I like the new Linkpane that shows inbound and outbound links. I am curious about what the Suggested links column is based on? Does it look at word frequencies and similarities? How is it generated? Is it similar to DevonThink 's concordance?
I like the ability of ziplinks using [[]]. In Obsidian and Roam Research it is possible to just select a word or words and then just type [[ which may bring up a dialog box with options once you start typing. Would that be a possibility - to select and just type [[ and instead of substituting the selected word it would put the brackets in front and at the back of the word?
That is not possible at present. But if you select a word and press Cmd+Shift+L a stub link is parked and a completion pop-up shown allowing you to type the target noteās name (auto-complete is offered) to complete the link.
You could make a feature request to info@eastgate.com explaining your idea. To help understanding it, donāt assume the reader has seen (or has installed) the other apps you mention, i.e. describe the feature needed fully rather than in terms of looking at another app.
The suggested links pane currently looks for text similarities, using a heuristically-tweaked version of tf-idf. So, right now, itās very similar to DEVONthink.
This will likely change in the future to give additional weight to common terms discovered by natural language analysis and perhaps to use some deep learning techniques.
āAccurateā is probably not something we can hope for. Tinderbox cannot know, after all, what you might be trying to do. You might link to something that contrasts with what youāre currently writing. Or, you might link to something that makes the same argument using very different language. Or, you might link to something that explores ramifications of this argument in a different field, or at a different time.
I think the best we can hope from this sort of ālink apprenticeā is that its suggestions might be interesting and sometimes useful. As it happens, I originally got into writing the first link apprentice to demonstrate how useless such facilities would be. Theyād make stupid mistakes! Theyād miss all sorts of nuance and get caught on lexical coincidence that would not fool a child.
All true: still, my little link apprentice did keep making suggestions that seemed surprisingly clever.
Thank you, Mark. I did not know the shortcut of Cmd+ Shift+L. That is very handy. I will put in a feature request. I am somewhat surprised about the current craze of Roam Research and Obsidian and other apps that promote bi-directional links as something totally new. But we cannot deny it is taking off. And wonder if some kind of marketing/PR push about the capabilities of TB is called for. Also, because of Web 3.0 IPFS and also Tim Berners-Lee initiative TB is well-positioned to take advantage of the current wave of fascination with links in note-taking apps. How can we ride the wave?
True. But it will be more accurate than just word frequencies etc. I trust it will be more sensitive to context and certainly more interesting. And I also find that the suggestions prompt thinking deeper about potential links that I may not have considered. For that reason alone, I applaud the feature! Thanks.
I also find Suggested Links (toggle with Cmd-7) quite useful. Currently, double-clicking seems to add a link in $Text to the suggested note. I wonder if it is possible to:
visit the suggested note instead of creating a link to it. Related options: Choose to visit in the same vs. new Tab. If same Tab, return with Cmd-ā.
choose whether a link is added in $Text or for the note as-a-whole. (I do not recall the TBX terminology for these two types of outgoing links.)
Basic links are from a noteāno anchor in $Textāto another note
Text links are links that use a link anchor within $Text.
Both types are uni-directional from source to a destination note. Noteās my have a ālink typeā (which in map view shows as a label.). The labelling is incidental. The real purpose is to all you, if you wish, to add semantic markers to your link. IOW, not just create a link by indicate why that link was made.
Suggested Links (SL) has been most useful when I work in Map View where I can only see sibling notes. While Map view is wonderful for arranging notes like a whiteboard, drawing associations, loosely grouping and ordering ideas, etc., one can lose sight of notes that exist in other containers.
SL has alerted me to related notes that lie in other contexts (containers). This triggers me to go have a peek at the suggested note, particularly the notes that surround the suggested note (which is why Iād like to be able to hop to a suggested note and hop back).
In this way, I find other places (topics) where Iāve made a similar note. This has served as a good reminder of relationships between topics. For example, I was made aware of a connection between a map/topic Iām working on today with a topic I had mapped 2 years ago.
