=Paper=
{{Paper
|id=Vol-1388/latebreaking_paper3
|storemode=property
|title=Detecting Filter Bubbles in Ongoing News Stories
|pdfUrl=https://ceur-ws.org/Vol-1388/latebreaking_paper3.pdf
|volume=Vol-1388
|dblpUrl=https://dblp.org/rec/conf/um/TranH15
}}
==Detecting Filter Bubbles in Ongoing News Stories==
https://ceur-ws.org/Vol-1388/latebreaking_paper3.pdf
Detecting Filter Bubbles in Ongoing News Stories
Giang Binh Tran1 , Eelco Herder1
L3S Research Center, Hannover, Germany
gtran,herder@L3S.de
Abstract. In this paper, we analyze differences in perspective between timelines
created by news agencies from various countries. By employing methods for date
and headline selection, which have been extensively evaluated in previous work,
we show several types of bias that exist in the media landscape. As users typically
select only a small number of news sources to follow, they necessarily experience
at least some ‘tunnel vision’, which is commonly associated with the so-called
‘filter bubble’. By recognizing and emphasizing the peculiarities of the users’
self-selected news sources, we can help them to break out of the bubble.
1 Introduction
One of the tasks of a journalist is to monitor, gather, curate and contextualize the rele-
vant information for the target audience. He needs to go through an enormous amount
of records with information of very diverse degrees of granularity, in order to put infor-
mation into context and tell his story from all significant angles, and, at the same time,
he needs to reduce the noise of irrelevant content.
Many newspapers regularly publish manually created timelines, which allow the
reader to gain a quick overview over events that span a longer time period - such as the
Egypt revolution - and to answer questions such as: how and when did the event start?
What were the main consequences of the initial events? What happened to the main
protagonists in the event?
Though convenient for the reader, creating a timeline is expensive, as it requires
substantial expertise. Moreover, creating a timeline is a very subjective task and there-
fore huge differences can be found in expert-generated timelines, even if they are on
the same topic [3]. In addition, there exist different models of reporting, in the West-
ern world varying from the Anglo-Saxon tradition of just reporting the facts, via the
Northern-European model of combining news reporting with opinions, to the ‘polar-
ized Mediterranean’ type of reporting [2]. Moreover, it is a known fact that newspapers
in different countries provide different perspectives and points of focus, reflecting dif-
ferences in national ideologies, priorities and opinions [5].
On the Internet, the filter bubble, as coined and popularized by Eli Pariser [4], is a
well-known phenomenon that explains why personalization leads users to mainly en-
counter products, news articles and other types of content that match the users’ own
interests and viewpoints. The tunnel vision that the filter bubble is said to create, is not
just a recent online phenomenon, but inherent to journalism: editors need to be selective
and therefore necessarily focus on matters that are of highest interest to their readers.
These are some of the reasons why summaries of events, such as timelines, differ
wildly, depending on the nationality, audience, political viewpoints and other sources of
subjectivity associated with a newspaper. At the same time, readers are known to select
a small number of newspapers - or even just one - that best match their personality,
viewpoints and interests. As argued by Paul Resnick in his keynote at UMAP 20111 ,
emphasizing differences in perspective is an important instrument for allowing users to
break out of their ‘filter bubbles’.
In this paper, we investigate methods for automatically recognizing and emphasiz-
ing differences between various timelines, and evaluate them by comparing timelines
from well-known news agencies and newspapers. These methods are based on our pre-
vious work on date and headline selection for news event summarization; by employing
these methods and the Shannon Diversity Index, we are able to recognize dates that were
considered important in one timeline but not in others, and differences in news coverage
on dates that were considered important in all timelines.
The work is carried out in the context of the EUMSSI2 project, in which cross-
modal analysis techniques are developed for analysing news articles, videos and audio
reports. This will allow journalists - and media consumers - to relate these messages
with one another and to understand the underlying events.
