Artificial Intelligence to the Rescue

Conducting an inventory of a collection is a demanding task, complicated by the need to decipher old handwriting and the risk of misinterpretation. The Natural History Museum of Neuchâtel is currently engaged in this challenging process, focusing on inventorying and digitizing its historical mollusk collection. This work proceeds in several stages: first, each specimen in the collection is photographed. Next, the information on the labels is deciphered and transcribed into a database. This step is far from straightforward, as it requires the reading of various historical handwriting styles that have evolved over the centuries. Finally, the collected data must be interpreted and categorized accurately.

Fortunately, the institution has benefited from valuable assistance: artificial intelligence. AI plays a crucial role in ensuring the accurate interpretation and categorization of biological data. In a test sample of 1,783 lots, AI successfully identified scientific names in 80% of cases and demonstrated strong performance in detecting collectors, with 83% of the lots correctly transcribed.

A Significant Time Saver…

The high performance of AI enables the rapid and reproducible processing of large volumes of data. For instance, it takes less than 10 minutes for AI to process 1,783 samples, compared to 30 hours for a human. There is still room for improvement, for example by calibrating the AI with reference lists, including the names of the museum's collectors and species thesauri with authors and their synonyms.

Artificial intelligence proves particularly effective in automating repetitive tasks. The museum is categorical on this point: it is a flexible and promising tool, offering significant potential for optimizing digitization processes and streamlining information processing.

…But Human Involvement Required

This new tool significantly reduces the time needed for data processing, allowing the institution's experts to focus on improving catalog data quality, enhancing documentation, and sharing collections. But AI doesn't do everything. The next step requires adapting the data according to specific lexicons. Here, human intervention remains essential, especially for the most complex lots.

Nevertheless, artificial intelligence represents a major advancement in the use of technology to better document, understand, and share the rich mollusk collections at the Natural History Museum of Neuchâtel.

—

SwissCollNet

The Swiss Natural History Collections Network (SwissCollNet) is an initiative from the Swiss Academy of Sciences (SCNAT), dedicated to the digitization and long-term management of Swiss natural science collections. SwissCollNet has supported 68 projects conducted in Switzerland’s museums, universities, and botanical gardens, covering extensive collections of animals, plants, fungi, rocks, fossils, and DNA samples. It aims is to improve their accessibility for research, education and the wider public.