A research team from the University of Jaén (UJA) developed an artificial intelligence-based system for recognizing the emotions of Twitter users… First applied to the Spanish language, this technology senses the mood of people who post on a social network, analyzes and categorizes them.
According to the Discover Foundation, the researchers are focusing their findings on areas such as identifying depression, anorexia, and bulimia, or offensive and aggressive language, among other areas.
Human language technology is a branch of artificial intelligence that focuses on the study of computer systems capable of understanding and generating language. This area is related to machine learning, which the ability of a software or machine to identify and learn complex patterns in the form of mathematical algorithms autonomously.
Experts are applying this technology to a dataset of tweets previously collected and analyzed by people in order to detect emotions in the text. Further, teach a machine to interpret new terms in Spanish using dictionaries and lexicons Into the system.
“This technology can be applied in a variety of fields to identify mental health problems or verbal abuse,” explained Flor Miriam Plaza, co-author of this study and researcher at the University of Jaén.
In a study entitled “Improving Emotion Recognition in Spanish Social Networks through the Use of Lexical Knowledge”, published in the journal Future Generation Computer Systems, specialists train the computer system with a series of tweets already collected and previously translated into Spanish.
In this way, generates a language model to recognize emotions such as anger, fear, joy and sadness. “This is a difficult job because it is not a binary classification of negative and positive emotions. There are many nuances that allow, for example, to detect joy, sadness or surprise, ”commented UJA researcher Maria Teresa Martin.
After this basic information was integrated into the system, the researchers incorporated new words from dictionaries and new words into expand the number of shades so that he can perceive and increase his accuracy. This gradual language learning, independent of the previously developed database, was intended to improve the efficiency of the system.
Once a tweet is found, the system analyzes it and assigns an emotion based on the generated language model, in this case Spanish. Experts noted in this study that the emotion most represented in tweets was joy, because The system was easier to detect than anger, fear, or sadnesswhich have great nuances.
University of Jaén researcher Luis Alfonso Ureña noted that “not perfect process since this technology does not clearly perceive speech images such as irony, sarcasm or standard phrases, and, in addition, new expressions are constantly generated. “Therefore, in order to improve this system, the machine must constantly ‘learn’ a particular language, for example Spanish from Spain or British English.”
In previous research, the Intelligent Access to Information Systems group focused on identifying anorexia and bulimia, as well as misogyny and xenophobic language on social media. Urenia explained that the research group is focusing on human language technologies used, among other things, to analyze feelings in Spanish. In the future, the idea “improve technologies based on artificial intelligence and machine learning which we use to apply it to a wider range of settings. ”
The study is funded by the Intelligent Access to Information Systems Research Group, the European Regional Development Fund (Feder), the Living-LANG project and the Spanish Government Networks project.