Presentation Title

Deep Neural Networks for Satire Detection on News Headlines

Presenter Information

Paul MandalFollow

Faculty Mentor

Rakesh Mahto

Start Date

17-11-2018 8:30 AM

End Date

17-11-2018 8:45 AM

Location

C162

Session

Oral 1

Type of Presentation

Oral Talk

Subject Area

engineering_computer_science

Abstract

Detecting sarcasm has been a problem in Artificial Intelligence (AI) because it is highly dependent on context and a human agent’s knowledge of the world. This means that for conventional AI, a large number of rules must be hard coded. Furthermore, naïve machine learning methods such as logistic regression simply generate lists of words that are frequently associated and dissociated with sarcasm. Thus, logistic regression is unable to take groupings of words into account in sentiment analysis problems. In this paper, we design a feed-forward neural network that is trained on word level information to determine whether a headline is sarcastic or genuine. We then applied this neural network to a corpus of 26,709 vector encodings of real and sarcastic news headlines taken from The Onion and Huffington Post. This feed-forward network can classify whether a news headline is real or satirical with 85% accuracy.

Summary of research results to be presented

First, we trained a logistic regression on our data. Depending on how the training and test sets were divided, accuracy varied anywhere from 80%-82%. We then trained a basic feed-forward neural network with two 16-unit hidden layers which yielded an accuracy of 83%. With hyperparameter tuning and combining previous network architecture, we were able to create a neural network that could classify the news headlines with 85% accuracy.

This document is currently not available here.

Share

COinS
 
Nov 17th, 8:30 AM Nov 17th, 8:45 AM

Deep Neural Networks for Satire Detection on News Headlines

C162

Detecting sarcasm has been a problem in Artificial Intelligence (AI) because it is highly dependent on context and a human agent’s knowledge of the world. This means that for conventional AI, a large number of rules must be hard coded. Furthermore, naïve machine learning methods such as logistic regression simply generate lists of words that are frequently associated and dissociated with sarcasm. Thus, logistic regression is unable to take groupings of words into account in sentiment analysis problems. In this paper, we design a feed-forward neural network that is trained on word level information to determine whether a headline is sarcastic or genuine. We then applied this neural network to a corpus of 26,709 vector encodings of real and sarcastic news headlines taken from The Onion and Huffington Post. This feed-forward network can classify whether a news headline is real or satirical with 85% accuracy.