Machine learning and botany combine to unlock secrets of ancient Greece

Machine Learning


Micah Gold, a junior at Yale University, can both pilot drones and write algorithms, and he'll be putting both skills to use this summer in central Greece as part of a unique archaeology project.

Gold is part of a research team at the Yale Ancient Pharmacology Program (YAPP), an interdisciplinary effort that uses ethnography, science and technology to uncover evidence of people's lives thousands of years ago.

In Antikyra, a port city on the Gulf of Corinth, the group will search for archaeological remains and botanical specimens that could lead to intriguing discoveries about medicines, olive oil, wine, perfumes and other concoctions made and consumed by the site's ancient inhabitants.

For Gold, an applied mathematics and computer science double major, archaeological fieldwork is an opportunity to use his knowledge of machine learning to find plants and places that may be rich in phytochemicals (noxious compounds that protect plants from infection and pests and may in some cases have medicinal properties), as well as an opportunity to broaden his intellectual horizons.

“I'm using what I learned in class, but I'm not just honing those skills,” he says, “I'm also learning new skills, like how to read archaeological texts and how to organize scientific research.”

This is Gold's second time assisting YAPP in the field. While working at the same location last summer, he learned to operate a drone, which was then used to take multispectral images of the area's rugged terrain, helping researchers map and locate remnants of ancient pottery and other artifacts. He then helped write algorithms that helped researchers organize and understand the multispectral data they collected.

“I'd never been to an archaeological dig before, so I had no idea what to expect,” he says. “It was an amazing experience.”

Gold and other talented undergraduates will have much to contribute to the interdisciplinary ancient pharmacology program, says Andrew Koh, YAPP principal investigator and research scientist at Yale University's Peabody Museum.

“Students contribute from day one,” he says. “We need and value their expertise. The average archaeologist doesn't know much about machine learning.”

Research team members, from left, Chris Renton, Andrew Coe, Trevor Luke, Micah Gold (back) and Savannah Bishop.
Research team members, from left, Chris Renton, Andrew Coe, Trevor Luke, Micah Gold (back) and Savannah Bishop.

'High-resolution footage from the past

Koh, who became a Peabody Research Fellow last year, is a pioneer in the use of organic residue analysis in archaeology to identify materials once contained in ancient vessels, an emerging field that is helping to reveal subtle ingredients people used in foods, drinks, medicines, fragrances and even drugs and hallucinogens.

It is an inherently interdisciplinary approach, drawing on chemistry, botany and other hard sciences, ethnographic knowledge of ancient languages ​​and cultures, and the latest technologies to find and analyze plants and ancient artifacts.

Combining these research methods can help recover lost knowledge, reconstruct potentially useful ancient materials, and provide a more nuanced and comprehensive understanding of ancient people's lives, Ko said. For example, rather than simply confirming that a ceramic vessel contained wine, YAPP's interdisciplinary approach could identify the individual ingredients used in making the wine. This could reveal important context, such as whether the drink had medicinal properties or flavor characteristics unique to a particular region, Ko said.

“We want to get as high a resolution in our understanding of the past as possible, which means we can't just rely on one discipline,” Coe says. “Chemistry can tell us something about the residues we find, but reading ancient texts might give us recipes or other clues about the materials people used. Putting all these pieces together creates a high-resolution view of the past.”

The Peabody's vast collection has provided Coe with a wealth of material to study, and he's already made some fascinating discoveries within it: Working with Agnete Lassen, curator of the Yale Babylonian Collection, Coe analyzed a 2,500-year-old Egyptian alabaster vessel and found residues in it that were a mixture of opioids.

Susan Butts, director of collections and research at the Peabody Museum, recalled taking Coe on a tour of the collection and having to pull him out of the storage facility to move on to the next collection.

“He was able to find a way to collaborate in a particular way with everyone he met,” she said. “Unlike other scientists at museums who oversee the care of specific collections as curators, Andrew is a free agent who can collaborate across collections. His research is analytical, and he's particularly well-suited for it because he looks at all kinds of materials to find new and interesting information.”

While the Babylonian and Peabody anthropology collections are obviously relevant to Coe's research, he's also working with the botanical collection, and sees nearly endless possibilities among the 14 million objects that make up the collection.

“We don't yet know how to collaborate with invertebrate paleontology,” he said.

While he unlocks secrets within his collection, his YAPP team is conducting fieldwork to unearth new discoveries.

