News Headline Comparisons
Ryan Weber
May 13, 2018
- News website comparison
- Example comparisons
- Interested in what stories are prioritized in real-time
- Typical front pages
- Look at word frequencies
- Look sentiment by day
- What is impacting this coverage primarily?
- After filtering out “trump”, positive sentiments
- With “trump” filtered out, how have the by day sentiment differences changed
- Sentiment differences by hour
- Negative sentiments
- What were the news stories on Saturday that were read as notably negative on CNN
- See which words are unique to each (tf-idf analysis)
- How do these results compare against the Google News Api results for the same sites over this time period?
- Comparing the overall headline results from this period (against the top headlines results) for each site reveals a marked difference from the earlier results seen above:
- Though some distinctions in overall sentiment were observed, an attempt to categorize the headlines into three distinct groups using Document-Topic Probabilities, was not at all conclusive:
News website comparison
Primary question: Using resources covered in this course, are there any quantifiable distinctions that can be made between information prioritized on sites across the political spectrum?
- Fox News
- CNN
- MSNBC
Example comparisons
Interested in what stories are prioritized in real-time
Approach:
Take the top three headlines from MSNBC, CNN, and Fox News websites over four days and analyze the results (scraped from each site at 5-10 minute intervals).
Compare against all headlines from these sites over the same time period to determine if there’s any distinctions that can be found between the sites.
Note, in this approach, since headlines are collected at 5-10 minute intervals throughout the day, if one headline stays up for a long time, this weighting will be reflected in the results, as it will appear one time in the data set for each time is was on a site when scraped.
Typical front pages
Fox news
“Kicker” elements precede each main headline, and may have a more emotional appeal than the article headlines themselves
CNN
Often one primary headline, then one or two additional featured stories
MSNBC
List of headliness, each associated with large image associated with video or article
Look at word frequencies
In coverage, “trump” is not surprisingly the most frequent headline word.
- CNN may emphasize narratives regarding “giuliani” and “iran”
- Fox has notable coverage of “rep” “waters”.
- MSNBC highlights coverage of “avenati”.
Look sentiment by day
What is impacting this coverage primarily?
After filtering out “trump”, positive sentiments
Associations seem reasonable
With “trump” filtered out, how have the by day sentiment differences changed
Relative difference is largely similar, but now all stories are seen as primarily “negative” (“No news is good news”.
Sentiment differences by hour
Trend is less obvious
Negative sentiments
What were the news stories on Saturday that were read as notably negative on CNN
| Text | word | |
|---|---|---|
| 1 | US policy shifts on Mideast increase tensions in a region on edge The State Department has been bolstering security at US diplomatic posts around the Muslim world in advance of the embassy move, anticipating violence | tension |
| 1350 | US policy shifts on Mideast increase tensions The State Department has been bolstering security at US diplomatic posts around the Muslim world in advance of the embassy move, anticipating violence | tension |
| 3551 | US policy shifts on Mideast increase tensions The State Department has been bolstering security at diplomatic posts around the Muslim world in advance of the embassy move, anticipating violence | tension |
| 5042 | World to see North Korea blow up nuclear test site The pledge to shutter the site and stop unannounced missile tests may be signs that the isolated nation is prepared to engage with others | blow |
| 5826 | Giuliani says Trump ‘denied’ a corporate deal, then backtracks The back-and-forth reignites controversy over possible political interference and sparks chatter about Giuliani making things more complicated for the President | chatter |
| 5853 | World to see North Korea blow up its nuclear test site | blow |
| 5951 | Giuliani says Trump ‘denied’ a corporate deal, then backtracks The back-and-forth reignites controversy over possible political interference and sparks chatter about Rudy Giuliani making things more complicated for Trump | chatter |
| 7076 | Mulvaney: McCain joke by White House staffer was awful, but not fireable offense | joke |
| 7208 | Mulvaney: McCain joke by White House staffer was awful, but not a fireable offense | joke |
| 9250 | Sanders berates staff over leak The White House press secretary ripped into the communications staff after an aide’s comment about John McCain was leaked | leak |
