CLFD: A Novel Vectorization Technique and Its Application in Fake News Detection
14702Authors
nan
Abstract
AbstractIn recent years, fake news detection has been an emerging research area. In this paper, we put forward a novel statistical approach for the generation of feature vectors to describe a document. Our so-called class label frequency distance (clfd), is shown experimentally to provide an effective way for boosting the performance of machine learning methods. Specifically, our experiments, carried out in the fake news detection domain, verify that efficient traditional machine learning methods that use our vectorization approach, consistently outperform deep learning methods that use word embeddings for small and medium sized datasets, while the results are comparable for large datasets. In addition, we demonstrate that a novel hybrid method that utilizes both a clfd-boosted logistic regression classifier and a deep learning one, clearly outperforms deep learning methods even in large datasets.
Keywords
Full Content
Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 3475–3483 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 3475 CLFD: A Novel Vectorization Technique and Its Application in Fake News Detection Michail Mersinias1, Stergos Afantenos2, Georgios Chalkiadakis1 1ECE, Technical University of Crete, Chania, Greece 2IRIT, Universite Paul Sabatier, Toulouse, France mmersinias@isc.tuc.gr, stergos.afantenos@irit.fr, gehalk@intelligence.tuc.gr Abstract In recent years, fake news detection has been an emerging research area. In this paper, we put forward a novel statistical approach for the generation of feature vectors to describe a document. Our so-called class label frequency distance (clfd), is shown experimentally to provide an effective way for boosting the performance of machine learning methods. Specifically, our experiments, carried out in the fake news detection domain, verify that efficient traditional machine learning methods that use our vectorization approach, consistently outperform deep learning methods that use word embeddings for small and medium sized datasets, while the results are comparable for large datasets. In addition, we demonstrate that a novel hybrid method that utilizes both a clfd-boosted logistic regression classifier and a deep learning one, clearly outperforms deep learning methods even in large datasets. Keywords: machine learning, document classification, fake news detection 1. Introduction The frequency and impact of fake news has substantially increased in the 21st century, due to the easy access to un- filtered information that is available to the public, mainly anonymously across the Web. It has been shown that fake news spreads faster than credible news (Vosoughi et al., 2018) while at the same time, fake news generation is greater than news fact checking (Rimer-Boston, 2017). According to a March 2018 Eurobarometer survey (Euro- barometer, 2018), fake news constitutes a threat to democ- racy for 83% of the EU population. Many governments have signed fake news laws which make it a crime to spread fake news online (Ikagai Law, 2018). This means that there is a growing demand for automatically detecting and stop- ping the spread of fake news, as its quick and effective de- tection can have a positive social impact. However, the complex nature of fake news makes its detection a chal- lenging task. There has been a considerable amount of work on fake news detection in the past few years. Existing research falls broadly within the realm of linguistic analysis, which uses natural language processing and machine learning meth- ods; or that of network analysis, which utilizes knowledge network graphs (Conroy et al., 2015). We follow the first approach and treat fake news detection as a binary text clas- sification task. Considering a fake news detection system, a constant flow of news articles has to be classified as ei- ther credible or fake. More generally, a dataset is com- posed of a number of news articles, or documents, each of which contains a certain number of features. The features always included are the title and the body text of the news article. Other features such as the author may sometimes be included. …
Content truncated to 500 words
Labeled Text
AbstractIn recent years, fake news detection has been an emerging research area. In this paper, we put forward a novel statistical approach for the generation of feature vectors to describe a document. Our so-called class label frequency distance (clfd), is shown experimentally to provide an effective way for boosting the performance of machine learning methods. Specifically, our experiments, carried out in the fake news detection domain, verify that efficient traditional machine learning methods that use our vectorization approach, consistently outperform deep learning methods that use word embeddings for small and medium sized datasets, while the results are comparable for large datasets. In addition, we demonstrate that a novel hybrid method that utilizes both a clfd-boosted logistic regression classifier and a deep learning one, clearly outperforms deep learning methods even in large datasets.
Publication Details
- Paper ID:
- 14702
- Year:
- 2020
- Publication:
- TIPSTER
- Publication Type:
- conf
- DOI:
- nan