Using an AI Large Language Model to Transcribe a Handwritten Log to ADIF
by Bill N6WS
Many of us have boxes, binders, and notebooks filled with handwritten logs from decades of operating. Those old paper logs often contain valuable DX contacts, contest QSOs, and historical operating records that never made it into electronic logging software. Until recently, converting those handwritten pages into computer-readable form was usually a tedious manual process.
Artificial Intelligence tools such as ChatGPT are now making that process dramatically easier.
A recent question on the DXLab reflector asked whether there was a practical way to convert old handwritten logbook pages into a format that could be imported into DXKeeper, the logging component of the free-ware DXLab Suite. The solution turned out to be surprisingly simple.
The basic process is to photograph or scan the handwritten log page, upload the image into ChatGPT, and ask it to transcribe the information and convert it into ADIF format. ADIF, which stands for Amateur Data Interchange Format, is the standard file format used by most amateur radio logging programs for importing and exporting contacts.
What surprised me most was how well the process worked.
I tested the method using a handwritten logbook page from 1969 containing CW contacts from an ARRL DX contest weekend. After uploading the image into ChatGPT, I simply entered the instruction: “Transcribe this image, and convert it into an ADIF file importable into DXKeeper.”
Within seconds, ChatGPT identified the callsigns, times, frequencies, modes, signal reports, and remarks from the handwritten entries and generated a valid ADIF file that could be imported directly into DXKeeper. The process is not perfect, of course. Handwriting quality matters. Callsigns written quickly during contests can sometimes confuse the software, especially characters like O and 0, or 1 and I. Faded pencil entries or poor lighting can also reduce accuracy. But for reasonably legible logs, the results can be remarkably good and far faster than manual typing. The quality of the photograph is important. A flat, well-lit image works best. If possible, avoid shadows, glare, or severe page curvature. Modern smartphone cameras are usually more than adequate. Scanning at higher resolution improves accuracy even further. One particularly useful aspect of the process is that ChatGPT understands amateur radio conventions. It can often infer the amateur band from the frequency, recognize standard operating modes such as CW or SSB, and properly structure the data into ADIF fields required by logging software.
Once the ADIF file is generated, importing it into DXKeeper is straightforward. Simply save the text with an .adi extension and use DXKeeper’s import function. The contacts then become part of your electronic logbook just like any other imported QSOs. This opens up interesting possibilities for recovering historical operating records. Many DXers have rare entities, deleted countries, contest contacts, or award credits buried in paper logs that were never entered electronically. AI-assisted transcription may now provide one of the easiest methods ever available for preserving and recovering those records.
The process also works well for:
- Old contest logs
- DXpedition notebooks
- QSL card transcription
- Award tracking records
- Station journals
- Propagation notes
While some manual review is still recommended before importing large numbers of QSOs, the amount of work required is dramatically reduced compared to entering everything by hand.
For operators with decades of paper records, this may finally make electronic conversion practical.
Recording Information Required by Awards
Logging