Anticyra, a culturally significant area with a history dating back to the Middle Bronze Age, is a good example.

Hellebore Hunting

Pausanias, a Greek geographer and early travel writer, discussed Phocis, the region the YAPP team is studying, in his detailed 10-volume work Describía de Hrecia, written in the 2nd century AD. The book describes the topography and cultural geography of Greece, particularly the temples of the Egyptian goddess Isis, who was worshipped throughout the Greco-Roman world.

Coe said the temple served as a kind of clinic, and the ruins could yield a treasure trove of vessels containing interesting remains. By triangulating the distances used by Pausanias to describe the temple's location, the location of the temple site could be pinpointed, he said.

Helicon's Hellebore
The team is collecting specimens of hellebore, a plant used by ancient people for both medicinal and harmful purposes.

The region has a special connection to hellebore, a legendary plant that can be used to both heal and harm, and the duality inherent in the original Greek word “pharmakon.”NumberThe 10th-century BCE medical writer Thessalus, son of the physician-philosopher Hippocrates, said their ancestor Nebulus advised invading troops to poison the water supply of a nearby city with hellebores, which led to an epidemic of debilitating dysentery and ultimately a massacre of the inhabitants. The episode is widely thought to have inspired the Hippocratic Oath, in which doctors swear not to harm others, Coe added.

This summer, a team of Yale researchers will be documenting multispectral and phytochemical analyses of hellebores from Mountain Helicon in nearby Boeotia to better understand their past and present relationship with the region.

Rather than digging based on gut feeling and guesswork, the team uses technology such as unmanned aerial vehicles (UAVs) equipped with multispectral and thermal sensors to target and record specific locations for excavation and investigation.

The aim is to identify promising sites quickly and efficiently, without clogging up the storage facilities of YAPP's local partners with thousands of pottery shards. At the same time, the team will use the same techniques to collaborate with local botanists to collect specimens of local plants that may have been ingredients in ancient medicines or drinks.

Bootcamp

The research team on Horse Island prepares for their expedition to Greece.
The team prepared for their expedition to Greece with a “boot camp” on Horse Island, a 17-acre site owned by the Peabody Museum off the coast of Branford, Connecticut.

In April, the YAPP team conducted a boot camp in preparation for their trip to Greece on Horse Island, a 17-acre piece of land owned by Peabody in the Thimble Islands off the coast of Branford, Connecticut. Koh said the isolated island, where drones can fly without disturbing neighbors, was an ideal place for the team to demonstrate their technology.

Gold and team member Murphy Tu, a graduate student in archaeology in the School of Humanities and Sciences, flew a drone equipped with an infrared sensor to test whether their algorithm could distinguish between pottery and rocks. (Combined with multispectral imagery, Ko says, the drone could very well be able to separate pottery shards and identify plants to genus, rather than species, over large areas, including areas not easily accessible by foot.)

Patrick Sweeney, collections manager for the Peabody Institute's botany department, led a session on collecting plant specimens. Coe said the training removes any unknowns and allows the team to hit the ground running.

“I would hate to go to Greece and find out there are technical problems,” he said.

By working smarter and faster (fieldwork takes weeks, not months), we can accommodate talented students with diverse interests and who are working on a variety of projects at the same time.

For example, after Gold returns from his two-week trip to Greece, he will begin an internship at Yale University's Wu Tsai Laboratory, where he will work with Abhishek Bhattacharjee, a professor of computer science at the Yale School of Engineering and Applied Science, on brain-computer interfaces for treating neurological diseases.

“It’s convenient because I don’t feel like I have to spend all my time on the YAPP project,” Gold says, “I can come in and get some very targeted work done and make a significant contribution.”

Gold met Coe during a seminar he took as a freshman with David Skelly, Yale's Frank R. Osler Professor of Ecology and director of the Peabody Museum, on working with museum collections. As a student, he helped Coe investigate the provenance of an Egyptian alabaster jar, which led to an opportunity to join a fieldwork team in Greece last summer. Since then, Gold has been developing vision algorithms that can analyze large amounts of multispectral and thermal data and isolate key regions of interest from areas. The algorithms allow researchers to analyze in minutes what would normally take days of tedious work.

Ko calls the ability to apply machine learning to analyze the resulting data a “game changer,” and says it highlights the importance of YAPP’s multi-node approach.

“In the old days, archaeologists would dig frantically to find useful sites, but through our technology nodes, we can use thermal data and machine learning to target activity,” he said. “And that's smart archaeology.”



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *