], embeddings[2]))
输出:
0.15320375561714172
0.8592830896377563
0.09080004692077637
https://github.com/facebookresearch/InferSent
https://github.com/Tiiiger/bert_score
此示意图演示了该方法,
https://en.wikipedia.org/wiki/Cosine_similarity#Angular_distance_and_similarity
https://towardsdatascience.com/word-distance-between-word-embeddings-cc3e9cf1d632
https://docs.scipy.org/doc/scipy-0.14.0/reference/generated/scipy.spatial.distance.cosine.html
https://www.tensorflow.org/api_docs/python/tf/keras/losses/CosineSimilarity
我们在组织中实施了一种非常简单但非常有效的方法。 我们不需要语义相似性,因此使用了以下方法。