Visualizing Word2Vec with TensorBoard

Let me introduce a great app — the Tensor Board Embedding Projector (https://projector.tensorflow.org/). It can graphically represent high-dimensional word embeddings. It lets you choose UMAP, T-SNE, PCA, and other dimensional-reduction techniques. In the right panel, you can enter a word. It can graphically represent high-dimensional word embeddings. It lets you choose UMAP, T-SNE, PCA, and other dimensional-reduction techniques. In the right panel, you can enter a word.

Figure 7.11 – An overview of TensorBoard

I entered the word building. It returns related words, such as construction, houses, constructed, built, and so on:

Figure 7.12 – An overview of the right panel of TensorBoard

The values in Figure 7.12 are the cosine similarities between the vector of building and the vectors of the other words. Remember, cosine similarity measures the similarity between two...