We use the Trektellen app to enter observations for this project. The data is uploaded the same day to the Trektellen website and shortly thereafter to the Estonian database PlutoF.

During the spring, I experimented with dictation. While observing migration, I used a smartphone voice recording app. The idea was to convert the dictations to text using Whisper speech recognition software and then parse the file into a format that could be imported into Trektellen, for example. In principle, it worked. Bird names (I used Finnish) were somewhat challenging for Whisper, so I prepared a database of “synonyms”—incorrectly heard bird names—that the parser used. For instance, the name for Barnacle Goose (valkoposkihanhi) was never recognized correctly, but a slightly shortened version (valkoposki) was easy.

A bigger issue was that Whisper often forgot some lines. The line was perfect in the dictation but did not appear in the text file. This wasn’t very common, but it happened almost every day. I found it best to run the Whisper output file through a preparation stage, then listen to the recording while reviewing the text file and making corrections. After that, the final parser could easily create the CSV files for import.

I believe this system is probably the best for avoiding errors when entering data—compared to apps and the traditional method of using pen and paper. However, it is not very practical for projects like the Põõsaspea project. For example, the last observer closing the count at 11 pm would not want to spend the next half hour on a computer, listening to their own voice and making corrections.

Panorpa

Argentina anserina