Nearest Neighbors Comparison

This interface allows you to retrieve a set of N near neighbor terms based on a submitted term or piece of text (pseudodoc). The terms returned are those in the LSA or word2vec space which are nearest the submitted term or pseudodoc. The system will compute a cosine similarity score for the submitted word compared to all nearest neighbors. Check the box below to view a 3D visualization of the nearest neighbors.

To try the system, enter a term or piece of text in the input area below. Then press the 'Submit Text' button.

Note that BERT is not implemented for the nearest neighbors comparison.


Select the word embedding method:

Number of terms to return:

Remove terms from return list that appear in corpus with frequency less than (<=):

Text to submit: