Show HN: A bookmark manager that uses TF-IDF vectors to group similar bookmarks
lxi.aiHello!, Everyone I know has a bookmark graveyard of bits of interesting information we wanted to come back to later but added into a general folder that ends up being a mess.
So I did what any good ML student would do and built a simple ML system (using TF-IDF vectors) that will recommend you similar bookmarks.
I use Obsidian.md as my second brain but needed a space to put the interesting links without spending the time to create a new note incorporate it into my organizational system. I think lxi.ai does a good job of being the middle ground for holding the interesting links for me to come back to when I really want to wrestle with the ideas. Hope some people like it too!
website: https://lxi.ai/
chrome extension: https://chrome.google.com/webstore/detail/lxiai/dkhfhajlhegd...
firefox addon: https://addons.mozilla.org/addon/lxi-ai/
Looks very good, will certainly give it a try.
I am assuming you built this with Python, right? Any chance you can share the approach? Would love to try building something for the web with the use of Tensorflow.
Hey thanks for giving it a try! And yes, built it with python - specifically I'm using a modified version of gensim's TF-IDF model implementation. If you haven't heard of TF-IDF before give http://www.tfidf.com/ a read through.
It might be overkill to use tensorflow for TF-IDF, but go for it!
With lxi.ai, I use the top 10 most important keywords for the actual comparison along with some other tricks to make sure the model can scale. Might be worth a blog post or something.
Thanks.
Found a small bug. I've added a few bookmarks and when I press the Random Bookmark link I get an Internal Server Error at this url (https://lxi.ai/bookmarks/achieve):
``` The server encountered an internal error and was unable to complete your request. Either the server is overloaded or there is an error in the application. ```
Ah yes that makes sense - it'll be fixed within the hour.
- How did build the graph view, looks very good? - Did you use Python to build the Chrome addon, too? If so, do you mind sharing a resource where I can see how to do that, cause when I was looking, did not find anything.
Thanks a ton in advance.
The graph is built with D3.js and some minor JS/CSS changes to create the hover effects.
The extensions are built with simple html/js/css - they're basically mini webpages with special permissions. For example, I use an activeTab api provided by the browser to read the page text when the user clicks a button.
I recommend google's getting started docs: https://developer.chrome.com/docs/extensions/mv3/getstarted/
Or find a youtube video and follow along!
That’s pretty cool, going to give it a try.