I’m just wrapping my dissertation as well, at SFU (Canada) in Education. I’ve been a satisfied DT client for many years and recommended it to every mac-head I’ve yakked with over many years.
My research and writing workflow has evolved with new tools and tricks but I’ve been pretty efficient with the following:
DT is my main working repository; I have largely by-passed Finder and stashed-indexed all my research files in my organized folder system. I replicate-duplicate and tag as I wish for convenience;
Other gems: what has been super-helpful that not many seem to reference is to organize my work into DT ‘workspaces’ - tabbed collections that enable me to toggle back and forth with open files; you have to remember to update the space if you change it up (Workspaces are found at the bottom of the ‘Go’ menu); I also think the OCR has been really helpful.
I use DT as a drafting space all the time, and I collate excerpts into new docs, organized as I wish (and usually saved to a specific workspace)
I haven’t done much final writing in DY; I usually copy out to Word. My faculty has a Word template that seems to constrain me to using it. It’s powerful but. it’s. Word. It’s not fun to use, never has been IMO. But to adjust to comments coming in from peer reviewers and committee members I’m think I’m stuck with it.
I haven’t scripted yet with DT but that may be coming in my future. I’ll need to hit the tutorials to get comfy with that.
I have Scrivener but have not found it convenient to use with Word. I use it for other drafting, but not academic writing.
my go to citation software is Zotero. It’s worked flawlessly. With. Word.
That’s kind of it for now. I rely on DT and when I’ve had any issue the community and DT help has been stellar at setting me back on course, for which I’m very grateful! - MM
This is actually another reason I ended up using Pandoc (from Scrivener). I can compile the pandoc-flavoured Markdown to a Word document, and a particularly nice feature is you can do a Word template and compile into the Word template. I can then send it and do the Word track changes etc.
I use Bookends to generate a BibLaTeX file, and to rename attachments to match the citation. I actually do the annotating of PDFs and so forth in DEVONThink, which I find greatly preferable for that task.
Another thing I should mention is that I’ve found DTTG super helpful for PDF annotation, especially the Apple Pencil support. I eventually realised that DTTG is actually better for that annotation task than a bunch of the specialised PDF or note-taking apps.
“I have recently successfully completed my DPhil at Oxford”
That would be a dream-come-true in my next life. Now, at 75, and not with great health, I won’t be embarking on any kind of vessel to take that kind of journey in this current life. But I am so happy for you! I hope it serves you well. And, yes, if you ever get the time to share what you did (if that is appropriate), in order to produce your doctoral thesis, I would love to at least try to understand the process. All the best, and again, “my hat is off to you, sir!”
I’m having some problem with the AI. Here is a typical example:
Not long ago, I was working on a newspaper article about alcohol consumption in the Philippines. For this purpose I imported two files into DT; one is a searchable PDF containing information regarding alcohol consumption per capita (15+) (in liters of pure alcohol), and the other one is a Safari webarchive containing general information from the WHO about total population number in the Philippines, life expectancy, total expenditure on health per capita, etc.
Based on content, DT suggests I move the first file into one of the following groups (which I created a long time ago):
Barbarians in Greek comedy
Comedy (Old Greek)
Drama (Old Greek)
Literature (Old Greek)
Old Greek Literature
Literature
For some reason, unknown to me, it just seems DT is absolutely obsessed with Old Greek literature. :–)
But wait, it gets better. DT suggests I move the second file into one of the following groups:
Guns in USA
Crime (USA)
US domestic policy
Politics (USA)
Countries
USA
Pornography
It’s not until I switch to “Based on tags” that I get useful suggestions.
Most of the files I have in DT are in English, and I basically assign tags to all my files in German. Thus I can be sure that it was me who performed the tagging … and not the original author of a file. I’m wondering whether this has any influence on the accuracy of the AI.
If “Based on content” is enabled, I rarely ever get meaningful suggestions.
In the above mentioned case, the old Greek groups that were suggested are indexed groups located as folders on an external disk, but the USA groups were all created in DT and are stored on my internal disk. So it doesn’t seem the glitch can be attributed to indexing.
I have also been thinking whether my group names are the culprit. Since most of my files are in English, it would perhaps be better to choose English names for my groups too rather than having them in German. Until now, most of my groups have German names. Could this be the reason for the poor AI performance?
