I just came across this topic and am very thankful for that. I’m an Obsidian newbie, so I may have some things wrong. But here are some random observations:
-
I’m a great believer in the Unix principle that a tool should do one thing, and only one thing, well.
-
With this in mind, IMHO this discussion conflates too many functions. The most obvious to me is (1) extracting metadata from a Zotero item to create a “source note” in Obsidian versus (2) extracting annotations from a pdf attached to a Zotero item.
-
The discussion of extracting annotations glosses over the wide variety of ways people annotate pdfs.
-
One example is a pdf that began life as a physical book or printed journal article and then was marked up with pen, pencil, and highlighter. At a later date, the printed & annotated document was scanned and converted into a pdf. On this electronic document, annotations are intelligible as such to humans but not necessarily to software that expects annotations to have a distinct digital format.
-
Another example is how some software can only handle some kinds of annotations, even when they are all done electronically. For example, for many years I have drawn straight lines in the margin, parallel to the page’s left & right edges, to identify important points in the document; I use one line for passages that are somewhat important and five parallel lines for extremely important passages. And I do this all on my iPad. Yet I’ve not found anything that can extract the passage as text and the lines as markup. For example, Matthew Meyer’s Obsidian-Zotero-Integration plugin extracts only: highlights, underlines, strikethroughs, notes, and rectangles. In turn, this limited set results from using an external pdf utility.
-
Since not all annotation styles are alike, any comparison of different software or workflows should include a list of exactly what kinds of annotations and annotation styles are supported.
-
Cobblepot mentions pandoc, and several other posters mentioned converting Obsidian documents to Word. Again, this presumes what, according to the “do 1 thing well” principle, should not be presumed. For example, I do lots of technical work, and therefore use LaTeX for writing. (Often with the LyX front end.) In this workflow, it’s actually a feature, not a bug, to use an intermediate .bib file rather than import directly from Zotero to Obsidian.
-
Closely related to this is the issue of the size of the Zotero library. I currently have over 17,000 items in mine. Sure, it’s a convenience to be able to access the entire library, but not at the cost of degrading performance. If I’m writing a paper, I may use only 100 sources to write it, with the final draft having perhaps 50; even a book will typically have at most a few hundred sources. Hence, if one uses a project-specific collection in Zotero along with the Bib(La)TeX plugin, instead of making an Obsidian import plugin search the entire library, one can restrict its searches to the project’s collection.
-
My workflow is typically (a) initialize a Zotero bibliographic entry, (b) attach a searchable pdf to it, (c) use Zotfile to store the document on my NAS drive (mounted via WebDAV), (d) use PDF Expert on my iPad to download the document from the NAS drive, (e) read, annotate, and markup the document on the iPad, (f) change the name of the annotated document by adding “(marked)” to its end, (g) upload the marked document back to the NAS drive in the same folder as the original, and (h) use Zotfile to link to this second version of the document. This allows me to keep clean and marked copies of the original document, in case, e.g., I want to use one for a class or handout.
-
This system works quite well, and I see no advantage to using a device that does not allow drawing directly on the screen. If the NAS drive sits behind a firewall and can be accessed over the Internet, the whole issue of synchronization becomes moot. Of course, if a team is working on the project, things become more complicated. But then, isn’t this what git is for?