Hmm - that’s frustrating.
You are putting just “ 21102021” in the first prompt (and nothing else) and then “Martin bought shares in company” in the next prompt?
I hope this isn’t off topic. If you want a timeline from Devonthink, you can use Create Metadata Overview to do interesting things with other apps.
For instance, I have custom metadata for Parent, Participant, Observer, Arc, and Location.
Those values correspond to fields in Aeon Timeline.
If I set those values on entries in DT, highlight what I’m interested in, and Create Metadata Overview, I get a tab delimited file I can directly load into Aeon Timeline, no modification necessary. I get events with dates and the matrix of relationships gets filled out, too.
One of my favorite tricks, even if my use is infrequent.
I imagine importing into other timeline utilities would be about as easy.
1 Like
Duh! You’re right; in my overexcitement, I was putting in the event as well as the date.
Oh well.
Thanks. That’s a relief as I must say it is nice and straightforward to enter the data for an individual event with this method.
Thanks, Amontillado.
It’s an interesting idea you suggest. I haven’t explored Create Metadata Overview and I must confess every time I try out AEON Timeline I get defeated by the complexity.
It happens 
.
Glad it is working for you.
If you start Aeon without using a template, and if you delete the very few extra relationships and entities it has by default, you can build what you need as you add to your timeline.
I’ve wanted to use Aeon for planning fiction. What I’ve ended up using it for is correspondence tracking.
I write a letter or email, I get replies, then reply to those, get more replies, include more recipients, etc.
A letter or email I write in a particular campaign is an event at a date and time. It has relationships for sender (me) and recipients. The event gets a link to the word processing document or email message in Devonthink, and also gets linked into a mind map view for that conversation.
Phone calls get the same treatment.
This gets me a timeline view with events grouped vertically by conversation, displayed horizontally by time. The mind map view gives me a view of the steps the dialog took.
The timeline file and all the documents and notes go in Devonthink.
If Devonthink had timeline and mindmap views, I would do the whole thing in Devonthink. Sigh… It’s all I want for Christmas, I swear.
3 Likes
Have you thought about uploading your document to Claude and asking AI to create a timeline?
I am guessing that might be similarly useful as Aeon Timeline but much quicker
An interesting idea. I’ll have to look into Claude. I’m a little AI-hostile, but that’s mostly based on generative AI and because I’m a stick in the mud. Searching and categorizing by AI is OK.
What I get out of Aeon is a timeline of, say, 75 documents, emails, and phone calls in both a timeline and hierarchical view.
I should have also mentioned when I add new correspondence I do that in Aeon’s mind map view. That sets the hierarchy. The timeline view is then automatic.
4 Likes
Agreed that AI is helpful for summarizing existing data, not for generating new content.
That said - I am not an attorney but our workflows may be similar as I am a testifying expert. I routinely use Claude to create deposition summaries and summaries/timelines of medical records; the results are impressive though it takes a bit of effort at first to deterimine the prompt that will achieve your desired output.
1 Like
Mm, when an AI summarizes something, I can never be sure what the machine has left out because it didn’t “think” it was important. If I don’t know the original data, this summary is almost worthless. And if I know the original data, I don’t need a summary.
I work in a profession where I am very often dependent on summaries. But I have them given to me verbally by people who know the files very well. If I don’t understand something or am missing information, I ask directly. In this way, I get summaries that are really substantial.
1 Like
What profession? Law? Engineering? Medicine? Something else?
AI is clearly helpful in all the above - albeit it is essential that it provide links/references to the sources. And it is still necessary to review the orginal source yourself.
That’s just as true if a human summarizes it as if AI does.
The test is whether AI adds something to human-only review; undoubtedly it does. Surely it does not replace it - I agree there.
I am not pursuing AI as I have so many other real interests that I have no time, really. I guess it will be done to me sooner than done by me.
That being said a “real” author asked me to review/edit a piece he planned to publish on numerous internet locations. I made a number of comments/changes (Word’s Revision feature, of course, not DEVONthink! but the documents reside in my DEVONthink). I was uncomfortable with many of the complex sentences, word choices, etc. but had nothing really to suggest other than I thought it too complicated. Using Word’s Readability tools, it scored low on readability and 14.4 years of education req’d! I shows that to the author until then who had been unaware, despite using Word for decades, of that Readability check feature. Oh well.
Point of my post: without me suggesting or knowing, he posted the draft to ChatGPT and asked the AI to “rewrite this draft to simplify in the style of [author name]”.
