Computer Science has a ected almost all elds of human knowledge, contributing to scienti c advances in many branches of Natural and Social Sciences. Journalism is one of the elds that is bene ting of the advance of computer science. Among the journalistic concepts that can be analyzed computationally is News Value. Novelty is one of the most important news value. A possible approach to get novelty elements in a story considers word frequency, through of the capacity to collect and analyze massive amounts of data. In this paper, we use the News Coverage Index dataset (NCI), maintained by the Pew Research Center, to analyze the novelty dynamics of news coverage, using the novelty signatures proposed by [12]. As a de nition of novelty, we used the rst appearance of a new lead newsmaker. Results show a good t of the model to the dataset. Furthermore, an analysis by media sector and broad topic shows interesting insights for the analysis of media coverage.
The Computational Science has a ected almost all
elds of human knowledge, contributing to scienti c
advances in many branches of Natural and Social
Sciences. For instance, the capacity to collect and analyze
massive amounts of data has transformed intensely
elds such as biology and physics [
In Social Science, despite the di culties to formalize
computationally many scienti c subjects of the human
behavior, \a computational social science is
emerging that leverages the capacity to collect and
analyze data with an unprecedented breadth and depth
and scale" [
Among the journalistic concepts that can be
analyzed computationally is News Value. News value as
a concept was thought by Johan Galtung and Mari
Holmboe Ruge's seminal publication in the Journal of
Peace Research. In 1965, the paper suggested a range
of attributes that establish news values in discursive
elements contained in newspapers and broadcast news.
Galtung and Ruge established the news values
elements as Frequency; Threshold; Unambiguity;
Meaningfulness; Consonance; Unexpectedness; Continuity;
Composition; Reference to Elite Nations; Reference to
Elite People; Reference to Persons; and Reference to
Something Negative [
News values are studied considering a range of
attributes contained in discursive elements. It is also
possible to verify the news value through a range
of \more speci c cognitive constraints that de ne
news values (Novelty, Regency, Presupposition,
Consonance, Relevance, Deviance and Negativity,
Proximity) [
The novelty can be understood by concepts as out
of the ordinary, least expected, or not predicted, news
values relating to the novelty, newness or
unexpectedness of an event/happening [
As previously observed, novelty is also elaborated
\mainly through using evaluative language, references
to surprise/expectations and comparisons" [
In this paper, we used a recent model of novelty dynamics to analyze news coverage. The main idea is to analyze whether di erent news sources present di erent novelty dynamics. This paper is organized as follows: Section 2 presents novelty signatures that emerge in some dynamical processes. Section 3 describes the data set used in our study. Section 4 presents the results of applying the novelty signatures to the NCI dataset, and Section 5 concludes the paper.
Tria et. al [
They were able to model novelty as a simple
mathematical model based on random draws sampling with
replacement of an Urn [
The rst signature is based on quantifying the rate at which novelties occur in a temporally ordered sequence of elements of length N by analyzing the growth of the number D(N ) of distinct elements in this sequence. This relation would imply in a Heap's law, which states that the rate at which novelties occur decreases over time as t , where is the coe cient of a power law distribution of D(N ) over N tted over the data.
The second signature is related to the frequency of occurrence of di erent elements in the data. The frequency-rank distribution would follow an approximate Zip an distribution (Zipf's law). In this distribution, the frequency of any element is inversely proportional to its rank in the frequency table, i.e., the frequency F (R) of an element at rank R is proportional to R , where is the coe cient of a power law distribution of F (R) over R tted over the data.
It is well known that and are inversely
correlated [
In our analysis, we used the data gathered by the Pew Research Center1. Every week, this institution produced the News Coverage Index (NCI) by identifying and annotating the main subjects covered by the U.S. mainstream media. The dataset used this research is the most updated dataset (2013), published by Pew Research Center. Until this moment, no other similar dataset that can be used to update the data or serve to comparison.