2 Related Work
Many studies specific to timeline summarization, such as [7, 10, 8], focus on the ex-
traction of salient sentences or headlines for generating the textual content of timelines.
They assume either that the dates are given in advance or they use simple measures such
as burstiness for date selection [1].
Prior approaches dedicated specifically to date selection include work by Tran et al.
[8] and Kessler et al. [3]. They use supervised methods that score dates independently
of each other, making use of frequency-based, temporal and topical features that are
extracted from a corpus of event-related newspaper articles. We, however, score dates
jointly, making use of interactions between dates in a graphical model.
Additionally, by using the Shannon diversity index [6] on the top of the graphical
model, our approach can highlight dates on which important events happened, but that
are likely to be ignored by many news agencies. This aspect has not been considered so
far in the previous work.
3 Approach
In order to recognize and emphasize differences between timelines, it is important to
first find events and corresponding dates that are ‘subjective’, in other words dates that
are not considered important by all news agencies.
Our approach for subjective date detection makes use of a corpus of news articles
(more details in Section 4) and consists of two steps. First, we use a random walk model,
which we proposed in [9], to rank the dates based on their importance with regard to
their impact on the future events3 . After that, we re-rank the top selected dates based on
1
https://presnick.wordpress.com/2011/07/17/personalized-filters-yes-bubbles-no/
2
http://www.eumssi.eu/
3
A demo of the WikiTimes system for automatic timeline creation is available at http://
wikitimes.l3s.de/
their Shannon Diversity Index (SDI) scores. SDI is widely used in ecology and biology
to measure the diversity of species in a community [6]. We use the SDI to express
the rarity and commonness of events, as reported by different news agencies. When an
(important) event is commonly reported by many news agencies, it is less subjective
and thus less likely to be filtered out. For this reason, our approach here is to give a high
rank to the date that has a low SDI score.
Formally, in the first step, we build a date reference graph, which is a fully directed
graph G = (V, E), where V is the set of dates mentioned in any text in corpus C,
including publication dates. The edges E = {e(di , d j )} indicate that at least one text
published on di refers to the date d j .
We represent each link between events as a multi-value tuple
e(di , d j ) = (Mi j , f req(di , d j ), Itemporal (di , d j ), Itopical (di , d j ))
to integrate different measures of date importance. The first value, Mi j = N1 expresses
the prior transition probability between 2 dates where N = |V|. The other values express
the strength of the connection between di and d j , modeled by the following aspects: fre-
quency ( f req), temporal influence (Itemporal ) and topical influence (Itopical ). Our random
walk model uses these perspectives to rank the collection of dates. The intuition is that
when a date d j is referred to from either a past or future news article (published on di ),
it is likely involved in the events that are reported in that article.
In the second step, we compute the SDI for the top-K ranked dates, based on the
distribution of news articles published on those dates. For example, a date that contains
only news articles from the BBC is considered less diverse than a date that contains
news articles from several large news agencies over the world. The computation of SDI
is sketched as follows:
XR
S DI(di ) = − pi ln pi
i=1
where R is the number of news agencies from which we collected data and pi is the pro-
portion of news articles from news agencies pi . That measure quantifies uncertainty in
predicting the agency identity of a news article that is taken randomly from the dataset,
hence, suggesting how subjective (or non-diversified) a date is.
4 Experiment
We present our preliminary results of two experiments on detecting subjective dates
in the Crisis data4 dataset, which contains ground truth (GT) timelines (written by
professional journalists) as well as a corpus of around 12K news articles that cover
events that happened in the context of four news stories: Egypt Revolution, Syria War,
Libya War and Yemen Crisis. The dataset is suitable for our purpose for the following
reasons: (1) it is a heterogeneous dataset that contains news articles and expert timeline
summaries from 25 well-known news agencies and; (2) it covers long-term stories that
have been happening since 2011, making the date selection problem non-trivial for any
system.
