Exploring the world of Philippine online news

Databeers Manila
@tjpalanca

Transcript

1. Exploring the world of Philippine online news Databeers MNL | 27 Jul 2017 @tjpalanca | tjpalanca.com Photo: Jason Howie/Flickr, CC BY 2.0
2. About TJ Palanca Data blogger @ tjpalanca.com Global data analyst @ uber Views expressed here do not necessarily reflect those of my employer or other organisations with which I have been affiliated.
3. Facebook has increasingly become the source of news for many Filipinos Not all news articles are created equal, and which ones vary depending on where you are standing Traditional news outlets have come under fire for alleged ‘biased reporting.’ This is not a new issue It’s key to understand what news outlets report on and how we interact with it!
4. How are the news articles that news portals post onto their Facebook pages distributed? How do people interact with these articles and topics? How do news articles construct their headlines with respect to these topics? 1 2 3
5. https://facebook.com/ abscbnNEWS gmanews ANCalerts CNNPhilippines dzmmteleradyo rapplerdotcom inquirerdotnet InterAksyon News5Everywhere ThePhilippineSTAR Yahoophilippines ……… MSNPhilippines We gathered data from 22 leading news pages in the Philippines from the Facebook Graph API Man faces raps after bullet from own gun kills sister, James Reid had an 'amazing' New Year's Eve, IN PHOTOS: Meet the newest Miss Tourism Queen of the Year, 7- anyos na batang binaril ng brgy. tanod sa Makati, pumanaw na, 11 killed in Sulu clash, military says, Athletes to watch out for in 2016, Philippine basketball’s All- Star stories from 2015 (Part 2), BIFF attacks military detachments on New Year's Eve, Oil firms slash LPG prices, LOOK: Solenn, husband share New Year's Eve kiss, How Pinoy lensman survived Dubai inferno, The year of kilig: Top Entertainment stories of 2015, Miss Colombia speaks up: I was humiliated, MMFF 2015 entries make P640-M in one week, IN PHOTOS: How celebrities spent New Year's Eve, New Year revelry leaves 500 tons of garbage in Manila, Why Kobe Paras feels like the 'luckiest man in the world', Youth group ends Kalayaan islands protest, LOOK: Pia Wurtzbach spends New Year's Eve in Times Square, China's new two-child policy law takes effect, The 20 restaurants that won 2015, Abe pledges to keep Japan out of war, Pangasinan town welcomes New Year with 'bongbongs', LOOK: Vice Ganda takes 'selfie' with Roger Federer, Longer blackouts loom over Mindanao in 2016, 'One hour, then that's it': Dubai fire survivor hangs from balcony, 'IS suicide attack' planned in Munich: German officials, 'Termite gang' sumalakay sa pawnshop, George Lucas says sorry for Disney 'white slavers' slam, Fires hit NCR in New Year revelries, Global markets end 2015 with a whimper, Fears of more death, devastation in flood-ravaged Missouri, Firecracker, stray bullet injuries down 53 pct - DOH, Binatilyo patay sa pamamaril sa Bagong Taon, WATCH: Tayag dances after DOH reports gains vs firecrackers, LOOK: NPA releases captive, Boy in coma, man critical after Makati shooting, LOOK: Kean Cipriano marries Chynna Ortaleza, How Davao welcomed New Year with no firecrackers, Fire victims in Cebu celebrate New Year in evacuation center, Thousands welcome New Year in Rizal Park, LOOK: New Year baby born in Cagayan de Oro named 'Christ', PANOORIN: New Year baby ng Fabella hospital, Belgium holds five over 'New Year plot', Philippine basketball’s All-Star stories from 2015 (Part 1), World welcomes New Year despite terror fears, INC breaks 3 Guinness records on New Year's Eve, Miss Universe Pia Wurtzbach sa NYC magba-Bagong Taon, Lalaki, sinindihan at niyakap ang ‘Goodbye PH’, patay, Watusi causes fire, 3,000 families lose homes, Families troop to Quezon Memorial Circle for New Year countdown, 2015: The Year That Was in Combat Sports (Part 1), Sounds from the sky: Top Region stories for 2015, Why the 2016 election campaign will look different, Year-end special: The most intriguing and entertaining Pinoy sports stories of 2015, Restoration of bombed NGCP tower halted due to land dispute, NPA releases captive soldier on New Year's Eve, 2015: The year in review, Fireworks in Sydney, 10 steps to financial independence in 2016, Changed lives, viral stories: How they unfolded on social media in 2015, Friends, family hold vigil for Fil-Am killed in fire, Fire engulfs Dubai hotel near world's tallest building, Hong Kong leader blames Facebook 'hack' for racy pics, Hopes, tragedies, and controversies: Top Global Filipino stories in 2015 #2, Israel sorry after diplomat used Singapore flag as table cloth, Filipina