This paper uses a systematic literature review method to find find the most recent papers online which are related to fake news detection on social medias. The paper aims to map the state of art of fake news detection, define fake news and find the most useful machine learning technique to do so.
Highlights:
- The paper finds the argument made by many researchers that information obtained from social media metrics is a key-feature for election prediction too simplistic due to the uncertainty over the real goal of political discussion on social medias, as many tend to be satirical.
- The use of social media information as a pre-processing step is however a favourable approach since it helps in identifying the starting point in the spread of rumours and used to understand the classifying process of fake news.
- The currently preferred method of handling the problem of fake news is the machine learning approach i.e. using composite classifiers which are in fact neural networks composed by classical classification algorithms which focus on a language-based analysis a s main feature of prediction.
- Bots can be viewed as catalysts for information propagation, either for good or bad purposes. Although bots don't favour a particular type of information, the scale at which they are able to disseminate fake news surpass those of a human being.
- The research concludes that contrary to many surveys, the current state of art of automatic detection of fake news is of using composite network analysis approaches on the machine learning techniques.