- Researchers have deployed machine-learning techniques to analyze patents for products made from animal parts, in an effort to identify trends in the wildlife trade.
- The team analyzed patents filed over a 50-year period to understand how animal parts have been used; they also discovered trends in patenting that could indicate potential interest for future use.
- In their recently published study, they show that stricter regulations haven’t led to a reduction in patent filing for products made from animal parts.
- The team say they hope the methodology they developed can help provide evidence that could eventually be used to make policy decisions.
Economists have long made the case that regulations foster greater innovation by business players. This apparently also applies to businesses involved in the wildlife trade, new research suggests.
In their recently published study, researchers from the University of Oxford in the U.K. found that stricter regulations haven’t led to a reduction in patents filed for products made from animal parts. This includes rhino horns and pangolin scales, both of which are banned from being traded internationally.
Co-lead author Amy Hinsley attributed the finding to a lack of awareness of regulations or a belief that they might change in the future. “Businesses thinking about their long-term strategies might see patenting of an illegal product as worth it if they think the law will change again before the patent expires,” Hinsley, a research fellow and co-director of the Oxford Martin Program on Wildlife Trade, told Mongabay in an email interview.
The study also found that the rate at which patents were filed for products made from animal parts was much higher than the universal rate of patent filing. The team used machine-learning techniques to scrape through and analyze patents filed between 1970 and 2020. They looked at patents filed for products from six animal groups — rhinos, pangolins, sturgeons, horseshoe crabs, caterpillar fungi and bears — to identify trends in how and where wildlife parts are being used for commercial purposes. While many products for which patents have been filed might not be in markets yet, the filings indicate a potential interest for future use.
“Patent is a company’s way of saying that there is a potential commercial opportunity,” study co-lead author Joss Wright, also co-director of Oxford Martin Program on Wildlife Trade, told Mongabay in a video interview. “If a company is actively patenting rhino horn products, it’s only a fair guess that there is some role for rhino horn in their products in the future.”
To pull out information from patents, the team applied a machine-learning method called topic modeling that uses specific keywords to identify topics and themes in patents. This was used not only to see if specific animals were mentioned in patents but to also classify them based on their use cases, like for medicinal or farming purposes.
Eliminating patents with confusing keywords was a task in itself. For instance, initially when they searched for the word “pangolin” in the patent database, a huge number of patents showed up in the results. It was only later that the team realized that a lot of the patents were biomedical ones that used a gene with a similar name found in fruit flies. Additionally, the team also had to refine the search procedure to corroborate if the patent was actually using an animal part or if it was only mentioning the animal incidentally. “A lot of patents would contain phrases like ‘This will help reduce fever, similar to the use of bear bile,’” Wright said. “We also had to check if it was actually talking about using rhino horn or if it was a patent for something like a cuddly rhino toy.”
The team also deployed another method called change-point analysis, which helped them identify points in time where the patent data showed a marked change in trends or characteristics. For example, the team observed a significant rise in the number of patent filings for products with caterpillar fungus (Ophiocordyceps sinensis) after it had been reported to be a performance enhancer during the 2022 Beijing Winter Olympics.
Both Hinsley and Wright said the findings allowed them to get a comprehensive understanding of how the wildlife trade is changing. This included the emergence of new products that point to certain species of concern. “We can use this to monitor changes in commercial interest over time and work with these businesses to find solutions to reduce the use of illegal wildlife products,” Hinsley said.
Dealing with the messy data and developing techniques that allow computers to make sense of it was the biggest challenge. “Machine learning works very nicely with structured numbers,” Wright said. “This was messy and unstructured data and the big task was to take the data, filter it down to remove conflating words, and make it fit into a structure that the computer can analyze.”
For now, the work is still part of ongoing research. However, Hinsley said the team is hoping it will help provide evidence that could “underpin policy and practical decisions” to reduce the illegal wildlife trade.
“It’s not just about the physical action of trade or predicting trade in the future,” Wright said. “But it’s also about taking into account patents as a data source and allowing us to understand the complicated human-wildlife interface.”
Banner image: A white rhino in Kenya. Image by Pablo Necochea via Flickr (CC BY-NC 2.0).
Abhishyant Kidangoor is a staff writer at Mongabay. Find him on 𝕏 @AbhishyantPK.
Citation:
Hinsley, A., Challender, D. W., Masters, S., Macdonald, D. W., Milner-Gulland, E. J., Fraser, J., & Wright, J. (2024). Early warning of trends in commercial wildlife trade through novel machine-learning analysis of patent filing. Nature Communications, 15(1). doi:10.1038/s41467-024-49688-x