Tech News

Researchers develop synthetic intelligence that may detect sarcasm in social media

data science
Credit score: Pixabay/CC0 Public Area

Pc science researchers on the College of Central Florida have developed a sarcasm detector.

Social media has change into a dominant type of communication for people, and for firms trying to market and promote their services and products. Correctly understanding and responding to buyer suggestions on Twitter, Fb and different social media platforms is crucial for achievement, however it’s extremely labor intensive.

That is the place sentiment evaluation is available in. The time period refers back to the automated means of figuring out the emotion—both constructive, unfavourable or impartial—related to textual content. Whereas synthetic intelligence refers to logical knowledge evaluation and response, sentiment evaluation is akin to appropriately figuring out emotional communication. A UCF staff developed a method that precisely detects sarcasm in social media textual content.

The staff’s findings had been not too long ago revealed within the journal Entropy.

Successfully the staff taught the pc mannequin to search out patterns that usually point out sarcasm and mixed that with instructing this system to appropriately pick cue phrases in sequences that had been extra prone to point out sarcasm. They taught the mannequin to do that by feeding it massive knowledge units after which checked its accuracy.

“The presence of sarcasm in textual content is the principle hindrance within the efficiency of sentiment evaluation,” says Assistant Professor of Engineering Ivan Garibay ’00MS ’04PhD. “Sarcasm is not all the time straightforward to establish in dialog, so you possibly can think about it is fairly difficult for a pc program to do it and do it properly. We developed an interpretable deep studying mannequin utilizing multi-head self-attention and gated recurrent models. The multi-head self-attention module aids in figuring out essential sarcastic cue-words from the enter, and the recurrent models be taught long-range dependencies between these cue-words to higher classify the enter textual content.”

The staff, which incorporates laptop science doctoral scholar Ramya Akula, started engaged on this downside beneath a DARPA grant that helps the group’s Computational Simulation of On-line Social Habits program.

“Sarcasm has been a significant hurdle to growing the accuracy of sentiment evaluation, particularly on social media, since sarcasm depends closely on vocal tones, facial expressions and gestures that can’t be represented in textual content,” says Brian Kettler, a program supervisor in DARPA’s Info Innovation Workplace (I2O). “Recognizing sarcasm in textual on-line communication isn’t any straightforward process as none of those cues are available.”

This is likely one of the challenges Garibay’s Complicated Adaptive Techniques Lab (CASL) is learning. CASL is an interdisciplinary analysis group devoted to the examine of complicated phenomena similar to the worldwide financial system, the worldwide info setting, innovation ecosystems, sustainability, and social and cultural dynamics and evolution. CASL scientists examine these issues utilizing knowledge science, community science, complexity science, cognitive science, machine studying, deep studying, social sciences, staff cognition, amongst different approaches.

“In face-to-face dialog, sarcasm may be recognized effortlessly utilizing facial expressions, gestures, and tone of the speaker,” Akula says. “Detecting sarcasm in textual communication is just not a trivial process as none of those cues are available. Particularly with the explosion of web utilization, sarcasm detection in on-line communications from social networking platforms is far more difficult.”

Scientists devise algorithm that detects sarcasm higher than people

Extra info:
Ramya Akula et al, Interpretable Multi-Head Self-Consideration Structure for Sarcasm Detection in Social Media, Entropy (2021). DOI: 10.3390/e23040394

Offered by
College of Central Florida

Researchers develop synthetic intelligence that may detect sarcasm in social media (2021, Might 7)
retrieved 8 Might 2021

This doc is topic to copyright. Other than any honest dealing for the aim of personal examine or analysis, no
half could also be reproduced with out the written permission. The content material is supplied for info functions solely.

Source link