Lightning Talks: Using Hyperblocks to Solve Word Analogy Problems

View on Snap!Con

Presented By: Ken Kahn


Abstract:

Word embeddings use machine learning to turn words into lists of 300 numbers. By doing some addition and subtraction on these lists using Snap!'s hyperblocks one can solve word analogy problems in a surprisingly straightforward manner. E.g. man is to woman as father is to X.

I'll demonstrate this project that works in 15 different languages.

How much effort would it take to add an extra language. I'm using your material in class but it would help to offer the children their mother tongue which happens to be Dutch ...