An interdisciplinary group of researchers from the University of Central Florida has managed to develop a deep learning system that is capable of identifying sarcasm and irony on social media. The new model optimizes communication, by shedding light on the real motivations underlying each message.
Social media has changed our lives in recent years. According to data from IAB Spain, in 2021 Spaniards use up to 5 different social networks per month. 45% of users recognize that these platforms directly influence their purchasing decisions. In addition, 85% of the Spanish population between 16 and 70 years of age uses social networks regularly.
Thank you for watching
For companies, this social reality has made it an obligation to set up communication and marketing channels through social networks. They require feedback from consumers, through comments or messages that mark the acceptance or rejection of each of the proposals, products, or services.
Despite this, one of the great bottlenecks of digital communication for companies and organizations is sarcasm and irony in written messages, which makes it difficult to understand the needs of users on social networks. Something similar happens when attempting interpersonal communication on these platforms: on many occasions, sarcasm and irony can complicate relationships and dialogue.
A smart solution to Artificial Intelligence detects sarcasm and irony on social media
Now, the new Artificial Intelligence scheme developed by US researchers solves this problem, identifying this type of expression in messages on social networks. According to a press release, in this way, the essence of what the user wants to communicate can be captured with greater precision.
As sarcasm and irony are forms of emotional communication and are generally manifested through gestures or vocal inflections, it is very difficult to detect them in written texts. However, through the deep learning system, the specialists were able to create a tool that learns to identify these feelings.
Deep learning is one of the most promising variants of Artificial Intelligence, considering that it supports working with a greater number of layers of information simultaneously. At the same time, these systems can learn from their own mistakes and thus improve processes. Taking these conditions into account, it is an ideal scheme for this study.
To apply it to the subject that occupies them, the specialists “taught” the Artificial Intelligence model which was the keywords that could indicate sarcasm and irony in a text written through social networks. According to the conclusions of the study, published in the journal Entropy, the system was later “fed” with a huge amount of data and information about interactions on digital platforms.
By combining disciplines as seemingly dissimilar as data science and psychology, the group of researchers succeeded in making the deep learning system prove effective in tests. In this way, the application is capable of detecting sarcasm or irony even though it cannot appreciate the tones of the voice or certain facial expressions: it only achieves this through written texts.
It would be desirable that these types of tools could be used to improve communication between companies and institutions with consumers, and not only as a new way of collecting data for marketing campaigns. In an increasingly informed society, the search for an efficient and constant dialogue with users should be part of the social responsibility of companies and organizations.