This section will describe further analysis that can be performed on the data after visualization. For example, exploring cosine distance similarity between different word vectors.
The following link is a great blog on how cosine distance similarity works and also discusses some of the math involved:
Consider the following:
- Various natural-language processing tasks can be performed using the different functions of
Word2Vec
. One of them is finding the most semantically similar words given a certain word (that is, word vectors that have a high cosine similarity or a short Euclidean distance between them). This can be done by using themost_similar
function formWord2Vec
, as shown in the following screenshot: This screenshots all the closest words related to the wordLannister
: This screenshot shows a list of all the words related to wordJon...