@SNOW/WWW, 2016, by Devipsita Bhattacharya and Sudha Ram
News consumption and distribution has undergone an unprecedented change with the rise of the Internet. News in this day and age has become synonymous with e-articles published on news websites. Use of web and electronic presentation technologies has enabled news providers to create content rich webpages to deliver news in a detailed and an engaging manner. News articles now contain a variety content such as text, images, podcasts, videos and real-time user comments. For anyone with an Internet connection, news is now an on-demand commodity.
With features of social recommendation, content sharing and micro-blogging; social media websites (e.g. Twitter) also play a critical role in electronic distribution of news articles. Users after reading online news articles, submit their recommendations on Twitter in the form of tweets, which are then viewed by other users leading to news article webpage visits. User activities such as these have indirectly enabled the news providers to make their audience aware of the content published daily on their websites. News article sharing has also helped news providers to reach out to a much wider audience in a very cost effective manner than was possible hitherto. Moreover, in response to the popularity of social media, news agencies have created official user accounts on various social media websites and use these accounts to regularly to post about selected news articles published on their respective websites.
Our previous work has methodologies for analyzing news article propagation on social media, e.g., Twitter. We have extracted and analyzed several Twitter based propagation networks to examine characteristics of user participation in news article propagation. Using these networks, we have formulated a framework for performance measurement of news article propagation [1; 2; 4]. Our framework includes measures that can be grouped into two major categories i.e. Structural and temporal properties of propagation. Structural properties measure unique characteristics of the propagation network such as length of longest cascade chain, average cascade length, user contribution, effective influence of news provider, article conversion ratio and multiplier effect. Temporal properties includes measures related to lifespan of articles, retweet response rate, and rate of spread of tweets. We have also extracted implicit networks of user-user relationships based on commonalties of article preference and tweeting activity and examined how users connect over time based on their article sharing activity . Our work has enabled us to examine news propagation in a unique and multi-faceted way by harnessing the power of network science.
Our current study focuses on similarities and differences in news propagation patterns on social media based on the primary channel of a news provider. The Internet apart from enabling e-articles has also transformed the news landscape with traditional news providers (printed, televised etc.) competing for news webpage article views. That is, news providers that previously competed with other providers primarily based on the channel of news distribution, now find themselves competing with a whole new set of participants. For instance, before the Internet, newspapers such New York Times and Washington Post were competing for subscribers and advertising revenues from their printed newspapers. Similarly, network news companies such as CNBC and CNN were competing with each other for audience engagement during prime time news hour. However, with each of these news providers now having news websites, the competition is no longer limited to their rivals in their primary distribution channel. News providers now also contend to attract advertisers and readers for their article webpages. In our study we compared the patterns of news article propagation on Twitter based on the primary channel of news distribution. News providers on the Internet can be grouped into different categories based on their primary channel of distribution.
Generally, Porter’s Five Forces model is used for strategic analysis of organizations in an industry. However, in our study, we develop a network based methodology for analyzing the competition among these news channel categories on social media. We collected a dataset of 24 million article tweets from Twitter for 32 news providers over a 3-week period, i.e., September 1- September 22, 2015.
Using this dataset, we extracted news propagation patterns for each news provider and analyzed the similarities and differences between their networks. Our Twitter based propagation network is a user – user network defined for a single news provider. Each node in the network represents a Twitter user participating in article sharing activity of a given news provider. Each edge (directed from source to target) represents the aggregate retweeting relationship established between two users. It is a network of aggregated propagation activity (i.e. across multiple articles) observed over a period of time.
We compared the networks using a number of structural properties of their propagation networks including the “density” of each network, proportion of disconnected users, average length of user cascade chains, number of retweeting relationships per user, the ability of users to form communities, and tweeting frequency of user(s).
We determined that when compared to networks of other channels, “online only” news providers have the smallest (number of nodes and edges) but the most dense networks. Interestingly, even with high density, their networks were found to have a higher concentration of disconnected nodes. This is expected since “online only” news providers have emerged only recently when compared to other news providers in our sample. For other news channels, we had mixed inferences when examining structural properties of their propagation networks. But, we were able to establish a statistically significant difference between the news channels based on their structural properties.
Our analysis of the news channels using a network based methodology makes several contributions.
- It allows news providers to benchmark their social media based propagation performance against other competitors in the same or in a different primary distribution channel. This is particularly important since on social media, even traditional suppliers of news (e.g. News agencies such as Reuters, Associated Press) are considered direct competitors for any news provider hosting an online news website.
- We identified features unique to our Twitter-based aggregate user-user networks. Important among these, is the presence of multiple disconnected communities of nodes. On an average, we found that a news provider’s propagation network had at least 4,000 disconnected communities containing two or more nodes. This highlights the importance of news article tweeting activities independent of those originating from news provider Twitter accounts.
- Network analysis adds a new dimension to competitive analysis which generally considers participation volume (number of users) to measure engagement. For instance, we observed that “news agency” (Reuters, Associated Press etc.) propagation networks had lower average counts of nodes and edges when compared to those of “television” (ABC News, CNN etc.) news networks (by a margin of 100,000). By considering these differences in values, television news providers emerge as “winners” in audience participation on social media when compared to “news agency” networks. However, we also ascertained that television and news agency networks had approximately equal values of network diameter (19.5 and 19 respectively). While on an average television based news agencies networks show higher tweeting and retweeting activity from their Twitter users, their audience’s ability to connect amongst each other to form the longest cascade chain over time is the same as that of “news agency” networks having lower average Twitter user participation count.
Our research points reveals that analysis of competition among news providers on social media needs a comprehensive consideration of various facets associated with user participation. It also shows that network science can provide important insights into the changing landscape of news on social media.
 Bhattacharya, D. and Ram, S., 2012. News article propagation on Twitter based on network measures – An exploratory analysis. In Proceedings of the 22nd Workshop on Information Technology and Systems.
 Bhattacharya, D. and Ram, S., 2012. Sharing News Articles Using 140 Characters: A Diffusion Analysis on Twitter. In International Conference on Advances in Social Networks Analysis and Mining (ASONAM), 966-971.
 Bhattacharya, D. and Ram, S., 2013. Community Analysis of News Article Sharing on Twitter. In Proceedings of the 23rd Workshop on Information Technology and Systems.
 Bhattacharya, D. and Ram, S., 2015. RT @News: An Analysis of News Agency Ego Networks in a Microblogging Environment. ACM Trans. Manage. Inf. Syst. 6(3):1-25.