| 9251 | Sanders berates staff over leak The White House press secretary ripped into the communications staff after an aide’s comment about John McCain was leaked | rip |
| 11160 | 1 person killed and 8 injured in Paris knife attack | kill |
| 11161 | 1 person killed and 8 injured in Paris knife attack | knife |
| 11162 | 1 person killed and 8 injured in Paris knife attack | attack |
| 11492 | At least 1 person killed in deadly stabbing attack in Paris | kill |
| 11493 | At least 1 person killed in deadly stabbing attack in Paris | attack |
| 11495 | At least 1 person killed in deadly stabbing attack in Paris | stab |
| 11714 | Deadly knife attack in Paris At least 1 person was killed and 2 ‘gravely injured,’ police say. The attacker was stopped by police. | kill |
| 11715 | Deadly knife attack in Paris At least 1 person was killed and 2 ‘gravely injured,’ police say. The attacker was stopped by police. | knife |
| 11716 | Deadly knife attack in Paris At least 1 person was killed and 2 ‘gravely injured,’ police say. The attacker was stopped by police. | attack |
| 11893 | Sarah Sanders berates staff after aide’s comment about John McCain was leaked | leak |
| 14388 | Deadly knife attack in Paris At least 1 person was killed and 2 ‘gravely injured,’ police say. The attacker was killed by police. | kill |
| 14389 | Deadly knife attack in Paris At least 1 person was killed and 2 ‘gravely injured,’ police say. The attacker was killed by police. | knife |
| 14390 | Deadly knife attack in Paris At least 1 person was killed and 2 ‘gravely injured,’ police say. The attacker was killed by police. | attack |
| 15305 | 1 killed, 2 gravely injured in knife attack in Paris | kill |
| 15306 | 1 killed, 2 gravely injured in knife attack in Paris | knife |
| 15307 | 1 killed, 2 gravely injured in knife attack in Paris | attack |
| 20465 | Suspect in deadly Paris knife attack shouts ‘Allahu Akbar,’ before police stop him | knife |
| 20466 | Suspect in deadly Paris knife attack shouts ‘Allahu Akbar,’ before police stop him | attack |
| 21211 | Suspect in deadly Paris knife attack shouts ‘Allahu Akbar,’ before police kill him | kill |
| 21212 | Suspect in deadly Paris knife attack shouts ‘Allahu Akbar,’ before police kill him | knife |
| 21213 | Suspect in deadly Paris knife attack shouts ‘Allahu Akbar,’ before police kill him | attack |
| 31808 | China’s navy takes historic step The first domestically built aircraft carrier heads out for a sea trial, another step in the country’s mission to rival world maritime powers | rival |
In what sense are “attack” or “blow” up negative. Perhaps if they appeal to a sense of fear in the reader.
In this sense, they seem appropriately classified in this data set to understand how the headlines may be operating on an emotional level.
See which words are unique to each (tf-idf analysis)
Example of where some of these words are coming from
kable(scrapedDf %>%
filter(site == "fox",
str_detect(str_to_lower(Text), "hillar")) %>%
count(Text))| Text | n |
|---|---|
| ‘VILE’ ATTACK Hillary Clinton crony slammed for mocking Trump Jr.’s wife, Vanessa | 6 |
| ‘VILE’ ATTACK Hillary Clinton crony slammed for tweet aimed at Trump Jr. | 37 |
| GRANITE STATE STAMPEDE With Hillary off radar, Democrat wannabes beat path to New Hampshire for Trump 2020 showdown | 6 |
| THE HILLARY FIX? Blue states lead charge to dump Electoral College, back popular vote | 1 |
| THE HILLARY FIX? Democrat states lead charge to dump Electoral College, back popular vote | 59 |
kable(scrapedDf %>%
filter(site == "fox",
str_detect(str_to_lower(Text), "maxine")) %>%
count(Text))| Text | n |
|---|---|
| MAXINE’S OUTRAGE Waters to GOP rep: ‘I resent the remark about making America great again’ | 86 |
kable(temp <- scrapedDf %>%
filter(site == "cnn",
str_detect(str_to_lower(Text), "product")) %>%
count(Text))| Text | n |
|---|---|
| Ford suspends F-150 production | 62 |
| Ford suspends production of F-150 trucks | 19 |
| Ford suspends production of the truck America loves | 36 |
| Ford suspends F-150 production A fire at a supplier has left both factories that produce the nation’s best-selling vehicle without parts | 9 |
kable(temp <- scrapedDf %>%
filter(site == "cnn",
str_detect(str_to_lower(Text), "world")) %>%
count(Text))| Text | n |
|---|---|
| The world is nervously watching as the gloves come off between Iran and Israel | 21 |
| World to see North Korea blow up its nuclear test site | 90 |
| China’s navy takes historic step The first domestically built aircraft carrier heads out for a sea trial, another step in the country’s mission to rival world maritime powers | 7 |
| Hope for the world’s last Cold War standoff Trump said the 3 prisoners’ release by North Korea was positive, but history suggests skepticism is warranted | 16 |
| Hope for the world’s last Cold War standoff Trump said the 3 prisoners’ release was a positive sign, but history suggests skepticism is warranted | 3 |
