# OCR-language information?

**URL:** <https://discourse.devontechnologies.com/t/ocr-language-information/67765>\
**Category:** DEVONthink\
**Created:** [November 14, 2021, 6:42pm UTC](https://discourse.devontechnologies.com/t/ocr-language-information/67765 "2021-11-14T18:42:09Z")\
**Posts on this page:** 12\
**Page:** 1

<div class="post-metadata">

**Author:** ![WhyO\_74](https://discourse.devontechnologies.com/letter_avatar_proxy/v4/letter/w/97f17d/32.png) [@WhyO\_74](https://discourse.devontechnologies.com/u/WhyO_74)\
**Post date:** [November 14, 2021, 6:42pm UTC](https://discourse.devontechnologies.com/t/ocr-language-information/67765/1 "2021-11-14T18:42:09Z")

</div>

Hello there!  
I bought my first document scanner in 2013 (Canon P-215), and have also used a flatbed scanner with various scan-programs on my Mac. Since 2015, Devonthink has been my digital library and archive, and my Private-database alone have 15.000 documents. Additional databases for scanned magazines, work-related, emails…

Those days, received and downloaded pdf’s were not often OCR’ed at all. And because I’m Norwegian, OCR’ing in Norwegian has only been supported by the ABBYY engine, and the P-215 (no-ABBYY software) has saved my scans with an English text layer.

I have also scanned documents from my studies in Germany, aside German and Swedish magazines, so you can tell I have a multi-language chaos in my archive 🤔

So - is there a way to find the OCR-language in a pdf, and can this be used to apply a Smart Rule to re-OCR relevant files with correct language? Devonthink has done this usually perfectly for a long time when converting documents.

Or have maybe my pdfs been rescanned and corrected in the updating process from DTPO2 \> DT3 Pro when the OCR engine was updated? I know Devonthink uses magic all the time 😃

---

<div class="post-metadata">

**Author:** ![jerwin](https://discourse.devontechnologies.com/letter_avatar_proxy/v4/letter/j/e495f1/32.png) [@jerwin](https://discourse.devontechnologies.com/u/jerwin)\
**Post date:** [November 14, 2021, 10:37pm UTC](https://discourse.devontechnologies.com/t/ocr-language-information/67765/2 "2021-11-14T22:37:36Z")

</div>

Select the files, and Select Tools|Create Meta Data Overview.

There should be a Language Field… Then re OCR the ones that aren’t Norwegian.

---

<div class="post-metadata">

**Author:** ![BLUEFROG](https://discourse.devontechnologies.com/user_avatar/discourse.devontechnologies.com/bluefrog/32/135_2.png) [@BLUEFROG](https://discourse.devontechnologies.com/u/BLUEFROG)\
**Post date:** [November 14, 2021, 10:37pm UTC](https://discourse.devontechnologies.com/t/ocr-language-information/67765/3 "2021-11-14T22:37:47Z")

</div>

You can’t OCR to a specific language via smart rules or scripting.  
If you change the primary language in **Preferences \> OCR** to Norwegian you should be able to re-OCR the files.

Can you ZIP and post an example Norwegian document?

---

<div class="post-metadata">

**Author:** ![WhyO\_74](https://discourse.devontechnologies.com/letter_avatar_proxy/v4/letter/w/97f17d/32.png) [@WhyO\_74](https://discourse.devontechnologies.com/u/WhyO_74)\
**Post date:** [November 15, 2021, 8:38am UTC](https://discourse.devontechnologies.com/t/ocr-language-information/67765/4 "2021-11-15T08:38:45Z")

</div>

My preset Primary language has always been Norwegian in Devonthink, with secondary languages chosen to those of my article variations (Swedish, Danish, German, English). DT does a perfect job in detecting the right one, as far as I have seen.

My problem is that many files have been imported into DT without OCRing/delete originals, and I know by random text copying, that the files are “bad Norwegian”, especially those scanned with the Canon.

But @jerwin seems to help me further down the road 🙂

---

<div class="post-metadata">

**Author:** ![WhyO\_74](https://discourse.devontechnologies.com/letter_avatar_proxy/v4/letter/w/97f17d/32.png) [@WhyO\_74](https://discourse.devontechnologies.com/u/WhyO_74)\
**Post date:** [November 15, 2021, 8:41am UTC](https://discourse.devontechnologies.com/t/ocr-language-information/67765/5 "2021-11-15T08:41:05Z")

</div>

This helps me a lot, thank you!