beauty wins Miss Tourism Queen of the Year International 2015, Being No. 1 draft pick 'not a target' for Kiefer, Belgium charges 10th suspect over Paris attacks, media.giphy.com, From Pope to Pia: Stories you liked in 2015, Davao singer belts out Whitney hit in 'Tawag ng Tanghalan', Pacquiao wants one more shot at Floyd before retiring, From pomp to pageantry: Images of 2015, 7 kilig moments between Kean Cipriano, Chynna Ortaleza, Vice Ganda grateful for overwhelming support, Butler providing the spark for Bulls in Rose's absence, It's official! Mariel, Amy join 'It's Showtime', Pope calls for end to injustice, apathy, Fire incidents on January 1, far lower in 2016, Obama calls for action vs gun violence in New Year message, Firecracker injuries hit 458: DOH, 'Nakakatakot!' Vice Ganda recalls witnessing Dubai fire, 1 dead, 41 injured as bus slams into Quezon resto, China restructures military as Xi eyes 'strong army', WATCH: Colombian burns Pia Wurtzbach effigy, Pauleen, Vic have house blessed on New Year's Day, Heat cool off Mavs as Wade contributes off the bench, LOOK: Derek Ramsay, girlfriend mark anniversary, Don't procrastinate: Tips on how to manage your time well, Former LTO chief Virgie Torres dies, Bangs Garcia engaged to British boyfriend, Timeline Photos
6. We applied a topic modelling technique known as Latent Dirichlet Allocation (LDA) Data used was from the headline and part of the blurb that is shown to the user after the headline. Any comments made by the page itself are also included. Only single words are used. Preprocessing steps included stopword removal, stemming/lemmatization for English words, spelling corrections, and removal of non-meaningful words. LDA number of topics k was chosen using cross validated rate of perplexity change.
7. Entertainment Filipino Language Sports News 1. Topics were derived through Latent Dirichlet Allocation (LDA) with 40 topics 2. For each topic, the top 30 words in terms of relevance as defined in Sievert and Shirley (2014) are displayed 3. The centres of each topic mass were determined by applying principal components analysis to the distance matrix, with the Jensen-Shannon divergence as the distance measure. This layout is intended to display the degree of semantic distance.
8. 1. Topics were derived through Latent Dirichlet Allocation (LDA) with 40 topics 2. For each topic, the top 30 words in terms of relevance as defined in Sievert and Shirley (2014) are displayed 3. The centres of each topic mass were determined by applying principal components analysis to the distance matrix, with the Jensen-Shannon divergence as the distance measure. This layout is intended to display the degree of semantic distance.
9. 1. Topics were derived through Latent Dirichlet Allocation (LDA) with 40 topics 2. For each topic, the top 30 words in terms of relevance as defined in Sievert and Shirley (2014) are displayed 3. The centres of each topic mass were determined by applying principal components analysis to the distance matrix, with the Jensen-Shannon divergence as the distance measure. This layout is intended to display the degree of semantic distance.
10. Trending news, morph to engagement (hot topics) Articles Trend over time
11. Reactions Trend over time
12. We’re only using word counts, though! We’re not actually learning and structure or meaning from the data. typhoon, storm, and weather are all slightly related, but are considered similarly as any other word in a bag of words model. x + y = 3 x - y = 5 Why are we adding letters together? Letters don’t belong in math! We’re going to literally add and subtract words! This is why we’re no longer friends.
13. Word vectorization (word2vec) Word vectorization allows us to encode more meaning into numerical representations by looking at the context of the word. “You shall know a word by the company it keeps.” J.R. Firth 'No window hours' traffic scheme gets mixed reactions from netizen NLEX, MIAA report lighter traffic on Monday Passenger traffic in Batangas drops VIRAL: Boy rides toy car into oncoming traffic in China Coming soon: Moms as traffic enforcers in Manila arusa sa pasaway: Manila to axe erring traffic enforcers Duterte says traffic jams come with infrastructure rebuild ???
14. Some cool analogies Duterte - President + Robredo = VP Philippines - Manila + China = Beijing Congress - Laws + Courts = Decision
15. Comparing content sets ABS CBN News - Rappler = DZMM Teleradyo
16. Comparing content sets ABS CBN News - GMA News = Inquirer
17. Thank you! tjpalanca/ facebook-news-analysis mail@tjpalanca.com https:// tjpalanca.com Data Roles @ Uber! MNL/SIN/SFO