Skip to content

Latest commit

 

History

History
32 lines (24 loc) · 1.32 KB

README.md

File metadata and controls

32 lines (24 loc) · 1.32 KB

matlabfakenews

an attempt to tackle the fake news problem using MATLAB

we broke the problem into a classification problem to determine whether articles were related to or biased against a certain headline.

main endpoints

demoObjects.mat contains the MATLAB objects for the Naive Bayes Classifier and the sentiment map.

doesAgree.m takes a headline and a body as strings (along with the learned model and sentiment map), and determines which of the four categories it falls under: unrelated, agree, disagree, or discuss

checkAllAgree takes a table with a Headline field, an articleBody field, and an observed stance field, outputting a table of results with the given stance along with the computed stance

getGrade takes these results and calculates a percentage score for the number of correctly classified (headline, article) pairs.

getFullGrade runs everything without the need to load anything. This will give the full score as long as the demoObjects.mat and competition test csvs are in the folder

competitionResults.mat contains the results from running getFullGrade over the whole competition dataset.

improvements

we should consider using ml to categorize the relevant articles into the correct bias categories. This involves finding another relevant input parameter. Since headlines are rather short, this is rather difficult.