Woh! I had been a noob and wrote bash wrapping on `--dependency`

+ Zotero released their ipad app in 2023. I have for consistency stuck to it for reading/synthesis. Just makes it easy that I don't have to manage it in different places for different devices.
- Goodnotes, their annotation tools even with premium seem somewhat deficient.

I have heard tales of browser add-ons that takes care of it for you.
Seems simple to program, a bit of work to maintain.

But that's all highly confident unfounded meta cognitive guess.

So I do to pay approximate same level of attention to both of them, which is higher than that of a white-noise TV show that I play in the background when cooking/cleaning.

The main difference for me is probably that podcasts are smaller and don't have "large" plot that needs to be followed.

I maybe in a different bucket slightly. I rarely listen to audiobooks for fiction, so both audiobooks and podcasts are for content rather than pure delight (which I read fiction as e-books).

I have come to count and let count audiobooks as books read. However, I would vehemently defend someone's argument how audiobook is passive reading while text medium is active reading.

Audio and Visual processing are not the same.

I am really curious to see what semantic task does not require modeling syntax.
I would expect a non-lexical abstract semantic task, maybe something exists in the corners of theoretical semantics where we stop thinking about forms and only work with features.

I think Semantic Parsing would be a good option, although to which representation probably depends on your requirements.
arxiv.org/abs/2206.10668 seems to have both side by side.