4
The Crisis data is currently available at http://l3s.de/˜gtran/timeline/
4.1 Experiment 1: Subjective Dates Selected by One News Agency
In this experiment, we aim to detect dates that have been included in the timeline of
only one news agency (and not in other timelines). We use a cut-off K = 50 to select
the 50 most important dates, using our random walk model, and rank them by their SDI
score. We then compare these selected dates with those in our ground-truth timelines.
The performance of the subjective date detection process is measured as the proportion
of the top-10 dates (this can be extended to a larger number) that are included in exactly
one timeline.
The result is described in Table 1, which shows that a significant number of impor-
tant dates and their events are reported by only one news agency. We took a closer look
into those events (see examples in Table 2) and found evidence that the events that are
included in only one timeline are typically not about the main theme of the ongoing
news stories, but about related aspects, such as business and human rights. As we dis-
cussed in [9], timelines often contain substories or side-paths that involve major actors
of the main story. Due to space limitations, journalists can only incorporate a limited
number of substories and the decision which stories to incorporate is often subjective.
Proportion ( %)
Egypt 90%
Libya 50%
Syria 70%
Yemen 50%
Table 1: Performance of the subjective date detection process
Date Event
Egypt 29-12-2012 The Egyptian Central Bank announces that foreign reserves drained to $15 bil-
lion from $36 billion in 2010 have fallen to a “ critical minimum ” and starts
new measures to stop a sharp slide in the value of the Egyptian pound . The
pound ’s decline slows but doesn’t stop , and now stands at just over 7 to the
dollar , compared to as strong as 5.5 to the dollar in 2010 . ... (AP)
Libya 18-08-2011 Libyan Foreign Minister Moussa Koussa says the country has decided on “ an
immediate cease-fire and the stoppage of all military operations . ” But sources
inside Libya say violence continues . ....(CNN)
Syria 14-08-2012 More than 23,000 people have been killed since the outbreak of the revolt , the
Syrian Observatory for Human Rights says . Former prime minister Riad Hijab
, who defected on August 5 , says that the regime is collapsing . Government
forces concentrate their operations on the two main cities Damascus and Aleppo
. ...(DailyStar)
Yemen 18-09-2011 Security forces open fire on tens of thousands of demonstrators in Yemen’s
capital , Sana , killing at least 26 protesters in one of the bloodiest days of the
9-month-old rebellion against President Ali Abdullah Saleh ...(L.A.Times)
Table 2: Example subjective dates and the corresponding events, which are included in only
timeline in our ground truth.
4.2 Experiment 2: Important Dates that Are Not Covered
In this experiment, we aim to find out how many important dates are not included in
any of the news agencies’ timelines. Similar to the previous experiment, we picked the
top-50 important dates from the results of our random walk algorithm and calculated
the proportion of dates that are not mentioned in any timeline. The results are shown in
Table 3. The high number of ‘missed dates’ does not mean, however, that the ground
truth timelines are of poor quality: as explained earlier, journalists need to be selective
when creating timelines. The SDI values also confirm that journalists usually do a good
job in selecting the dates to be included: for all stories, except for Libya, the average
SDI of dates included in at least one timeline is higher than the average SDI of dates
that did not make it into any of the timelines. In other words, dates that are included in
at least one timeline are typically covered by more news agencies than dates that do not
appear in any timeline.
Proportion ( %) SDI included SDI excluded
Egypt 62% 2.47 1.52
Libya 24% 2.34 2.25
Syria 74% 1.65 1.70
Yemen 56% 2.10 1.70
Table 3: Proportion of important dates that are not included in any timeline, along with the average
SDI of included and not-included dates
To better understand which events were left out of the timelines, we analyzed several
of those dates with a high SDI score but that were not included. Some example events
are shown in Table 4. Except perhaps for the Libya article, these are events that readers
may be interested in, but that they will not be able to find in the timelines of the news
agencies that they are subscribed to.
