Introduction
1.1 Background of the Study
Fake news is defined as a misleading information that facades itself as real news (Jennifer Allen, 2020). Its main purpose is to promote a specific cause (Erin May, 2017), and is often used to influence a person’s perspective, usually for political agenda. Experts now recommend avoiding the term “Fake news” as it is usually associated with politics, but it is basically the same as “False information”.
There are several kinds of fake news. Some are not meant to harm, some are meant as jokes, and some are made to intentionally mislead people (Claire Wardell, 2017). This implies that fake news has been used throughout history often enough to have different defined types.
Recently, the amount of misinformation is increasing. This is especially true in the online context, as there is a growing amount of fake news and rumors proliferating the internet, which leads to an increase in the number of people who fall for misinformation. This backs up the common belief that information on the internet is generally unreliable (Hiroko Kanoh, 2018).
There are many ways that this phenomenon impacts people and society. One such way is in the field of politics, where many politicians, political parties, and non-impartial parties take advantage of the accessibility of the internet and the gullibility of the general public to spread information that is not necessarily true and smear other parties for their own personal interests. It is even said that fake news is funded and run by professionals (David, 2019). This is particularly relevant in the Philippine context, where there has been a recent abundance of fake news spread by troll farms and by those less informed or those misinformed themselves, creating a lot of unhealthy discourse and a general distrust of online information.
Misinformation has been rampant since the popularization of social media, and usually, these news are made to influence the public’s view for profit, push a political agenda, or cause mass hysteria. As such, the modern world needs a way to distinguish truth from distortion of facts and protect the general public from being misinformed.
1.2 Research Questions and Project Objectives
This study aims to lessen the false information consumed by the public by developing an automatically updated website containing a feed of aggregated election news and articles that are considered to be “real”. This website will be updated in real-time with minimal human interference through the use of Machine Learning techniques and will automatically filter out news that are deemed to be fake or satire, with the aim of protecting the general public from fake information and making real news accessible to the general population in time for the national elections. In order to make this website possible, datasets containing real and fake news will be analyzed in this study to produce insights regarding the key characteristics of news articles.
The researchers aim to answer the following questions:
1. How can we identify fake news from real news?
2. What are the common characteristics present in real news but not in fake news, and vice versa?
3. What metrics can be used to determine the legitimacy of a news article?
4. How can these insights be applied to Philippine news?
5. How can these insights be used to develop the algorithm for the website?
1.3 Scope and Limitations
The study will focus on three publicly-available news datasets for the analysis, and online articles reachable by the web crawler for the development and maintenance of the website. This study will mainly target the general population who have access to the internet and are hereby more prone to receiving false information online. The research will only be done at home with the use of online collaboration tools that will allow the researchers to communicate and get the necessary information needed for this research. The data to be used will only be collected from publicly-available datasets, will be processed completely through automated Machine Learning algorithms, and will not include any preferences, opinions, or perspectives of the researchers themselves. The insights gathered from the datasets will mainly be generated with the creation of the election news website in mind and are not recommended to be applied in other contexts.
One limitation of this study is that the datasets were mainly extracted from non-Philippine contexts, so one challenge for this study is applying the generated insights to Philippine context through additional analytical methods. This study is also not designed for news and transcripts created in video, audio, or photo formats, and will only be analyzing and gathering text-based articles and content. For the website itself, the amount of content generated and displayed will also inherently depend on the capacity of the server host itself and the online content that are reachable by the web-crawling algorithm used.
1.4 Review of Related Literature
There is a need for Machine Learning classifiers to detect fake news automatically (Alim Al Ayub Ahmed, 2020). The paper of Alim Al Ayub Ahmed et al. has criteria to include and exclude papers. The program then assesses the quality of papers.
There are a number of computational techniques that can be used to determine a fake article. Ahmed et al. included extracting language features like n-grams to train models to raise Machine Learning accuracy, while Wang used textual features and metadata to train his neural network while also using concatenation to determine patterns (Iftikhar Ahmad, 2020).
In discerning between real and fake news, we need to be critical and observant. There are often indications whether a news story is real or not (Justin Harrison, 2018), and it is crucial to know those indications to be aware of misinformation. As such, this kind of knowledge is needed for this study – machines are designed to function only the way a programmer functions it to, and a programmer needs to have enough knowledge of the topic to be able to apply them effectively in the creation of a program as its designer (Dale Stokdyk, 2021).
There can be a lot of different indicators of fake news depending on perspective, but they are all generally the same, just summarized in a different way. An example of this indication is that fake news has gaps or contradictions in its story. The location, logic, and statistics are also important indicators (Nanice Ellis, 2020).
Fake news cannot be verified, meaning that it cannot be found anywhere else and has no proper author (University of Connecticut, 2021). It cannot be searched in other trusted sources. It also has no known or verified author, and if there is, it generally isn’t an expert in the field.
