Inteligencia Artificial

The thirty participants have developed seven artificial intelligence projects with social impact, including one aimed at detecting the degree of loneliness in the elderly. The organizers of the program, among which is the Chair Cuatroochenta, are committed to holding a second edition to bring AI to more people and to address deep learning techniques.

The Universitat Jaume I (UJI) of Castelló has hosted this Saturday, February 19, the DemoDay to close the first edition of the Saturdays AI Castellón training course, in which students have presented the seven projects developed during the course. LactIA, a chatbot to support lactation, and Machine Orange, a system for the early detection of diseases in oranges are the winners of this first edition. Thanks to the help of volunteer mentors, the participants have learned, during 14 Saturdays between October and February, how to apply artificial intelligence to technological solutions aimed at improving people's daily lives.

7 open source projects based on machine learning with social impact

LactIA: Lactation Support Chatbot

It has been the winner of the jury in this first edition and its four members have obtained a discount on the Saturdays AI online master's degree in deep learning. They have developed a chatbot that includes natural language processing (NLP) techniques, one of the most promising artificial intelligence technologies and transformer algorithms that perform semantic processes. It responds quickly to questions about lactation and, moreover, does so at any time, facilitating access to information.

"On the Internet, there is a lot of information about lactation, but sometimes it is difficult to find the right one. We thought it would be interesting to create a chatbot so that lactating mothers could have access to get their questions answered continuously and whenever they needed it".

Pilar Espinosa, Lactia group memberer

Machine Orange: Disease detection in oranges

This prototype received the public award with 32% of the votes. It is a semi-automatic detection system that makes it possible to determine whether or not an orange is diseased while it is still on the tree. It is designed for growers to take a photo and from that image, the system detects if it is suffering from any type of pest, such as cotonet, thrips or red louse, some of the most common citrus diseases.

"In the future, a drone could integrate this solution to detect diseases in crops, not only oranges, but also any other vegetable.”

Antonio Rubio, a member of the Machine Orange group.

AIsolation: Detecting loneliness

Motivated by the increase in social isolation caused by the pandemic, they have developed a model to determine the state of loneliness of a person. This prototype has taken advantage of the Cuatroochenta Chair's database that investigates frailty and unwanted loneliness in the elderly. Through various questions and personal information, the system is able to determine the degree of loneliness of a person.

AIntiscam: Detection of scams in RRSS

Its objective is to identify fraudulent methods in comments on social networks. The algorithm they have created analyzes the texts of comments on YouTube and locates possible fraudulent patterns to predict whether there are signs of scams. It is a supervised machine learning model to locate the patterns used by fraudsters through bots.

VolcanAI: Anticipation and prevention of volcanic eruptions.

Design of a sequential model, based on a personalized neural network, that seeks to anticipate volcano eruptions, just as rainfall or heat wave prediction systems do.

Line brAIk: Detection and calculation of reading speed

A tool for estimating reading speed by tracking the student's gaze in order to detect the moments of greatest concentration and optimize study time. The system estimates the student's productivity by detecting the number of lines read correctly during a given time.

NutridAlet: Meal recommendation

An app that takes into account the nutritional values and preferences of the user to recommend healthy and appetizing recipes. Through an initial survey, the algorithm learns what your preferences are. In addition, it is based on the ingredients you have available in your fridge, as it has the function of scanning your shopping receipt and creating an online pantry with the food you have at home.

All the projects have a social impact, are aligned with at least one of the Sustainable Development Goals (SDGs) and are published in the open-source repository Github. For almost four months, the thirty students have acquired the necessary knowledge in machine learning to subsequently collaborate as a group and invent a technological solution to improve a social problem.

The closing event of the course took place in the Assembly Hall of the FUE of the UJI. The directors of the Cuatroochenta Chair, who participated in one of the initial sessions of the course, attended to learn about the projects. The event was also attended by representatives of local partners Cuatroochenta, Asociación de Big Data y Analytics (ANBAN), IOTSENS and Xarxatec. The organizers hope to launch the second edition of Saturdays in Castellón. A program that was born in 2018 and has trained more than 3,000 people in 30 cities in Europe, Latin America, the USA, Canada and Africa.

The Chair Cuatroochenta of Artificial Intelligence Welfare and Health of the Universitat Jaume I has supported the deployment of the Saturdays AI training course in Castellón.