Hello,
Strange things in the world today. For me with DEVONthink now too. With all the PDF+Text files in DEVONthink I’ve tried, the Find Inspector search does not show any occurrences of any text string I search for.
When I export any PDF+Text file from DT and open it in Preview, all Find operations there work just fine.
I’m using DT 4.2.2 and MacOS 26.3.1. I didn’t find any other posts related to this problem, except for a long time ago (2004).
Thanks,
Francis
Can you post a screenshot please? And what do you get when you convert the PDF to text?
Here is the screenshot:
Here is the text (I used File | Export | As Text) 
:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
Path-LLM: A Multi-Modal Path Representation Learning by
Aligning and Fusing with Large Language Models
Primary
Yanta N Rd
ABSTRACT
Youyi E Rd Youyi E Rd
The advancement of intelligent transportation systems has led to
Youyi W Rd
a growing demand for accurate path representations, which are
essential for tasks such as travel time estimation, path ranking,
and trajectory analysis. However, traditional path representation
learning (PRL) methods often focus solely on single-modal road net-
Options
9 min
3.9 km
12min
3.8 km
Xi’an Museum
Track
1.2km
Primary
0.7km 1.2km
Yongning Rd
0.6km
Cehui W Rd
Jianshe Rd
Cehui E Rd
Tiyuchang N Rd
0.7km
12 min
Wenyi S Rd
1.2km
Airforce Hospital
work data, overlooking important physical and regional factors that
Zhuque Ave
influence real-world traffic dynamics. To overcome this limitation,
we introduce Path-LLM, a multi-modal path representation learn-
Textual Road
Road Name;
Road Type;
Road Length.
(a) Two multimodal paths
Holiday Hotel
Changan N Rd
0.7km 1.9km
Trunk
2nd Ring South Rd
9 min
ing model that integrates large language models (LLMs) into PRL.
Our approach leverages LLMs to interpret both topological and
textual data, enabling robust multi-modal path representations. To
effectively align and merge these modalities, we propose TPalign, a
contrastive learning-based pretraining strategy that ensures align-
ment within the embedding space. We then present TPfusion, a
multimodal fusion module that dynamically adjusts the weight of
Yanta N Rd
And another example from a PDF I created:
Screenshot:
Exported Text File (first lines)
Chapter 3 cognitive types and nuclear content.
3.1 from Kant to cognitivism
If Kant had considered scheme out early enough, Pierce said, they would’ve overgrown
his whole work(WR 5:258-59). In the previous chapter, I suggested it was precisely the
problem of schematic ism that obliged to undertake a change of direction in the third
critique. But we might say something more: if we were to reconsider the problem of
Kantian, schematism, much of the semantics of this century, from the truth – functional to
the structural variety, would find itself in difficulty. And this is what has happened in the
area usually referred to as " cognitive studies "
Drag that PDF document to your desktop.
Control-Click > Compress… it.
Upload the PDF here or hold the Option key in DEVONthink and select Help > Report Bug. Attach the ZIP.
First, you must press return after entering a string in the search field. That starts the search. Did you do that?
2nd, I downloaded the original PDF from the ACM (https://dl.acm.org/doi/epdf/10.1145/3696410.3714744). Converting that to pure text gave a very different result from yours. Which leads me to the question: How did you obtain this PDF? Did you run OCR on it after downloading?
Yes, I hit return after entering the search string.
I got the PDF file (the journal article) online. I don’t remember if it was ACM or a different outlet/publisher.
In any case the second PDF is my own creation and search also doesn’t work with it. I’ll zip up the PDFs and submit them with a bug report.
Since search in a PDF file does not work with any PDFs I’ve imported to DT, I assume the files are not the problem, but something else is.
Francis
Here are two sample files
SamplePdfFiles-fh.zip (1.5 MB)
On these files, using DEVONthink’s Search (CMD-F) and using words I can see in the documents, works here. Can’t think what is stopping this working for you.
Thanks. An important result.
That suggests I might simply need to reinstall DEVONthink. Unless you think otherwise, I’ll try that tonight.
Francis
Reinstalling software rarely solves a problem with it. IMO, it’s better to figure out what is wrong exactly.
Did you open that PDF in another app and annotate it there, for example? I suppose that the original text layer got somehow destroyed. Which, in my experience, does not happen all by itself by just importing the document into DT.
Are those two exported directly from DT?
Reinstall usually is not a solution to a software issue. I would wait till @BLUEFROG gives advice. I think he already suggested you put in a support request.
Well, reinstalling software has worked many times in the past for me.
No, the PDF were copied by their creation/download directly to DT.
What you have written before suggests the text layer in the PDFs is fine, right? I mean you had no problems with the Find operation on the files in DT. This observation brings me to believe it is a software problem
Well, then do it I guess. Your machine.
Before undertaking drastic steps like that, you could try to restart your Mac. Which also “has worked many times in the past” for many people.
To which question does this answer refer?
Are those two exported directly from DT?
Did you open that PDF in another app and annotate it there?
Did you ran OCR on the PDF?
The text layer in my PDF is fine. And that one is the one I copied directly from the ACM site. The text layer in your PDF is far from “fine”. But I have no idea (nor did you clearly answer any question related to that) where the PDFs originated that you posted here, what you did or did not do with the PDFs you imported into DT etc.
I was answering your question. Your follow-up was:
Are those two exported directly from DT?
Did you open that PDF in another app and annotate it there?
Did you ran OCR on the PDF?
My answers
The text layer in my PDF is fine. And that one is the one I copied directly from the ACM site. The text layer in your PDF is far from “fine”.
Interesting. What software produces the text layer? I took it as a hypothesis to experimentally test and ran OCR in DT on it. In spite of the warning, “Are you sure you want to convert this searchable PDF again?” I proceeded and DT created a new PDF file. I tried the Find operation. The result was the same as with the earlier file. I attach the file for your consideration. Does the result of this experiment mean DT is not created a “proper” text layer?
Thanks for the help,
Francis
Eco2000-Ch3-1-OCRed.pdf (750.4 KB)
The original PDF creator has produced the text layer in this case. The original document was LaTeX, and ACM (or software provided by them) converted that to PDF. Which creates automatically a text layer, as most software does nowadays.
No. All that means that in your case something bizarre happens. I think @BLUEFROG is better positioned to solve this problem.
In DEVONthink, hold the Option key and select Help > Report Bug.
I have OCR turned on as part of my import process. Do you have that set? Maybe OCR the document and try again?
The issue turned out to be an out-of-date index. A quick local backup and a File > Rebuild Database got things back in shape.
