The system, devised by researchers from the University of Huelva, predicts its arrival with the data that users provide in a mobile application
Hit the beach, plant your umbrella, and head off to shore for a refreshing dip. But surprise! The sea is full of jellyfish. Those annoying creatures that, despite being in their midst, become the number one enemy of vacationers. Now however bathers have a new ally and it is at their fingertips, or rather, the mobile phone.
“I am in La Malagueta and there are a lot of jellyfish on the shore”. By introducing this data in an app we are already helping a group of inresearchers from the University of Huelva who have developed a prediction model for the appearance of jellyfish on the beaches, thanks to artificial intelligence (AI). To do this, they use the information that citizens enter in the Infomedusa app, a mobile application developed by the Aula del Mar de Málaga.
It is what is known as citizen science, that is, one in which citizens themselves participate in any of the phases of the research process. “Thanks to artificial intelligence, we correlate the data we get from the information shared by users with scientific data”, Explains to NIUS Juan Carlos Gutiérrez, researcher at the UHU. In this way, they are able to predict the arrival of jellyfish to the bathing areas.
Thus, experts have established what is the dynamics and the path of these marine species when they approach the beach. Thus, they have verified that the jellyfish first reach the coasts closest to the Strait of Gibraltar and, from there, are progressively carried by the currents and pushed by the wind towards the Levant. Already close to the beaches, they reach the bathing areas if the wind is agitated with intensity and perpendicular to them.
The great difficulty for AI: colloquial language
The great difficulty, acknowledges the expert, are the colloquial phrases that artificial intelligence is not able to interpret. “Users enter the data in a kind of chat that the application has and often write in any way, which makes the process difficult,” admits the expert.
To avoid mistakes, they have used a glossary of 557 base words, i.e. keywords that are related to the data they are looking for and two researchers have assigned numerical values to the adverbs of quantity of each message so that the machine can analyze them correctly. “It is about assigning a category to these indeterminate adverbs of quantity,” he points out. In this way, if a user writes “Today there are many jellyfish on the beach”, the system associates ‘many’ with a numerical value of 5 to indicate a large number of these invertebrates on the beach, the expert explains.
Despite the difficulty sometimes entailed in the colloquial language commonly used in chats and social networks, the reliability of the information extraction is quite high, exceeding 80 percent. However, they work on different methodologies that facilitate the interpretation of the data provided by citizens, so that the system improves its margin of error. “We are even thinking about ‘direct’ users to enter information in a specific way. Thus, it would be much easier to interpret it and it would not require human intervention ”, Gutiérrez summarizes.
At the moment, they only work with the beaches of the Malaga coast but the objective is to cover the entire coast of Andalusia. “This will allow us know a couple of days in advance if there will be jellyfish on the beach we want to go to and, therefore, plan our trip”, Says the expert. A new ally to prevent sun and beach wishes from being thwarted.