Emoji Math
Because why not? I put a minimal effort into this project, so please have low expectations.
pip install [email protected]+git://github.com/whitead/emoji-math.git
Usage
emoji-math
computes the given python expression and returns either the value or the nearest 5 emojis as measured by cosine similarity.
>emoji-math 👑 - 🚹 + 🚺
Best Matches:
👑-🚹+🚺 = 👸
👑-🚹+🚺 = 👑
👑-🚹+🚺 = 🤴
>emoji-math 🚹 @ 🚺
🚹 @ 🚺 = 0.32784234338655205
>emoji-math np.sin(🏰)
Best Matches:
np.sin(🏰) = 🏯
np.sin(🏰) = 🏰
np.sin(🏰) = 👸
>emoji-math 🚹 + 🚺
Best Matches:
🚹+🚺 = 🚻
🚹+🚺 = 🚺
🚹+🚺 = 🚹
Options
You can choose between using the Euclidean distance or cosine similarity for reporting results. Cosine works better for multiplication/division because it only cares about direction. Default is Euclidean.
>emoji-math --cosine 🚹 + 🚺
Best Matches:
🚹+🚺 = 👚
🚹+🚺 = 🚺
🚹+🚺 = 🚹
Google Colab
Credit
Made by @andrewwhite01
Vector embeddings from emoji2vec as described in
@misc{eisner2016emoji2vec,
title={emoji2vec: Learning Emoji Representations from their Description},
author={Ben Eisner and Tim Rocktäschel and Isabelle Augenstein and Matko Bošnjak and Sebastian Riedel},
year={2016},
eprint={1609.08359},
archivePrefix={arXiv},
primaryClass={cs.CL}
}