Methodology
2.1 Research Design
This study will be focusing on gathering data from concrete, publicly-available datasets containing news that are classified as either fake or news. The program will use a combination of different algorithms with Machine Learning integrated in a website for accessibility. The program will also be using web-crawler algorithms in order to gather the data for the website.
The algorithms will find patterns in the determined elements of the respective type of news and will use different metrics to decide whether to pass the news as real or not. It will also learn the characteristics of news as it receives more data.
2.2 Requirement Analysis
The project will create a program that will find a relationship between the elements identified and the two kinds of news. The program will use a combination of algorithms and Machine Learning techniques that will be integrated in a website. This project will require:
- A computer
- A programmer with knowledge about algorithms
- A person that knows how to discern between fake and real news in detail
- A server host
- A programmer with knowledge on how to create a website
- Internet access
Since there is a lot of complex knowledge involved in creating this project, there may be more requirements. An available source of technical information is also needed for additional information.
2.3 Data Gathering Procedure
For the program to work as intended, the required data will be extracted from datasets from the website Kaggle, which is an international machine learning and data science platform and community that contains free-to-use datasets.
Once the program reaches a certain threshold of accuracy in discerning news, the future data will be gathered and filtered using web-crawling algorithms based on the insights provided by the analyses, after which they will be added to the website itself.
2.4 Data Analysis
The data gathered from datasets will be analyzed through the use of Natural Language Processing techniques in combination with Convolutional Neural Networks (CNNs), which are methods that enable machines to process and analyze text. Through this, the characteristics found across all real news and all fake news can be identified, and can provide insights on how to distinguish these real news from fake or satire news. Statistical analysis will also be used to detect and filter out machine-generated text, in order to combat “Neural Fake News” that have specifically been designed to mimic human text.
The program will conduct certain steps to determine the elements involved through repetitive learning with different data. Techniques used in machine learning such as backpropagation will be implemented to increase the accuracy of the output. Different neural network models will also be used to determine which produces the most accurate results.
References
- Ahmad, I. (2020, September 4). Fake News Detection Using Machine Learning Ensemble Methods. Hindawi. https://www.hindawi.com/journals/complexity/2020/8885861/
- Ahmed, A. (n.d.). Detecting Fake News using Machine Learning: A Systematic Literature Review. Cornell University. https://arxiv.org/ftp/arxiv/papers/2102/2102.04458.pdf
- Alhasan, M. (2020, February 17). Fake News Dataset. Kaggle. https://www.kaggle.com/mohamadalhasan/a-fake-news-dataset-around-the-syrian-war
- Allen, J. (2020, April 3). Evaluating the fake news problem at the scale of the information ecosystem. ScienceAdvances. https://advances.sciencemag.org/content/6/14/eaay3539
Amin, H. (2019, March 1). Fake News. Kaggle. https://www.kaggle.com/hassanamin/textdb3 - Bisaillon, C. (2020, March 27). Fake and real news dataset. Kaggle. https://www.kaggle.com/clmentbisaillon/fake-and-real-news-dataset/code
- Ellis, N. (n.d.). 9 Ways to Discern Fake News from Real News. Wake up World. https://wakeup-world.com/2020/07/10/9-ways-to-discern-fake-news-from-real-news/#:~:text=Fortunately%2C%20by%20using%20a%20combination%20of%20logic%20and,you%2C%20don%E2%80%99t%20take%20anyone%E2%80%99s%20word%20at%20face%20value.
- Explained: What is False Information (Fake News)? (n.d.). Webwise. https://www.webwise.ie/teachers/what-is-fake-news/
- Kanoh, H. (2018, August 28). Why do people believe in fake news over the Internet? An understanding from the perspective of existence of the habit of eating and drinking. ScienceDirect. https://www.sciencedirect.com/science/article/pii/S1877050918313851
- Quilinguing, G. (2019, September 28). The problem with fake news: UP experts speak on the impact of disinformation on politics, society and democracy. University of the Philippines. https://up.edu.ph/the-problem-with-fake-news-up-experts-speak-on-the-impact-of-disinformation-on-politics-society-and-democracy/
- Thompson Rivers University. (n.d.). Fake News. TRU Libraries. https://libguides.tru.ca/fakenews/falling
- Wardle, C. (2017, February 16). 7 Types of Mis and Disinformation. The University of Iowa LIBRARIES. https://guides.lib.uiowa.edu/c.php?g=849536&p=6077637
- Why do people fall for fake news? (n.d.). TRU Libraries. https://libguides.tru.ca/fakenews/falling
- Harrison, J. (n.d.). What is fake news? University of Victoria. https://libguides.uvic.ca/fakenews#:~:text=Some%20indications%20of%20fake%20news%20include%3A%20Information%20or,no%20credentials%20or%20relevant%20connection%20to%20the%20subject
- Stokdyk, D. (2021, April 6). What Do Programmers Do, Anyway? Southern New Hampshire University. https://www.snhu.edu/about-us/newsroom/2017/01/what-do-programmers-do
2021-7-8-1625773418