Dear colleague/fellow participant :-)
Thank you for taking the time to check out our work at ASNAC 2016.
Please feel free to contact us c/o m.cheong AT monash.edu
This is a list of references from our poster slides as presented at ASNAC 2016, as well as from our extended abstract below. More of our work can be found in the Our papers section at the end.
Aleti Watne, T., M. Cheong and W. Turner (2014), "#Brand engagement or @Personal engagement? How Australian 'Mass Brewers' and 'Craft Brewers' Communicate with Consumers through Twitter". Proceedings of the Academy of Marketing (AM) conference, Bournemouth, UK, July 7-10.
Borra, E. and B. Rieder (2014). "Programmed method: developing a toolset for capturing and analyzing tweets," Aslib Journal of Information Management, 66(3), 262-278.
Byrne, J. and M. Cheong (2016). "Social media (meta)data analysis: perspectives from social anthropology". (Work in progress).
Cheong, M. (2013). Inferring social behavior and interaction on twitter by combining metadata about users & messages (Doctoral thesis, Monash University, Australia). Retrieved from http://arrow.monash.edu.au/vital/access/manager/Repository/monash:120048.
Ghazarian, Z. and M. Cheong (2016). "Social media and political interaction: The case of the 2016 Australian national election". (Work in progress).
May, J. (2011). "Burning issues". The Age. August 13: 15-17.
Meikle, J. and Jones, S. (2011). "UK riots: More than 1,000 arrests strain legal system to the limit". The Guardian. August 10.
Robinson, D. (2016). "Text analysis of Trump's tweets confirms he writes only the (angrier) Android half". Variance Explained. Retrieved from http://varianceexplained.org/r/trump-tweets/.
Rogers, S., Sedghi, A. and Evans, L. (2011). "UK riots: every verified incident - interactive map". Retrieved from http://www.guardian.co.uk/news/datablog/interactive/2011/aug/09/uk-riots-incident-map.
Here is a copy of our Power of Metadata poster slides.
Abstract:
The Power of Metadata: The role of statistical pattern recognition and inference algorithms in Twitter studies.
The literature on Twitter in academia has mushroomed over the past 7 years, with many research areas opening up periodically. Hot topics nowadays focus on extrinsic properties of Twitter such as issue mapping, social network analysis, text mining, and 'big data' visualisation and analysis - powerful standalone techniques in their own right. However, one area which has remained somewhat a niche is the usage of raw metadata for inference generation and as the basis for various statistical pattern recognition techniques.
Analyses of metadata from users and messages on Twitter can be more beneficial (and can strengthen) 'surface level' analyses of tweets and the user network. This is evident in a simple yet pertinent example: how a single metadata attribute, the "Twitter client" can be used to distinguish between US Presidential Candidate Donald Trump's personal tweets and his campaign staffers' (Robinson, 2016).
In this presentation, we will draw from Twitter metadata-based research in Cheong (2013)'s PhD thesis as well as related work (conducted from 2009-2013) to provide examples of the relatively cost-effective, simple, yet powerful techniques that can be employed on Twitter metadata to generate meaningful inferences. These were based on large-scale empirical studies, experimental metadata-collation techniques, bespoke Twitter API harvesting scripts, and eclectic studies of the Twitter ecosystem; which pre-dates most large-scale data collection and analysis platforms such as TCAT (Borra & Rieder, 2014).
We will also cover how these niche computer-science based techniques are still relevant to more contemporary studies and popular analysis techniques. We believe that they can augment findings from current analysis methods found in popular studies, especially in line with the recent surge in popularity of 'big data' and increased awareness of machine learning techniques. Examples include proofs-of-concept of how statistical correlation and clustering on Twitter metadata - staples of data mining techniques - can be used to reveal latent patterns in day-to-day happenings as reported by Twitter users. As an analogy, this is not dissimilar to how hidden neural network patterns are discovered by Google DeepDream.
Also, in the interest of multidisciplinary research, this presentation will review existing (and propose new) applications of our metadata-based techniques in several other disciplines: including but not limited to marketing (Aleti Watne et al, 2014), political science (Ghazarian & Cheong, 2016) and social anthropology (Byrne & Cheong, 2016).
Ultimately we believe than by combining the best of multiple disciplines, our niche techniques will provide valuable understanding of human phenomena - resulting from cross-disciplinary research that is greater than the sum of its constituent parts.
Here is a list of our papers resulting from research since 2009...