hosh icon indicating copy to clipboard operation
hosh copied to clipboard

similarity as alternative way of searching

Open dfa1 opened this issue 5 years ago • 1 comments

Problem Provide a command to search by similarity, according to different well known algorithms such as Levenshtein, Cosine, Jaccard.

Solution In practice, create a tiny wrapper around this library: https://github.com/tdebatty/java-string-similarity

Normalized values would require also a new value type (decimal or float).

Benefit Being able to search not only by exact match or regexp but also by similarity.

dfa1 avatar Jun 03 '20 19:06 dfa1

Smart synonym search would also be magnificent. Some embedding like Word2Vec or more modern from GPT-2. I'd love play around with them. Look, lots of guys play with them for fun, pre-training those enormous models on some specific area. Ask them to train a model on bash scripts and comments. It will then write code for you, from the comments :-)

kosiakk avatar Jun 03 '20 21:06 kosiakk