| King of the world Trump takes his political chaos theory beyond US to rack up wins for his political base | 1 |
| North Korea summit set for Singapore ‘We will both try to make it a very special moment for World Peace!’ President Trump tweeted Thursday of the meeting scheduled for June 12. | 10 |
| The world watches nervously as Israel and Iran trade blows There is a possibility Trump’s withdrawal from the Iran nuclear deal is greasing the wheels to a regional war | 6 |
| Trump remakes the world in his image Trump takes his political chaos theory beyond US to rack up wins for his political base | 16 |
| Trump reshapes the world to win at home Trump takes his political chaos theory beyond US to rack up wins for his political base | 1 |
| Trump rules the world Trump takes his political chaos theory beyond US to rack up wins for his political base | 2 |
| Trump shakes up long-standing global forces The President’s often head-spinning imprint on the world is delighting his supporters and alarming critics | 1 |
| Unbridled Trump storms global stage The President’s often head-spinning imprint on the world is delighting his supporters and alarming critics | 11 |
| US policy shifts on Mideast increase tensions in a region on edge The State Department has been bolstering security at US diplomatic posts around the Muslim world in advance of the embassy move, anticipating violence | 19 |
| US policy shifts on Mideast increase tensions The State Department has been bolstering security at diplomatic posts around the Muslim world in advance of the embassy move, anticipating violence | 21 |
| US policy shifts on Mideast increase tensions The State Department has been bolstering security at US diplomatic posts around the Muslim world in advance of the embassy move, anticipating violence | 31 |
| World nervously watches as the gloves come off between Iran and Israel There is a possibility Trump’s withdrawal from the Iran nuclear deal is greasing the wheels to a regional war | 4 |
| World to see North Korea blow up nuclear test site The pledge to shutter the site and stop unannounced missile tests may be signs that the isolated nation is prepared to engage with others | 8 |
kable(temp <- scrapedDf %>%
filter(site == "msnbc",
str_detect(str_to_lower(Text), "schneiderman")) %>%
count(Text))| Text | n |
|---|---|
| Cohen, AG Schneiderman cases collide in court filing | 13 |
| How Cohen case intersects with Schneiderman abuse | 111 |
kable(temp <- scrapedDf %>%
filter(site == "msnbc",
str_detect(str_to_lower(Text), "pimp")) %>%
count(Text))| Text | n |
|---|---|
| Giuliani on Avenatti: I don’t get involved with pimps | 110 |
- CNN seems to be only site emphasizing fire at Ford factory.
- Fox emphasizes stories on Hillary and Maxine Waters, as well as having an emphasis on emotional words in this tf-idf collection (“slam”, “fight”, “resent”, “outrage”, “tasteless”).
- MSNBC places an emphasis on Mueller, Avenati, and, somewhat surprising for the ‘liberal’ site in this grouping, notably covered the abuse scandals of the democrat Eric Schneiderman, who recently resigned as Attorney General of New York. Looking at the results above, the majority of this coverage regarded it’s relation to the Michael Cohen case, which this site emphasizes more than the others.
How do these results compare against the Google News Api results for the same sites over this time period?
Overall headlines
Top headlines
Relatively similar, though as expected, single news items don’t carry the same weight in the analysis.
For instance, “giuliani” does not have the same emphasis for CNN and Maxine “water”s does not have the same emphasis on Fox.
Similarly, “russia” is seen in the more general results on MSNBC, where as “probe” and “giuliani” are more common on the top headline results.
Notably, the number of available stories from the Google API for MSNBC over this time period was almost half of those available for the other two news outlets, perhaps explaining the closer similarity between the two results here.
Comparing the overall headline results from this period (against the top headlines results) for each site reveals a marked difference from the earlier results seen above:
Overall headlines
Top story headlines
In general, we can see that the top headlines looked to be much more negative.
Also, the difference between CNN and Fox News that had been seen in the top headlines, does not appear to be present in the overall headlines for this time period.
Further, we can see in the CNN results the outsized effect that one or two particular “negative” stories can have (Saturday for CNN) when weighting their effect for presence (as persisting top headlines on the website).
Though some distinctions in overall sentiment were observed, an attempt to categorize the headlines into three distinct groups using Document-Topic Probabilities, was not at all conclusive:
