Hacker News new | ask | show | jobs
by ahnick 2441 days ago
Sounds interesting. Can you expand on what you mean by "with embedding cosine as character replacement weights"?
1 comments

Edit distance allows for insertion, deletion and substitution. Some allow custom cost for substitution (e.g. characters closer together on a keyboard have lower substitution cost than other; for typo normalization). What I describe in my comment is to use the cosine similarity between learned character embeddings as substitution costs.