---

<div class="post-metadata">

**Author:** ![WhyO\_74](https://discourse.devontechnologies.com/letter_avatar_proxy/v4/letter/w/97f17d/32.png) [@WhyO\_74](https://discourse.devontechnologies.com/u/WhyO_74)\
**Post date:** [November 15, 2021, 8:43am UTC](https://discourse.devontechnologies.com/t/ocr-language-information/67765/6 "2021-11-15T08:43:12Z")

</div>

![Skjermbilde 2021-11-15 kl. 09.24.49](https://devontech-discourse.s3.dualstack.us-east-1.amazonaws.com/uploads/original/3X/5/a/5ad9064fdf753c544bcd6349d6dbfc3d48af678f.png)

---

<div class="post-metadata">

**Author:** ![WhyO\_74](https://discourse.devontechnologies.com/letter_avatar_proxy/v4/letter/w/97f17d/32.png) [@WhyO\_74](https://discourse.devontechnologies.com/u/WhyO_74)\
**Post date:** [November 15, 2021, 9:24am UTC](https://discourse.devontechnologies.com/t/ocr-language-information/67765/7 "2021-11-15T09:24:25Z")

</div>

Some comments/fun fact:

1. Sorting by language did not help me much, because those Canon-pdfs are gibberish Norwegian (OCR-engine library not Norwegian, I suppose).  
This was tested on the marked file (OCR \> to searchable PDF), and the (renamed) new PDF with DT was both much smaller AND perfect.

2. I knew from opening pdfs in Preview, that the Info pane showed both Content Producer (Canon) and PDF-Producer (FineReader OCR Pro, from earlier rescans before DT).  
I looked and found a Creator metadata field! And now I can easily sort by the older “OCR-versions”

3. Metadata info is visible only when the Group is opened, so I thought, these metadata could be used in a Smart Group (“English pdfs”). But I realize that the important metadata is Creator, and will create a Smart Group “Bad OCR” (Creator matches Canon) to rescan these files. I just have to remember to activate the “Move original to the trash” in the OCR-settings first 😃 (normally deactivated)

4. I have also noticed a big size difference between OCRs from older DTPO 2.x and the new ABBYY FineReader Engine 12, so I probably just have to re-OCR every pdf older than those…

Hope someone else out there can use this information. Devonthink shows new strengths and solutions again 💪 👏 👌

---

<div class="post-metadata">

**Author:** ![jerwin](https://discourse.devontechnologies.com/letter_avatar_proxy/v4/letter/j/e495f1/32.png) [@jerwin](https://discourse.devontechnologies.com/u/jerwin)\
**Post date:** [November 19, 2021, 5:33am UTC](https://discourse.devontechnologies.com/t/ocr-language-information/67765/8 "2021-11-19T05:33:58Z")

</div>

Out of curiosity, is Devonthink’s version of Finereader compiled for Apple Silicon? I recently used the OCR to read some screenshots, and it seemed faster than my Finereader PDF installation.

---

<div class="post-metadata">

**Author:** ![cgrunenberg](https://discourse.devontechnologies.com/user_avatar/discourse.devontechnologies.com/cgrunenberg/32/7172_2.png) [@cgrunenberg](https://discourse.devontechnologies.com/u/cgrunenberg)\
**Post date:** [November 19, 2021, 8:41am UTC](https://discourse.devontechnologies.com/t/ocr-language-information/67765/9 "2021-11-19T08:41:08Z")

</div>

> [@jerwin](#):
>
> Out of curiosity, is Devonthink’s version of Finereader compiled for Apple Silicon?

Not yet.

---

<div class="post-metadata">

**Author:** ![jerwin](https://discourse.devontechnologies.com/letter_avatar_proxy/v4/letter/j/e495f1/32.png) [@jerwin](https://discourse.devontechnologies.com/u/jerwin)\
**Post date:** [November 19, 2021, 9:34am UTC](https://discourse.devontechnologies.com/t/ocr-language-information/67765/10 "2021-11-19T09:34:55Z")

</div>

Oh OK. I’ll chalk it up to the comparatively small size of the selection captures I was processing.

---

<div class="post-metadata">

**Author:** ![WhyO\_74](https://discourse.devontechnologies.com/letter_avatar_proxy/v4/letter/w/97f17d/32.png) [@WhyO\_74](https://discourse.devontechnologies.com/u/WhyO_74)\
**Post date:** [November 19, 2021, 9:52am UTC](https://discourse.devontechnologies.com/t/ocr-language-information/67765/11 "2021-11-19T09:52:13Z")

</div>

I have FineReader OCR Pro too, and that app is slooow. But very good if you need to adjust color, contrast, crop aso a pdf.

---

<div class="post-metadata">

**Author:** ![jerwin](https://discourse.devontechnologies.com/letter_avatar_proxy/v4/letter/j/e495f1/32.png) [@jerwin](https://discourse.devontechnologies.com/u/jerwin)\
**Post date:** [November 19, 2021, 10:10am UTC](https://discourse.devontechnologies.com/t/ocr-language-information/67765/12 "2021-11-19T10:10:59Z")

</div>

My favorite feature of Finereader PDF (version 15) is that it will read Fraktur and prerevolutionary Cyrillic. The user interface is painfully slow on my m1.