Identify trust value for fake news detection from a social media

TSEGU, EUAEL

dc.contributor.author	TSEGU, EUAEL
dc.date.accessioned	2020-03-16T09:10:16Z
dc.date.available	2020-03-16T09:10:16Z
dc.date.issued	2020-03-16
dc.identifier.uri	http://hdl.handle.net/123456789/10347
dc.description.abstract	Trust analysis on web sources is mostly done using methods like account-based, content-based and activity of the source. In account-based attributes like followers’ type, the number of followers, number of likes, type of follows and friends are used to analyze the trustworthiness of the source. When using content-based method attributes like Hashtag (#), includes (@) and links that are found on the post are used to analyze the trustworthiness of the source. When using the activity of the source to analyze its trustworthiness attributes like how many times it posts, the time interval of posts is used. Those methods are useful for posts that are used for marketing purpose adverts or for spammers. When it comes to real liars, intentionally want to deceive people and hence, it is hard to use those methods for analyzing the trust value of the sources. Because those people try their best, to make their post genuine and they don’t make things that makes them look like a spam. In this research, we are using text content to extract the fact and check if the fact is valued and try to calculate the validity of the facts in order to calculate the trust value of the source. We develop a method that uses sentiment analysis, knowledge base, and a voting system. The source sentiment to a specific entity is used for calculating the sensationalism of a source to an entity. This is used as a bias in the calculation the trust value of the source with coordination of knowledge base which tries to analyze the past trust value history of the source and the voting system which is used for cross-checking with other sources. From this research, we were able to get an accuracy of 44.3 % for fake news and 79.3 % for genuine news based on an initial trust value of 0.5. the genuine source accuracy is higher than the fake sources because the genuine sources were small in number but their number of posts were higher than the fake news on the entity Donald Trump, whereas the fake source were large in number but they don’t post more than 3 or 4 posts. We observe from the data that fake sources write about diverse topics even though they don’t have the knowledge or information. This is why they don’t have consistency in their posts.	en_US
dc.language.iso	en	en_US
dc.subject	Computer Science	en_US
dc.title	Identify trust value for fake news detection from a social media	en_US
dc.type	Thesis	en_US