Representing the dict

Through word2vec, we were wondering about the representation of one vector. Onto the generative poster, I used some kind of 3D terrain representation where each height represent one dimension of the vector.

But what about the whole dictionary ?

Indeed, reducing the dimension of vector using a Principal component analysis (PCA) algorithm: from the whole dimension, the PCA compute a given number of dimension. Once we get the 3 dimension for each word, we mapped them between two different value to easily add them into a 3d space.

Here I tried to make it using Aframe.js, a framework for VR and 3d environnement. This environment represent the vector space of Poème Saturnien, written by Verlaine.

It kind of made some sort of cylinder through the space. So I need to continue in order to find where that come from

How to deal with bigger dataset ? In fact, the file will be too heavy to be load on the web and in aframe. Main idea about this is to, pick a limited random words or for a given term, search the closet to draw only some part of the 3d word.