mirror of
https://github.com/jackyzha0/quartz.git
synced 2025-12-24 21:34:06 -06:00
16 lines
1.4 KiB
Markdown
16 lines
1.4 KiB
Markdown
---
|
|
title: "How AI-Powered Link Suggestions Supercharge Note-taking"
|
|
date: 2022-06-24
|
|
lastmod: 2022-06-24
|
|
---
|
|
An essential part of connecting notes is to [[posts/learn-zettelkasten-by-example-3-new-note-or-not|find a note that needs to be connected]]. But there are many times when I forget I've made a link. Thus, I either spend a lot of time finding the relevant note or create a note with a title similar to a note I've already made. So, when creating [Fleeting Notes](https://fleetingnotes.app) (A wiki for quick notes), I felt it was paramount to make finding past notes fast and efficient.
|
|
|
|
The solution I've settled on is to order the links by relevance when link suggestions are triggered (i.e. the overlay that shows when pressing `[[` ).
|
|
|
|
Essentially, the text that is written + every link is fed into a machine learning model. Then, the model spits out the most relevant links for the given text. If you're curious about the details of this machine learning model, see the [library](https://www.npmjs.com/package/@tensorflow-models/universal-sentence-encoder) I used to accomplish this.
|
|
|
|
Here is a demo of the sentence similarity ranking in action:
|
|
{{< youtube 3ySN4o0o7Hw >}}
|
|
|
|
|
|
Fleeting Notes is a free application, but AI-powered link suggestions is a paid feature in Fleeting Notes. If you want to try it out, use the promo code `AILINKS` for a 1 month free trial (Note: This coupon code expires July 31st, 2022). |