Date Event
Egypt 11-02-2012 Three people including an Australian journalist, an American student and their
Egyptian guide are arrested in the city of El-Mahalla El-Kubra for allegedly
offering inducements to people to join. Labour activist Kamal al-Fayyumi is
also arrested in El-Mahalla El-Kubra...
Libya 12-06-2012 Libyan leader Muammar Gaddafi plays chess with Kirsan Ilyumzhinov , the
president of the international chess federation....
Syria 08-03-2012 Syrian dissidents reject a call by Kofi Annan to stop fighting and seek peace
talks...
Yemen 29-11-2011 Opposition candidate Mohamed Salem Basindwah was chosen to lead the na-
tional unity government , and pledged on Nov. 29 to tackle issues including fuel
shortages...
Table 4: Example of dates and corresponding events that did not make it to any of our ground-
truth timelines.
5 Discussion and Conclusions
In this paper, we have shown that it is possible to automatically find dates and corre-
sponding events in news stories that are likely to be covered in only one or a subset of
manually generated timelines. Further, we have shown that there are important dates
that are not incorporated in any timeline.
As we argued earlier, it is unavoidable that timelines are selective and, to a certain
extent, subjective. Still, it would be very desirable to make readers aware of the pecu-
liarities of the timeline that they currently inspect. This can be achieved by highlighting
dates with large differences in coverage between news agencies or timelines, by adding
links to other timelines for dates that are not incorporated in the current timeline, or,
alternatively, by constructing an annotated ‘timeline of timelines’ for those who wish to
contrast the bias in their personal selection of news sources with news sources that they
usually do not visit.
Conversely, journalists will be able to create better balanced - or, alternatively, even
more argumentative - timelines with feedback on their current selection of dates and
events. Even though - as we have shown in previous work - it is possible to automati-
cally construct timelines by selecting the most relevant dates and headlines, still manual
processing and editing would be needed to enhance the communicative qualities of the
timeline, and to adapt it to the needs of the readers.
6 Acknowledgments
The work was partially funded by the European Commission for the FP7 project EUMSSI
(611057)
References
1. H. L. Chieu and Y. K. Lee. Query based event extraction along a timeline. In Proceedings
of SIGIR’04, pages 425–432, 2004.
2. F. Esser and A. Umbricht. Competing models of journalism? political affairs coverage in us,
british, german, swiss, french and italian newspapers. Journalism, 14(8):989–1007, 2013.
3. R. Kessler, X. Tannier, C. Hagege, V. Moriceau, and A. Bittar. Finding salient dates for
building thematic timelines. In Proceedings of the 50th Annual Meeting of the Associa-
tion for Computational Linguistics: Long Papers-Volume 1, pages 730–739. Association for
Computational Linguistics, 2012.
4. E. Pariser. The filter bubble: What the Internet is hiding from you. Penguin UK, 2011.
5. S. Seo. Hallidayean transitivity analysis: The battle for tripoli in the contrasting headlines of
two national newspapers. Discourse & Society, 24(6):774–791, 2013.
6. C. E. Shannon. A mathematical theory of communication. SIGMOBILE Mob. Comput.
Commun. Rev., 5(1):3–55, Jan. 2001.
7. R. C. Swan and J. Allan. Timemine: visualizing automatically constructed timelines. In
SIGIR, page 393, 2000.
8. B. G. Tran, M. Alrifai, and D. Quoc Nguyen. Predicting relevant news events for timeline
summaries. In WWW, 2013.
9. G. Tran, E. Herder, and K. Markert. Joint graphical models for date selection in timeline
summarization. In Proceedings of ACL, 2015.
10. R. Yan, X. Wan, J. Otterbacher, L. Kong, X. Li, and Y. Zhang. Evolutionary timeline sum-
marization: a balanced optimization framework via iterative substitution. In Proceedings of
SIGIR’11, pages 745–754, 2011.