It would be great to have a plugin (or, this might actually be part of the core functionality) that would run different distance measures to compare existing notes in order to identify and suggest:
notes that are potential duplicates (through cosine similarity);
notes that share similar keywords (through tf-idf). This option will be particularly valuable because then Obsidian itself will be suggesting connections. A graph view can then show the existing connections, the suggested connections — together or separately.
Other distance metrics can be included for slightly different outcomes.
Similar to the backlinks, this linking should be based on text similarity.
Similar texts should be displayed as the third link category in the Link Pane.
Recently Linus Lee posted an essay about auto-linking and suggestion algorithms based on text similarity. In his post he refers to his own implementation based on using n-grams. There is a link to a public deployment of the app, so everyone can play with the interface to see how it works.