Researchers at MIT CSAIL and Project CETI believe they have solved something like an “alphabet” for sperm whales with the help of machine learning techniques. The findings, published under the title “Context and combinatorial structure in sperm whale vocalizations,” represent an important advance in understanding cetacean communication.
This study deals with codas, which are sequences of clicks that serve different linguistic functions. “What we discovered is that there are previously unexplained variations in the coda structure,” CSAIL Director Daniela Russ told TechCrunch. “We found that the coda types are not arbitrary, but rather form a newly discovered combinatorial coding system.”
Whale sounds have been an important subject of research for decades, but the team behind this new study says it has uncovered a previously unknown level of nuance in the talkative marine mammals. Suggests. The paper points out that 150 different sperm whale codas have been recorded in previous studies.
“These subsets have been shown to encode information about the identity of the caller and clan,” it explains. “But almost everything else about the sperm whale's communication system remains unknown, including fundamental questions about its structure and ability to transmit information.”
The research team drew on the work of pioneering marine biologist Roger Payne, who passed away last June. Paine's most influential work was on humpback whale songs. “He inspired us to want to use cutting-edge technology to better understand whales,” Russ says.
The research team deployed a machine learning solution to analyze a dataset of 8,719 sperm whale codas collected by researcher Shane Gero off the coast of the small Caribbean island of Dominica.
“We took the input and then tuned the machine learning to visualize it and understand it better,” Rus explains. “Then we will analyze the results together with biologists.”
The team's approach marked a change from older analyzes that studied individual coders. A richer picture is formed when sounds are studied in context as interactions between whales. Contextual details are categorized using musical terminology. It includes tempo, rhythm, ornamentation, and rubato. From there, the team isolated what they called the sperm whale phonetic alphabet.
“This phonogram allows us to systematically explain the observed variability in coda structure,” says Russ. “We believe this may be the first example outside of human language where communication provides an example of the linguistic concept of patterning duality. refers to a set of elements that can be combined to form a larger meaningful unit, similar to creating a “.''
Those “words” have different meanings depending on different contexts. The paper added:
Our results demonstrate that sperm whale calls form a complex combinatorial communication system. The seemingly arbitrary list of coda types can be explained by a combination of rhythmic, tempo, rubato, and ornamental features. Large-scale combinatorial vocal systems are extremely rare in nature. However, their use by sperm whales shows that they are not unique to humans and may result from dramatically different physiological, ecological, and social pressures.
While this breakthrough is exciting for all involved, there is still much work to be done. It could spread first to sperm whales and then to other species like humpback whales.
“We decided to study sperm whales because they have a broad dataset and the potential to collect even more,” Russ says. “Clicks also form a kind of discrete communication system, which is much easier to analyze than a continuous communication system. But Roger Payne's research also shows that humpback whale songs are not random. There are repeating segments, and there's an interesting structure there that we just haven't been able to explore in detail.”