Gosh darn. What came back was terrific. In my role as editor I had no quibbles with anything that ChatGPT wrote. Word Readability scores much better.
Amazing and scary.
2 Likes
Agreed - you would not want ChatGPT to author original ideas in a professional brief.
But it is extremely helpful for spelling, grammar, and style.
Moreover while you would not want it to author your content it can be very englightening to take your own report and ask ChatGPT (or other AI) “Rebut this report and point out errors/omissions/criticisms”
2 Likes
@stephenjw @dtlow I have been using Stephen’s script to create a folder of chronology snippets (with some notes and links to other documents related to the event). I’ve also used the Table of Contents function in DT to create a consolidated rtf document of events.
This works well for me as I can click on the entry for a particular event, go to the note and see the original document reference and, if necessary, click on that to go to the document itself.
My problem is I need to share the chronology with an external, non-Mac person but, for that to be useful for them, I need to include at least the document reference and there are over 200 events.
Has anyone succeeded in extracting that additional information as well or (recognising this is a fairly specific issue) have suggestions of where to look for possible solutions?
Hi @wrothnie. I have sometimes shared the chronology but haven’t looked at creating one with working links to the source documents that can be actioned outside of DT.
If you want the links to work for the outside user, you could put together a composite PDF with the chronology and the source documents, but I think you would then have to manually hyperlink the chronology entries to the relevant page of the PDF.
If it is just a document reference you are after, you could use a custom meta data field for the source reference and have the script reproduce this item name in the chronology (eg “18.12.2020 - Plaintiff consults Dr Smith - Source Report of Dr Smith dated 24.3.2024”).
It would likely mean manually inputting the document reference into the custom field (if the documents are all in a single PDF and the page numbering is the same as the page count of the PDF you might be able to use the Page Count placeholder to automate or partly-automate this).
Alternatively, if the document reference can be the name you have given the document (eg the name is “Report of Dr Smith dated 24.3.2024”), you could perhaps use the document name and page count to display “Source Report of Dr Smith dated 24.3.2024, page” and the page count).
The page count might not reflect the pagination of the source document if its page numbering doesn’t start from “1” (as sometime happens if there is a cover page or blank pages in the scan).
If you’re going to have 3 pieces of information for each event (date, event and source), it might be better to present it in a table. I’ve used Word to convert to table but haven’t tried using sheets in DT for this purpose.
Thanks.
I should say, if I haven’t already, I love easy it is to make the individual notes.
I came up with a solution to my problem using Hazel. I had read that Hazel can be used to rename documents like bank statements to include the date. I used this methodology with a custom list (I think that’s what they’re called) to scan each chronology note for the discovery number and then renamed the file to include that (separated from the date - event by “ - “.
Then I used the DTPro tool Create Table of Contents on the folder as an rtf.
In the resulting, native rtf file I replaced each “ - “ with a tab and pasted the whole lot into an Excel spreadsheet so that each of date, Event and Discovery Number were in three columns.
Unlike the rtf, the Excel doc does not have the link back to the original doc. However, I couldn’t get the rtf file to create a three column table - there were three columns but the data all went into the same column.
(The link back to the original document in DTPro would only help me of course, not the people I have to share the chronology with.)
Sounds like a good workaround.
But it is extremely helpful for spelling, grammar, and style.
Which in my option is why AI is of no interest to me. I don’t want my style altered / changed. I don’t want commas where I don’t want them. And as another has said I don’t want to have to read a long document twice to make sure the AI version hasn’t left out something I would find important.
3 Likes
Especially the last part is also one of the – in my opinion main – counter arguments to AI in qualitative research, at least in analysing and coding long interviews
1 Like
That sounds to me like cutting off your nose to spite your face.
Do you ever find a human editor or colleague to be helpful? It’s pretty rare for that to not be the case.
I never use AI to write nor even for grammar. But I do often find it useful to ask AI questions about both my own writing and documents I receive from others such as:
- Is this document likely to persuasive? Why?
- Which of these documents gives a stronger argument?
- What is missing from this document?
- What inconsistencies are in this document?
Often its suggestions are templated or otherwise not helpful. But it’s pretty rare for AI to not yield at least one useful observation or suggestion. Perhaps I will change my report; perhaps I will just make a note in my mind where the strengths and weaknesses are in my own work or in the work of others. That’s pretty useful to me in any case.
1 Like