使用Gensim Doc2Vec寻找'Doctag'和'infer_vector'之间的距离? [英] Finding the distance between 'Doctag' and 'infer_vector' with Gensim Doc2Vec?
问题描述
使用Gensim的Doc2Vec，我如何找到 Doctag 和 infer_vector()之间的距离?
Using Gensim's Doc2Vec how would I find the distance between a Doctag and an infer_vector()?
非常感谢
推荐答案
Doctag 是doc-vector的键的内部名称. infer_vector()操作的结果是一个向量.因此，正如您从字面上提出的那样，它们是不可比的.
Doctag is the internal name for the keys to doc-vectors. The result of an infer_vector() operation is a vector. So as you've literally asked, these aren't comparable.
您可以通过在训练期间通过 model.docvecs [doctag] 提供的doc-tag密钥，向模型询问已知的doc-vector.这可以与 infer_vector()调用的结果相提并论.
You could ask a model for a known doc-vector, by its doc-tag key that was supplied during training, via model.docvecs[doctag]. That would be comparable to the result of an infer_vector() call.
手头有两个向量，您可以使用 scipy 例程来计算各种距离.例如:
With two vectors in hand, you can use scipy routines to calculate various kinds of distance. For example:
import scipy.spatial.distance.cosine as cosine_distance
vec_by_doctag = model.docvecs["doc0007"]
vec_by_inference = model.infer_vector(['a', 'cat', 'was', 'in', 'a', 'hat'])
dist = cosine_distance(vec_by_doctag, vec_by_inference)
您还可以查看gensim的 Doc2VecKeyedVectors 如何在模型中通过其 similarity()已知(通过其doctag键名)已知的矢量之间的相似度/距离.和 distance()函数，位于:
You can also look at how gensim's Doc2VecKeyedVectors does similarity/distance between vectors that are known (by their doctag key names) inside a model, in its similarity() and distance() functions, at:
这篇关于使用Gensim Doc2Vec寻找'Doctag'和'infer_vector'之间的距离?的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！
