The Proliferation of Fake News - Research Proposal Example

Fake News Detection Using Deep Learning and Machine Learning

1.0Introduction

1.1Problem definition

Both the public and governments have raised concerns over the proliferation of fake news that has been caused by rapid technological developments. Fake news takes advantage of the mainstream media and social media channels to dispatch and disperse phony information. Online news consumers are usually the targets of fake news because most social media sites do not have strict policies for regulating the flow of information on their websites. It is approximated that 72.3% of fake news starts with mainstream media and online social networks. The reasons behind this are many. Firstly, the rise in social media usage has reduced the barriers to entering the media industry. Blogs and online communities have become common in recent decades, allowing everyone to post pieces of news online. It is worth adding that fake news is generated and dispersed intentionally to mislead the readers. Therefore, fake news requires a large audience base for rapid dispersion. The second reason why fake news has become an issue of concern is many social media users. Conspiracies being shared on social media networks increases the spread of fake news exponentially and uncontrollably. Thirdly, the fact that the general public's confidence levels on mass media make them more susceptible to fake news circulating on social media networks as they browse the headlines and ignore the content.

Thesis statement: this research paper aims to describe developing a fake news detection system using an algorithm based on deep learning or machine learning.

If ‘a’ was the news article whose definition is comprised of several characteristics such s its title, the text in it, images/photos, publisher, author, sources of information/data, etc., then the function f, that is dependent on 'a' and its characteristics can be represented as:

F(a)= 0 if a is fake

F(a)= 1 if a is true

Related work

In the recent past, developers have made numerous attempts to develop software that will be able to recognize and filter fake news using algorithms based on deep learning and machine learning. Most of the studies and research work have focused on detecting hoaxes through social media channels as they are regarded as the main spreading channels of fake news. Models that have been proposed calculate the probability of a news post on a social media channel to be false based on its characteristics. Characteristics of the posts that are considered are the number of followers, the number of likes, and the number of times that particular post has been shared. This is achieved by applying various deep learning and machine learning methods such as classification trees and SVM. Such methods have shown success results of over 93% based on the accuracy of detecting clickbait fake news.

A fake news detection system should be able to classify an unseen document as either false or deceptive. Such a system is based on the framework of computational linguistics. Basically, fake news detection is done using two approaches, the linguistic and the network approach.

Linguistic approach

Mihalcea and Strapparva first used natural language processing (NLP) techniques for fake news detection in 2009. Previous studies have also been done regarding fake news detection based on sentimental analysis, lexical analysis, the similarity of content, the similarity in style, and the consistencies of semantics. Data sets sourced from the crowd can also be analyzed to detect deceptive text by applying a semi-supervised approach, as proposed by hai et al., (2016). The text contained in the news is also analyzed to detect their authenticity besides the analysis of the features of the user who shares the news.

Network-based approaches

Fake news detection can also be achieved by analyzing the structure of the network and its behavior. Network structure and behavior are crucial features used in the development of a knowledge graph used to verify the truthfulness of news based on facts derived from the relationships that exist between network entities. According to the new model developed by Ciampalgia et al., () the probability of news can be derived using the network effects of variables. The accuracy of such methods was approximated to be 61% to 95%. Based on this approach, social network behavior is an interesting area for further study.

Neural network-based approaches

Research problem

This research paper proposes to study fake news detection in online social networks based on specific news articles, their creators, and their subjects. The study will consider diverse sources of information, the text contained in them, the authors' profile & description, and the relationship between the news articles with their subjects to detect fake news on online social networks simultaneously. The fake news detection problem proposed in this study is one based on credibility inference. Real news will have a higher credibility rating, while fake ones will have lower credibility.

Fake news detection has been described as difficult by previous scholars based on the following:

• Formulation of the problem: since the research problem being studied in this paper belongs to a relatively new area of study, it will be necessary to define and formulate the problem before studying it formally.
• Use of textual information: a model that will be able to extract textual information about articles, their creators, and their subjects from online social networks will be required. The model should be able to capture the signals to be used to measure the validity of the news. Such a model is not easy to develop.
• The complex fusion of information: news articles have a very close correlation, easily identifiable using author information and the relationship between the article and their subjects. Developing a learning program that integrates these aspects will enhance the credibility of the fake news detector's results.

1.2Background of the Study

This section of the research paper will introduce the whole issue of fake news while giving case scenarios. This section will also provide an overview of how the discussion paper contributes to the existing knowledge in fake news detection using deep learning and machine learning. Various theoretical concepts and frameworks regarding fake news detection using algorithms based on deep learning and also machine learning will be introduced in this section of the research paper.

Limitations and challenges of the research will also be discussed in this section, including methods of collecting data.

2.0Chapter 2: Literature Review

Various research articles on fake news detection, including the ones attached in the assignment requirements, will be reviewed in this chapter. The literature will also be linked with the main themes of the paper to fill knowledge gaps in the discussion area. The relevance of the research articles in answering the research question based on validity, reliability, consistency, accuracy, and adequacy of the information provided will also be considered in this section of the research paper.

3.0Chapter 3: Data

A brief description of the materials/data used in the research study and a summary of the statistical/empirical information will be provided. Also, data presentation using figures or tables will be done in this chapter.

4.0Chapter 4: Analysis and Discussion of Results

This will be the main body of the research paper, where the information/data presented in the previous chapter will be analyzed and discussed in relation to the research question. It will also provide a comparative analysis of the data to prior studies on fake news detection using deep learning and machine learning.

Conclusion

In this section, I will evaluate whether the paper has answered the research question by providing a summary of the main findings. I will also provide the need for future research on the subject of fake news detection. While also giving valid recommendations.