New Computer Learning System Can Recognize Irony And Sarcasm


Computer science student Eden Saig of Technion-Israel Institute of Technology said that it is possible to detect the intent of social media users in their messages based on subtle hints from the way they express their ideas in the text.

Saig supported his argument by developing a new computer learning system that is said to be capable of detecting emotions and sentiments found hidden in text messages and emails. The said system has been designed to identify certain keywords and individual grammar habits that give a unique characteristic to the sentence structure.

"Now, the system can recognize patterns that are either condescending or caring sentiments and can even send a text message to the user if the system thinks the post may be arrogant," said Saig.

Saig added that voice tone and inflections are important in conveying the meaning from a verbally communicated message. In the case of using images or emojis, however, Saig said that this manner of expression lacks the subtle or complex feelings that are found in real-life verbal communication.

"These icons are superficial cues at best," stated Saig.

Saig examined 5,000 posts on social media pages and performed a statistical analysis on each post. These pages, labeled as "superior and condescending people" and "ordinary and sensible people" were initially created to contain humorous posts, which could explain why they have enjoyed huge popularity in Israel.

Social media users are then invited to post their suggestions of a phrase, which they believe could convey a stereotypical saying that may be suitable for a particular page.

By applying so-called "machine-learning" algorithms to the submitted contents, Saig was able to get valid results that would help him automatically detect stereotypical manners hinted in the everyday postings of people who enjoy social network communications.

Seeing that the contents mostly used colloquial, everyday language, Saig further expressed that "the content could provide a good database for collecting homogeneous data."

Saig hopes that his system of sentiment analysis would provide a way to see intent even when such intent is not verbally expressed.

"I hope that ultimately I can develop a mechanism that would demonstrate to the writer how his or her words could be interpreted by readers," said Saig.

ⓒ 2018 All rights reserved. Do not reproduce without permission.
Real Time Analytics