Thanks for your reply. Frankly, I don’t at all rely on DEVONthink’s Auto Classify … not because I have noticed the frustrations you report, but in general I trust my brain more than the computer. I keep a minimum number of “Groups” and don’t try to be cleaver about where stuff goes. I rely on the more powerful search, Smart Groups, etc. to find stuff. I view having lots of specific Groups–which act more like Apple Finder which I do use of course but do not count on for information management–and finely segregated Groups to be sort of like paper filing cabinets of decades past where indeed it was important to put the item in the right drawer to have any hope of retrieval. No longer.
I really can’t comment on your questions and observations. This seems like a topic of its own so rather than lump it in with using DEVONthink for a doctoral thesis (which is a broad question), perhaps start a new thread here on seeking guidance how to get more out of the Auto Classify functionality.
What exactly couldn’t you do with CSL style + Pandoc’s Citeproc? I found it to be extremely flexible and easy to customize using the Visual CSL Editor, the Zotero editor, and, nowadays, even by hand as the syntax is not difficult.
I am happy to say that, with a lot of tweaking and the aid of a few Lua filters, I managed to re-implement all the BibLaTeX appearance and features I needed for my own area (archaic & classical greek philosophy); that is, using the appropriate abbreviation for each classical work, multiple bibliographies to split primary and secondary sources; and citation backlinks (from each bibliography entry to all the occurrences of the ref in the text). (I also used it to create a glossary/index, but this is for another time).
The main advantage I got from this effort to transition into Pandoc’s Citeproc is that now I can rely on a single bibliography system to compile with Quarto or Scrivomatic to multiple outputs (e.g. LaTeX, PDF, html/chunkedhtml/Website, ePub, docx, and so on) directly from Scrivener. Not to mention that Citeproc is very fast in comparison with BibLaTeX’s nightmarishly long compile times!
I made the decision quite a while back, and I’ve had a look at your filters. I hadn’t realised that the Visual CSL Editor existed - that’s a game changer.
The specific things that pushed me towards BibLaTeX were the abbreviations for ancient texts and for dictionary/encyclopedia entries, the ability to automatically construct an abbreviations section at the start of the thesis and an author index at the end with almost no extra effort, the proper and systematic use of abbreviations where the style calls for them, and some things that are super fiddly to do otherwise such as resetting initial/subsequent citations per-chapter. I also had in mind to use LaTeX’s
My guess is that I can do most or perhaps even all of this with a combo of the style editor and pandoc filters. But by the time I became aware of that I was pretty invested in the LaTeX approach and focused on the writing and editing.
I am pretty keen to explore this, because being able to use pandoc citeproc instead of BibLaTeX means that I can target so many different output formats.
I am also keen to learn more about Lua filters myself. Let’s share notes on this at some point. I still miss the \citefield functions (e.g.\citeauthor and \citetitle). I definitively can get around without these, but it seems silly not to have it, because it is totally doable with a Lua filter. Just no one took the time to write one yet. (I have been trying, but not close yet).
That part is pretty easy to achieve. Just add the abbreviations to a particular BibTeX field, such as shorttitle, note, annote, then edit the Oxford CSL style to give preference to it whenever it is present.
I use a filter to take the section with the title # Abbreviations and place it in a specific metadata field so that the latex template can place in correctly at the beginning of the document. This means that I have to do it “by hand” (using Bookends, that is), which I don’t mind because I was not entirely happy with the BibLaTeX Philosophy-Modern printing the whole entry alongside the abbreviation.
While we should consider this feature still missing, it would not be hard to properly recreate it using a Lua filter to collect all the cited references with a shorthand or shorttitle field. The same goes for all of the other things, such as the author index and the resetting initial/subsequent citations per-chapter. You definitively won’t find anything quite ready yet to accomplish this. But, with some effort…it is doable.
A lot can be achieved already with multibib and citations-backlinks. As you can see in the image, when you hover the inline citation, the pop-up provides you with the link to all the other occurrences of the same reference in the document, making it easy to jump-around. It could be just a matter of making an bibliography with authors names and indexing them by citing them separately: @Aquinas states about @Aristotle* @Cat I 1 1a01 that...
Well the problem with Bookends it seems to me is that one can’t search it with Finder or Houdah Spot. It is not a big deal just to go into it and search in the app to be realistic. I like everything to show up with Houdah spot and DEVONthink 3 stuff does, as does Ulysses. The PDF’s stored in Documents by default, are searchable as it happens. Which is why I keep some in Bookends now.
