Year: 2021
Type: article
Abstract: —Clickbait is an article title or a social media post that attracts readers to follow a link to the article’s content. It is one of the major contributors to the spread of fake news. To prevent a wide spread of fake news, it should be detected as soon as possible. This paper presents a content-based feature called headline2vec that is extracted from a concatenation layer of a convolutional neural network (CNN) on the well-known word2vec model for high dimensional word embeddings, to improve an automatic detection of Thai clickbait headlines. A pioneer dataset for Thai clickbait headlines is collected using a top-down strategy. In the experiment, we evaluate the headline2vec feature for Thai clickbait news detection using 132,948 Thai headlines where the CNN features are constructed using a non-static modeling technique with 50 dimensions of word2vec embedding with a window size of two, three, and four with the epoch of 5. Using the proposed features, we compare three classifiers, naïve Bayes, support vector machine, and multilayer perceptron. The result shows that the headline2vec with multilayer perceptron achieves up to 93.89% accuracy and it outperforms the sequential features that utilize n-gram with tf-idf.
Language: English
Subfield: Sociology and Political Science
Field: Social Sciences
Domain: Social Sciences
Open Access status: diamond
Award: G4605047560
OpenAlex ID: W7219376205
