An open-source version of Semantle/Cémantix developed in React and using a database of Italian words.
We use pre-trained Italian word embeddings from fastText.
The list of 1k common words from Napolux is used for acceptable solutions. The list of words in Debian's Italian spell checker is used to screen for valid words.
The Jupyter notebook in backend/Semantle_generator creates .json files with word embeddings. The files should be copied in the public folder of the frontend. All other computations occur in the frontend.