Fact is I have a good idea in my own mind what I have in Bookends, that is the truth, I don’t have thousands of citations, most of my ‘piles’ are just in DEVONthink 3 as PDFs which are easily searchable and have always the citation data with them in one form or another.
But what I used to do, I no longer bother, was create a PDF of everything that was in Bookends. A full bibliography if you prefer, in my default format, then I store that in DEVONthink 3 and it is searchable as a PDF. So suppose I look for “Russell” “Leibniz” in either DEVONthink 3 or Houdah Spot I will find the corresponding citations: then I know I have it in Bookends really and can go there if I need to insert it into a doc or whatever.
It is File > Export References > Hits > Bibliography Format
You do have to update it now and again, it isn’t quite current depending on how one uses it. I used to do it every few weeks if I remembered, deleting the previous one. You can also, as you know, keep importing them using DEVONthink 3 as single files. I do that sometimes if I am going to annotate. I have no coherent system really.
I am exactly like you. I do use the classify thing, but only to look at what is in some way similar. I can’t explain that fully, I like doing it, but I don’t actually ever organize according to it.
It’s always interesting to see other academics describe their use of DT. For me, DevonThink is one of the two “essential” productivity apps I started to rely on during my Bachelors degree. I have since finished my Masters and PhD degrees, and I’m still using it daily as a university faculty. I’ll probably continue to use it until I retire… (FYI the second “essential” app which received same consistent use would be OmniFocus).
Thanks for the explanation. I guess I haven’t felt the need for this because most of my BE items have PDFs attached, and they will contain the metadata and be searchable by Finder and, if the relevant folders are indexed, by DT3.
I found the same in the end. The thing is that most PDFs how have the metadata and all you need for citations really within the pdf itself. I don’t use that method any more really. I use BookEnds really so that the citation is properly formated if I copy and paste. Though I stil find myself endlessly fiddling with citation formats. Any answers to that one!?
@Bernardo_V oh yes — I remember reading that initial post and getting very interested before telling myself off and going back to the writing and editing!
I’ll find some time this week to download that sample project and try it out. It certainly looks like you’ve already done most of the work. Quarto is a bit of a game changer too I think, and seems to have quite a bit of momentum behind it which is a great sign.
I’m not sure if it’s a better way, and I’m not writing a dissertation or anything, but my workflow is to dump all of my citation information into Bookends, and since it creates a static UniqueID for every source, I use that number everywhere in my related notes in DT.
Aside: I’ve fiddled with Bookends citation templates, but since my use for it is genealogy, the information I need to include for each source usually has too many details/variables, so I have three custom fields in Bookends: footnote citation, first citation, and short citation that I fill in manually. I looked at using EndNote, but EndNote generates dynamic IDs.
I save all the digital copies of a source in a Bookends Attachment folder in OneDrive with the beginning of the filename updated to match the Bookends UniqueID and then attach the file to the Bookends record.
That attachment folder is indexed in DT. Once a file shows up in my indexed DT folder, I add the DT link to a custom field in Bookends, and I add the Bookends UniqueID to a custom metadata field in DT, and then in DT I create an annotation for it from a custom template, which generates a link to the source in Bookends from the custom metadata field.
My default annotation template pulls in the BookendsID from the custom metadata field, so as long as I remember to add the Bookends UniqueID before creating the annotation, DT does a bunch of magic for me. Part of my template is a section on source information with a reminder that “all of the info should be in Bookends not the annotation but here’s the space for it if you’re being lazy”.
When I want to review all my notes on a source, all I have to do is search for the ID. (I don’t specify the metadata field in my search, since that is only for the actual source, but instead do just a plain text search.)
The downside of using the custom annotation template is if I’m in a hurry and don’t delete the parts of it I don’t need, DT’s See Also thinks a whole bunch of files are related when they really aren’t. But that’s on me.
Thanks, I only really have very standard citations. Even then it can get fussy I found, nice solutions. There is, I note, a lot of scope for invention using these apps and that is one thing I really like about them. I hope Jon the Bookends guy sees this one.
It’s interesting that you mention complex genealogical source citations. I am writing a book (cited via Chicago style endnotes) that depends heavily on government records and am finding similar issues with idiosyncratic citations. The US National Archives, for example, often preserves agency file structures in their entirety. Census and other genealogical information similarly doesn’t fall neatly into a collection/box/file system
All of my annotations are in DEVONThink, so if I really am stuck writing custom manual citations for ~80% of my source documents, is it even worth adding a citation manager?
