research question

Analyzing the most commented categories how do people feel about it?

v11svg-ch 10 20 5 2 1 INTERVIEW MOVIE EXTRACT NEWS REALITY SHOW REPORTAGE SPEECH TRAILER Ratio Amount of videos Logarithmic scale 1000 100 10 1 AMATEUR DOCUMENTARY EDUCATION INTERVIEW NEWS REPORTAGE SPEECH TV SHOW VLOG Ratio Amount of videos Logarithmic scale USA YOUTUBE Amount of comments par category (ratio) CHN Baidu video and Bilibili Amount of comments par category (ratio) USA YOUTUBE Comments tone analysis par category CHN Baidu video and Bilibili Comments tone analysis par category AMATEUR INTERVIEW REPORTAGE VLOG ANALYTICAL ANGER CONFIDENT FEAR JOY SADNESS TENTATIVE Hover on the flows to hilight MOVIE EXTRACT TRAILER ANALYTICAL ANGER CONFIDENT FEAR JOY SADNESS TENTATIVE Hover on the flows to hilight INTERVIEW

Description

Here we decided to focus on the most commented videos in order to uncover all the different users’ inclinations and their reactions to the video contents.

For the Chinese part we just considered all the results we collected (relating to each particular category) since the general amount was not that huge. On the other hand, relating to Youtube videos we had to someway reduce the amount of comments to better focus on them. This meaning that we considered the first five most commented videos and scraped them analyzing only the comments with more than a hundred likes. This for a specific reason: choosing the most shared opinions.

Protocol

prot11 - filter: most commented- considering the first 5 rows - all the comment with more than 100 likes - new dataset with all the comments together RAW GRAPHS VIZ Illustrator(refining) VISUALIZATION Analizing the most commented categories how do people feelabout? MOSTCOMMENTEDVIDEOS QUERY JSON CONVERTER U.S.NORTH KOREANDEFECTORS 脱北 CHINA Excel(cleaning) CORPUS DEFINITION BBEdit(claning+merging) Excel(claning+merging) English translation YTDTVideo Infoand Comments Web Scraper Excel .json Tone analyzer

We used the DMI tools to scrape all the Youtube comments while for Baidu Video and Bilibili we used Webscraper when possible. After translating all the Chinese results we separately analized the tone to better understand how the two different points of view.

For evaluating the comments tone we employed Tone Analyzer tool which detects joy, fear, sadness, anger, analytical, confident and tentative tones after driving a linguistic inspection through the text.

Data

Timestamp: 11/2016 - 11/2017

Data source: Youtube, Baidu Video,Bilibili

We organized the excel file filtering the most commented videos and among them the most liked comments. Since the considered comments related to the categories in different amounts we proportioned them in order to make them comparable. To do so we calculated the ratio dividing the number of comments by the number of video in each category. As a result we decided to tone-inspect all the videos with the highest ratio-volume. After converting the .json outputs of the tone analysis we manually cleaned them and built another dataset.