#Semantic Space

Tools for generating and exploring word embeddings with Transformers. Collectively, the tools allow you to a) generate powerful thesauruses from electronic dictionaries, and b) conduct sophisticated searches on these thesauruses.

semantic-space can perform autoregressive searches — word alchemy. Provide the program with multiple search queries and watch the results become more precise with each additional term. Perfect for seeking that word on the tip of your tongue.

#Utilities

• embed-dictionary
• embedding-compressor
• semantic-search
• makedict

#How to use:

1. Download makedict and semantic-search to begin.
2. Download a dictionary. Supported dictionaries are the free GNU Collaborative International Dictionary of English and the stardict-packaged OED (version 2), obtained elsewhere.
3. Point makedict.py at the dictionary together with the name of a Hugging Face-hosted encoder model like bert-uncased-base or roberta-large. Wait for the process to finish.
4. Place the resulting file in its own directory and point semantic-search.py at it. Explore!