Chris Pollett > Students >

    ( Print View)



    [CS 297 Proposal]

    [Comparison between hash2vec and word2vec -pdf]

    [Different Approaches for word2vec from reference paper -pdf]

    [Deliverable 1]

    [Deliverable 2]

    [Deliverable 3]

    [Deliverable 4]

    [Deliverable 5]


    [CS 298 Proposal]



Among different distance metrics, cosine similarity is simple and most used in word2vec. It is normalized dot product of 2 vectors and this ratio defines the angle between them. Two vectors with the same orientation have a cosine similarity of 1, two vectors at 90 degrees have a similarity of 0, and two vectors diametrically opposed have a similarity of -1, independent of their magnitude.

I have applied cosine function to compare the linear relationship between words. E.g. if I compare words depression and geology those words should be more similar i.e. the distance between them should be less and hence cosine similarity would be more for these words. This particular property of word2vec shows its linear compositionality.