Can AI detect sarcasm with the help of ‘friends’?—Yes!

In a groundbreaking way that could potentially reduce marriages, friendships and countless other misunderstandings, scientists say they have built an AI-based sarcasm detection tool.

A group of researchers at the University of Groningen in the Netherlands claim to have developed an AI system that can tell whether you’re honest and serious, or displaying a sly and humorous wit.

Their work, presented today at the Acoustical Society of America meeting in Ottawa, combines text and speech analysis. The team is led by graduate student Xiyuan Gao, who is “addressing the problem of figurative language” as part of his academic research.

Gao did not immediately respond to a request for comment. decryption.

The research paper outlines the team’s approach to sarcasm detection.

First, they trained the neural network on multimodal data, including audio clips, transcribed text, and annotated emotional content from satirical scenes from sitcoms like “Friends” and “The Big Bang Theory,” obtained from the MUStARD database.

Next, they developed an algorithm to map emotional signals in audio and text and assigned appropriate emojis based on sentiment analysis. This makes AI more powerful at detecting sarcasm based on a multifaceted approach than other tools that rely solely on text and voice pitch changes.

“Our methodology leverages the strengths of each modality: emotion recognition algorithms analyze emotional signals in audio data, and sentiment analysis processes text,” the research paper explains. “The integration of these modalities aims to compensate for the limitations of pitch perception.”

AI detected sarcasm in scenes from a new sitcom with about 75% accuracy. However, researchers acknowledge that humans still have an advantage over machines in this particular cognitive test.

“When you start studying sarcasm, you become extremely aware of the extent to which we use it as part of our normal way of communicating.” said Matt Coler, another researcher involved in the project. tutelar. “We have to literally talk to our devices as if we were talking to a robot. Because it is.

“It doesn’t have to be this way,” he added.

This sarcasm detection tool is just the latest attempt to uncover hidden sentiment or meaning in human language, and it is becoming increasingly important as millions of conversations are held every day with large language models (LLMs) due to the explosive adoption of AI chatbots.

For example, Hume AI says it has developed an AI system that detects different inflections in the voice to identify different emotions, not just changes in tone or mood.

Hume is designed to understand and respond to human emotions in a more nuanced and empathetic way, making AI interactions more natural and engaging.

Meanwhile, a research team from Haga-Helia University of Applied Sciences and the University of Oulu, Finland, published a research paper last month exploring contactless technology in multimodal emotion recognition.

Likewise, the team found that the most accurate way to analyze emotions was to merge visual, auditory, and textual clues together. However, they emphasized that this type of understanding is still limited due to cultural variables, inherent model biases, misinterpretations, and lack of understanding of past contexts.

Other researchers have tried to analyze the use of emojis to adequately detect the emotional states of those who use them. After all, a text message followed by a winking face emoji can have a very different meaning than a message with no picture or any other emoji.

The ability to adapt to the rich nuances of conversation will help AI chatbots better respond to human messages, but it could also have far-reaching implications for society as a whole. In addition to detecting unwritten negative tones in language or identifying hate speech, it can also facilitate interpersonal interactions, providing greater clarity in moments of confusion. This is especially true for people who communicate beyond language or suffer from neurological conditions.

Edited by Ryan Ozawa.