The NCI captured and analyzed 52 news
outlets in real time to determine what was
being covered and what was not in the U.S.
news media. The analysis was conducted
weekly, Monday - Sunday. The key variables
included source, story date, big story, broad
story topic, placement, format, geographic
focus, story word count, duration of
broadcast story and lead newsmaker. The outlets
studied came from print, network TV, cable,
online, and radio. They included evening and
morning network news, several hours of
daytime and prime time cable news each day,
newspapers from around the country, the top
online news sites, and radio, including
headlines, the long form programs and talk [
By focusing on the topic of the story, the index measures by what percentage of the analyzed news hole is about that topic. Data were collected from January 2007 to May 2012. Table 1 presents the number of news stories collected per year. Note that the year 2012 has a few stories because the collection period ranges from January to May, rather than January to December. The codebook includes variable names, de nitions, applicable procedures and changes that were made to certain variables. For each story, it was annotated the date, source, broadcast start time (morning, noon, afternoon, evening and night, or not broadcast), duration in seconds, word counts, placement prominence, story format, big story, geographic focus (local, US national, US international, non-US international), broad 1Formerly Project for Excellence in Journalism (PEJ). story topic, media sector (cable TV, network TV, newspaper, online and radio), and lead newsmaker. The number of outlets and individual programs vary considerably within each media sector, as do the number of stories and size of the audience.
The index is a good source for analyzing, through time, how stories emerge and sink. Other possibilities include how the character or narrative focuses of the story change and how much of the broad topic's categories get more coverage, when compared to the others. However, the index does not provide information for additional possible questions, such as tone, sourcing or other matters.
The key variable chosen in this study was \lead newsmaker", a variable that \determines the person whose actions or statements constitute the main subject matter of the story". In the NCI, the derivation of the \lead newsmaker' variable used a methodology that examined the outlets daily by the coding team. The researchers establish as a de nition: variable lead newsmaker determines the person whose actions or statements constitute the main subject matter of the story discussed with at least 50% of the story (in time or space).
Therefore, in our analysis, a news story is agged as a novelty whenever the rst appearance of a new lead newsmaker occurs, considering an ordered sequence of histories by date in the NCI. Obviously, this approach does not completely capture all the aspect of novelty in news coverage. It is perfectly possible (and indeed very common) that some new factor is being published by some lead newsmaker who appeared before. However, this approach does capture some aspect of novelty, in a sense that di erent subjects are being noticed in the media. Furthermore, the approach sheds some interesting insights, as discussed next. 4
In this section we present the results of the novelty
signatures as proposed by [
Figure 1 shows the two novelty signatures for all
stories in the NCI dataset, for the Heaps' law and Zipf's
law, respectively. The graphs show a very good t (the
blue line in the graphs), indicating that the dynamic
of novelties also follows the model proposed in [
This is an interesting result per se, but we can move beyond that by conditioning the analysis by some news groups. Figure 2 does this, where we have split the analysis by the media sector (newspaper, online, radio, 101 102 103 104 105 0 10000 Analysis by Media Sector
MediaSector
CableTV β=0.76 NetworkTVβ=0.85 ROandlinioe ββ==00..8830 Newspaper β=0.83 40000 MediaSector
CableTV α=1.07 NetworkTVα=0.98 ROandlinioe αα==00..9927 Newspaper α=0.89 104 103 broadcast TV and cable TV). Figure 2(a) shows how novel lead newsmakers appears in the news sequence, for each media sector collected by the NCI. The interpretation of these results is, the steeper the line, the more novelty the media sector has (according to the de nition of novelty used in this paper). Surprisingly, newspapers is the media
sector with the larger ratio of lead newsmakers per story, followed by online portals, radio, network TV and cable TV. Figure 2(b) shows an orthogonal insight for this result, which shows the rank distribution of lead newsmakers for each sector. As Heaps' law and Zipfs' law are inverse correlated, the interpretation of these results are, the steeper the line, the more a media sector concentrates the coverage in a few lead makers. Cable TV repeats lead newsmakers more often than other sectors, and (proportionally) uses fewer leads newsmaker than the other media sectors. Newspapers, on the other hand, proportionally use the top ranked lead news makers less often, and have a larger number of histories with di erent lead makers.
