The Netherlands eScience Center's engineers write about research software, covering Rust, FPGAs, Python, C++, GraphQL, and word embeddings.
HN commenters cite its post on king minus man plus woman to show the word2vec analogy result is partly an artifact of how it was computed.
First off, this interface is very nice and a pleasure to use, congrats!
Are you using word2vec for these, or embeddings from another model?
I also wanted to add some flavor since it looks like many folks in this thread haven't seen something like this - it's been known since 2013 that we can do this (but it's great to remind folks especially with all the "modern" interest in NLP).
It's also known (in some circles!) that a lot of these vector arithmetic things need some tricks to really shine. For example, excluding the words already present in the query[1]. Others in this thread seem surprised at some of the biases present - there's also a long history of work on that [2,3].
[1] https://blog.esciencecenter.nl/king-man-woman-king-9a7fd2935a85
[2] https://arxiv.org/abs/1905.09866
[3] https://arxiv.org/abs/1903.03862
A lot more info about how it worked here https://blog.esciencecenter.nl/king-man-woman-king-9a7fd2935a85 and an interactive demo https://dash.gallery/dash-word-arithmetic/ (blog post about the demo https://medium.com/plotly/understanding-word-embedding-arithmetic-why-theres-no-single-answer-to-king-man-woman-cd2760e2cb7f)
Thank you for the comment!
This led me to do a bit more research, and I see indeed the queen result is in itself infact "cheating" a bit: https://blog.esciencecenter.nl/king-man-woman-king-9a7fd2935a85
#TheMoreYouKnow
I removed my blog post from medium's distribution. So it should now be freely accessible!
https://blog.esciencecenter.nl/king-man-woman-king-9a7fd2935a85
That never exactly worked for word2vec either. https://blog.esciencecenter.nl/king-man-woman-king-9a7fd2935a85