Identifying biases in articles published in the news media is one of the most fundamental problems in the realm of journalism and communication, and automatic mechanisms for detecting that a piece of news is biased have been studied for decades. In this paper, we compare the WiSARD classifier, a lightweight efficient weightless neural network architecture, against Logistic Regression, Gradient Tree Boosting, SVM and Naive Bayes for identification of polarity in news. Motivated by the fast pace at which news feeds are published, we envision the increasing need for efficient and accurate mechanisms for bias detection. WiSARD presented itself as a good candidate for the task of bias identification, specially in dynamic contexts, due to its online learning ability and comparable accuracy when contrasted against the considered alternatives.
Evaluating weightless neural networks for bias identification on news
De Gregorio M;
2017
Abstract
Identifying biases in articles published in the news media is one of the most fundamental problems in the realm of journalism and communication, and automatic mechanisms for detecting that a piece of news is biased have been studied for decades. In this paper, we compare the WiSARD classifier, a lightweight efficient weightless neural network architecture, against Logistic Regression, Gradient Tree Boosting, SVM and Naive Bayes for identification of polarity in news. Motivated by the fast pace at which news feeds are published, we envision the increasing need for efficient and accurate mechanisms for bias detection. WiSARD presented itself as a good candidate for the task of bias identification, specially in dynamic contexts, due to its online learning ability and comparable accuracy when contrasted against the considered alternatives.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
