I’ve finally settled on an efficient system for those who prefer logging with pen and paper instead of phones or laptops. Writing in a notebook while out in the field has always been the easy part. The hard part—the soul-crushing, time-draining part—was transcribing those notes. I used to need two cups of strong black coffee just to get through one activation. Now, I can get by with one. Here’s how.
First, take clean, well-lit pictures of your log. A photo capturing two pages at a time works fine. Feed those images into the ChatGPT 4o engine and ask it to transcribe them. At the start, it made mistakes—misreading columns, bungling signal reports, that sort of thing. But with a few tweaks to the prompt, like telling it what to expect in the first column or clarifying the formatting, the output quickly improved. Soon enough, I was getting transcripts with only minor conversion errors.
Next, I ran the output through a text editor for quick corrections and light reformatting. I’d comment out observations or extra remarks I didn’t want included, then pass the cleaned text to FLEcli to generate an ADIF file.
For this workflow to shine, it helps to write your logs in the same format expected by FLE (Fast Log Entry). That’s always been my approach anyway, so I could test the system on old logs as-is. You’ll also need to have legible handwriting—something ChatGPT handles surprisingly well if you follow its tips for maximizing OCR accuracy, which seems all common sense to me, but I’ve included at the end of this post for your convenience.
Years ago, I tried various handwriting OCR software to process my logs. They all failed miserably because most relied on dictionaries to “correct” misconversions, and callsigns, of course, don’t follow typical English patterns. The only exception was a machine learning-based handwriting project on GitHub, which could bypass the dictionary correction entirely. ChatGPT, though, didn’t need that level of directive. It recognized the callsigns well enough right out of the box, requiring little effort to make it work.
FLEcli: GitHub - on4kjm/FLEcli: Multi-platform "Fast Log Entry"(FLE) processing tool
Handwriting Recommendations for Better OCR Accuracy
To improve the accuracy of ChatGPT’s image-based character recognition (OCR), follow these tips for fast and clear handwriting:
1. Handwriting Style
- Use Block Letters or Neatly Written Cursive:
- Printed block letters are easier for OCR to read than cursive.
- If you prefer cursive, ensure it’s legible and well-spaced.
- Avoid Overlapping Letters:
- Ensure each letter is distinct and does not blend into others.
- Consistent Letter Size:
- Write letters of a uniform size, avoiding overly small or large characters.
2. Writing Tools
- Use a Dark Pen or Marker:
- Black or blue ink works best. Avoid pencils or light-colored pens.
- Medium or Thick Tip:
- Thin tips may result in faint or hard-to-read characters.
3. Writing Surface and Alignment
- Use a Plain, White Background:
- Write on unlined, white paper for a clean contrast.
- Avoid Tilting Text:
- Keep your writing straight and aligned horizontally on the page.
- Leave Margins:
- Avoid writing too close to the edges, as parts of the text might get cropped when photographed.
4. Spacing and Layout
- Leave Adequate Spacing Between Words:
- Ensure words and lines are spaced enough to avoid crowding.
- Avoid Overwriting:
- Do not write over erased or scratched-out areas.
- One Column per Page:
- Avoid multiple columns or cluttered text.
5. Lighting and Image Quality
- Good Lighting:
- Use bright, even lighting to avoid shadows or glare on the page.
- Stable Image Capture:
- Take photos directly above the paper, ensuring the text is flat and not distorted.
- Use a high-resolution camera for sharper details.
6. Write Symbols Clearly
- Distinguish Between Similar Characters:
- Example: Differentiate “1” from “l”, “O” from “0”, and “S” from “5”.
- Underline or Circle Key Points:
- If you include special symbols, make them prominent.
7. Practice Consistency
- Stick to a consistent handwriting style.
- This reduces variability and increases the chances of accurate OCR.
By following these tips, you can significantly improve OCR accuracy, ensuring that your handwritten notes are processed accurately and efficiently.