We can speculate that the higher frequency of novel lead newsmakers in newspaper media is due to that this media needs competitiveness in relation to other media (digital and electronic), which are characterized by dissemination of news in real time. As the newspaper is a diary media, it always needs to have something di erent to present than what was published on the previous day on TV, radio and, Internet. Despite being late in relation to events in one day (it generally publishes stories from the eve), the newspaper still continues to be a source for other rival media because it intends always having something new in their pages. On the other hand, TVs have a rotating audience, and focus on a narrow range of topics. Thus, the presented histories focus in a few lead newsmakers. Radios an online media are somehow in between these two extremes.
To gain some insight in the online versus o ine scenario, we break down the analysis in online versus ofine media, as shown in Figure 3. The interpretation of the graphs is the same as of 2. Figure 3(a) shows that online sector introduce more lead makers in their stories, and Figure 3(b) indicate that the same lead maker appears less often in online media. As can be seen from the graphs, online media have stronger novelty signatures. Therefore, online media have a bias towards introducing more di erent lead newsmakers, and a lower tendency to echo the same leading maker in future stories.
A possible reason for this is that online outlets have a high propensity to show new stories due to the di erence in media consumption from the target audience. In general, the audience for online news sources is of younger people (as discussed in the previous section). These users have a less tendency to in-depth stories, foComparison of online versus non−online media Analysis by Broad Story Topic 102 f)( R 101 0 20000 10000 cusing in the headlines. They also are more connected, and access the news more often, thus the necessity of novelty in the news stories.
We did a similar analysis, but conditioning on the ve most frequent broad story topics. The broad story topic variable identi es which of the broad topic categories is addressed by a story. NCI has 32 broad story categories, but most of them have low frequencies. These low frequencies di cult an analysis, due to a lack of data. Figure 5 shows these results. The interpretation of the graphs is the same as of 2, except for the fact that instead of media sectors, we have topics in these graphs. Figure 4(a) shows how novel lead newsmakers appear in the news sequence, for each of the ve most frequent topics collected by the NCI. In these gures, one traditional attribute of news value, Negativity (any reference that is negative), emerges as through Crime broad story topic. Crime is the topic with the largest rate of novel lead makers, followed by economy/economics, US foreign a airs, government agencies/legislatures and campaigns/elections/politics with the lowest rate. Figure 4(b) indicates that the most frequent lead makers appear proportionally less often in the news than the most frequent lead makers in campaigns/elections/politics. Furthermore, crime is the sector with the largest proportion of lead makers to appear in fewer histories.
A similar analysis was performed by start time of the program, as shown in Figure 6. The interpretation of the graphs is the same as of 2, except for the fact that instead of media sectors, we have the program start time in these graphs. Figure 5(a) shows that, in general, morning programs introduce more often new lead makers, while night programs have few novelty lead makers. On the other hand, Figure 5(b) shows that night program cites more often the more noticed lead makers than morning programs. We believe this also is related to the target audience, which in the evening/night has a higher prevalence of elderly people, which is more interested in-depth coverage. 5
In this paper, we examine the dynamic of novelties in
the NCI dataset. We used the lead newsmaker as the
main variable to de ne the concept of novelty in our
framework. We veri ed a very good t of these data
to the two novelty signatures discussed in [
We obtained interesting and insightful insights when conditioning the analysis do media sector and broad story topic. Regarding media sector, we veried that newspapers is the sector with largest novelty, in terms of the introduction of new lead newsmakers. Furthermore, online media have a largest novelty, when compared to non-line media. In terms of story topic, crime is the sector with more novelty, also in terms of lead newsmakers. 6000 )(4000 ND (a) Heap's Law 101 102R
We believe these patterns somehow tend to follow the interest of the public in order to get her attention. Thus, there is the necessity to provide news on topics to better reach a target audience, tailoring the audience. An interesting future work is to analyze whether these patterns would be similar in the next years, because the online young audience became a generation more mature. Would the behavior be the same, and the sectors have to adapt to the news consumption patterns of this generation or they will change their tastes, showing a similar behavior or their previous generation as accessing the in-depth stories?
This research has two obvious limitations. First, our adopted de nition of novelty does not capture all aspects of novelty, as new information can be published about lead Newsmakers which already appeared in the sequence. However, we believe this de nition do capture some aspects of novelty, and were able to provide some interesting insights on the topic. Furthermore, the data set has a bias towards the U.S.A. media coverage. An interesting further research direction is to broaden this research to di erent sources.