Show simple item record

dc.rights.licenseRestricted to current Rensselaer faculty, staff and students. Access inquiries may be directed to the Rensselaer Libraries.
dc.contributorSibel, Adalı
dc.contributorHolzbauer, Buster
dc.contributorXia, Lirong
dc.contributor.authorYu, Ziniu
dc.date.accessioned2021-11-03T09:10:39Z
dc.date.available2021-11-03T09:10:39Z
dc.date.created2020-04-30T11:43:06Z
dc.date.issued2018-05
dc.identifier.urihttps://hdl.handle.net/20.500.13015/2416
dc.descriptionMay 2018
dc.descriptionSchool of Science
dc.description.abstractThe polarization in politics often leads to a large number of news articles with different points of view. Some of the news contains biased or incorrect information. Some may include conspiracy theories or even made up stories. It is hard for people to decide which story is correct given that they are sometimes misled by misinformation or misrepresented opinions.
dc.description.abstractThe phenomenon of depicting a story in different ways is noticed and studied by some recent studies. In this thesis, we introduce the concept of alternative narratives in the field of news media and propose an algorithm to identify such narratives automatically. A narrative is just a way to depict an event and an alternative narrative article to a given news report is a different representation of the same story. Recent studies have shown that these type of narratives share some properties with content sharing of articles in news networks.
dc.description.abstractSo far, finding the alternative narrative articles to a given news report is often accomplished with manual tagging. As the first step to develop an algorithm capable of tracking down alternative narratives automatically, we first identify a list of feature keywords that are unique to the event in the original story. We generate the keyword list by user input, or by removing stop words from the title or choosing the top frequent entities in the text. To improve the keyword matching quality, we propose a subroutine, keyword expansion and restriction, that uses the original keyword list to generate an enhanced keyword set. After calculating the weight of these keywords using td-idf and other weighting mechanisms, we rank the articles in the target sources by assigning score to each article using keyword weights. We conduct a series of test cases to show that our algorithm performs well in different situations and analyze some possible optimal parameter values in the algorithm. However, since this is the first attempt to develop an algorithm to find such narratives, we can get a list of articles with the highest ranks being potential alternative narratives, but we are not able to set an automatic cutoff to filter the result or select articles based on the difference of their stance on specific issues. This is left for future work.
dc.language.isoENG
dc.publisherRensselaer Polytechnic Institute, Troy, NY
dc.relation.ispartofRensselaer Theses and Dissertations Online Collection
dc.subjectComputer science
dc.titleAn algorithm to find potential alternative narrative news articles
dc.typeElectronic thesis
dc.typeThesis
dc.digitool.pid179714
dc.digitool.pid179715
dc.digitool.pid179716
dc.rights.holderThis electronic version is a licensed copy owned by Rensselaer Polytechnic Institute, Troy, NY. Copyright of original work retained by author.
dc.description.degreeMS
dc.relation.departmentDept. of Computer Science


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record