I am experimenting with Tinderbox as my zettelkasten. I have imported several Readwise highlight notes from my Obsidian vault. When I opened a note in Tinderbox and went to “Get Info” for the note, the dialog box opened and pointed to attributes and AI. What is the “Sentiment” and “Sentiments” values? Are these embedded codes in the note that identify the semantic meaning for the note that is used for relating to similar notes? Is there an AI function built into Tinderbox?
It’s part of the “tagger” process. More info here at aTbRef. This is not AI in the sense of a GPT process.
Thank you. After reading the reference you pointed me to, I think that that is a functionality that is several steps beyond my understanding of Tinderbox at this point. For now, I will just let Tinderbox do its magic on its own.
1 Like
This aTbRef article is also pertinent: Natural Language Processing. So, there is no build-it AI in terms of a “please do my work for me” AI button. The built-in AI attributes are an experiment in term (aka tag, keyword) extraction from Text, generally $Text.
Some users have been integrating AI (OPenAI, Claude, etc.) by making calls out from Tinderbox to such services. It necessarily involves code, but implemented correctly you don’t need to understand that code once all is working.
Some threads on AI integration by Tinderbox users:
2 Likes
Tinderbox automatically scans notes to look for a few things that it can extract from language with some reliability:
- names
- places
- companies and similar “entities”
The sentiment attributes look at each paragraph, and try to assess whether that paragraph is positive (+1) or negative {-1) in sentiment, and also to gauge the entire note. It’s the sort of thing that most people just don’t need at all, but saves the day when you do need it.
1 Like
Noting the last, is there a non-spammy/SEO desciption of what real-world code sentiment analysis. I’ve trust issues about it. I idea makes sense, the execution less so. I understand why people want it (insights, for free) but that is another reason to be cautious about how algorithms have validity here.
This isn’t to gainsay the last above. If you know what it is and that you need it, you need it, you do. But, as with ‘AI’,. even if ‘sentiment/AI-curious’, we probably don’t. This is where ‘fear of missing out’ (FOMO) plays badly: the app is a toolbox, we don’t all need every tool.
One example would be a set of notes, gathered from magazines and newspapers, describing a performance or a social event in some past historical era. It would be interesting to know, when examining these accounts, whether the writers were sympathetic or hostile to the event. You could do this yourself for ancient sources, where you have at most four or five; for a 19th century event, you might have dozens or hundreds.
2 Likes
Thanks for a useful explanation of meaningful use of this feature. I’ll borrow that if I may. I think social media research (‘twitterology’) gave the overall concept of sentiment analysis a bad name.
I think this example would be useful in aTbRef to help explain the presence of the feature.
1 Like
As a first step, I’ve added a bit to this aTbRef note, Sentiment Analysis, based on @eastgate’s helpful description above.
1 Like