by Briana Giasullo
The Academy of Natural Sciences Library and Archives is excited to announce the launch of a new digital project. Using Zooniverse, an online crowdsourcing platform with almost three million volunteers around the world, we are inviting the public to interact with our collections like never before!
Our Zooniverse project, Field Journal Fix-Up, asks volunteers to help us transcribe field journals from our archives in a new way that combines Artificial Intelligence (AI) and crowdsourcing. Field journals are detailed notebooks written by scientists as they traveled on expeditions. They contain tons of information including lists of specimens collected, environmental and weather reports, travel itineraries, and more. The Academy’s archives holds many field journals, as well as many other unique materials that can be researched by appointment.
Transcribing field journals allows us to extract information from them and share it with researchers around the world. But traditional transcription, which requires multiple people working directly with fragile materials in a shared space, poses many challenges and is difficult to manage. Knowing the importance of this work, the Library and Archives team wondered: How can we make the transcription process easier?
Libraries are increasingly using AI to assist with various tasks, raising questions and concerns to be explored. AI for transcription is especially popular in part because it resembles standard Optical Character Recognition (OCR), which has been around for years. OCR takes a static image and extracts any text present in the image, making the text machine-readable. Machine-readable text can be indexed and searched by a computer, which is why it’s so helpful for research. However, OCR only works on typed text and cannot transcribe handwritten text.
Unlike OCR, AI transcription can predict words and does a much better job at transcribing handwritten materials, giving us a good starting point. AI has its own problems, though. In addition to environmental and ethical concerns, AI doesn’t actually understand what it’s writing so is prone to making errors that a person would not make. For the transcripts to be reliable, they need to be reviewed by people before they are published online. This is where Zooniverse comes in!
The Field Journal Fix-Up project blends AI with “the wisdom of the crowd.” Volunteers who navigate to our Zooniverse website are presented with a digital image of a field journal page, plus a text box containing an AI-generated transcript of that page. Working directly in the text box, volunteers can make corrections to the text and submit them to our project staff. After a set number of volunteers have reviewed a page, it becomes “retired” and is no longer displayed.
Completed field journal transcripts will receive a final review by project staff before being published onto the Biodiversity Heritage Library (BHL), where transcripts are displayed alongside their respective journal pages. BHL is completely free to use and contains over 300,000 volumes of biodiversity literature contributed by natural history museums around the world!
This project is a huge step in expanding access to our collections and has the potential to build a community of passionate volunteers. Every Zooniverse project contains a “Talk” section that acts as a public forum where volunteers can talk to each other and to the project staff. The forum will be monitored by project staff to ensure questions are answered. Subject-specific questions about identification (species names), locality, or scientific terms may be answered by scientists from the Academy.
Overall, we are excited about this new way of using technology to connect our collections and staff to the world! We believe that AI can be used responsibly to make science accessible and to build reliable sources of information on the internet. We welcome you to take part in our project by visiting this link: tinyurl.com/field-journal-fix-up
