Loading…

Sign up or log in to bookmark your favorites and sync them to your phone or calendar.

Session 1D [clear filter]
Tuesday, July 12
 

10:30

Session 1D: Analytics & Data Mining
Location: PSH (Professor Stuart Hall Building) - LG01, 
Goldsmiths, University of London, Building 2
Campus Map 


Moderators
avatar for Grant Blank

Grant Blank

University of Oxford

Tuesday July 12, 2016 10:30 - 12:00
PSH (Professor Stuart Hall Building) - LG01 Goldsmiths University, Building 2

10:31

Accuracy of User-Contributed Image Tagging in Flickr: A Natural Disaster Case Study
Location: PSH (Professor Stuart Hall Building) - LG01, 
Goldsmiths, University of London, Building 2
Campus Map 

Contributors:
  • George Panteras, George Mason University, United States
  • Xu Lu, Arie Croitoru, George Mason University, United States
  • Andrew Crooks, George Mason University, United States
  • Anthony Stefanidis, George Mason University, United States

Social media platforms have become extremely popular during the past few years, presenting an alternate, and often preferred, avenue for information dissemination within massive global communities. Such user-generated multimedia content is emerging as a critical source of information for a variety of applications, and particularly during times of crisis. In order to fully explore this potential, there is a need to better assess, and improve when possible, the accuracy of such information. This paper addresses this issue by focusing in particular on image tagging (i.e. user-assigned annotation) in Flickr. We use as case study a natural disaster event (wildfire), and assess the accuracy of user-generated tags. Furthermore, we compare these data to the results of a content-based annotation approach in order to assess the potential performance of an alternative, user-independent, automated approach to annotate such imagery. Our results show that Flickr user annotations can be considered quite reliable (at the level of ~50%), and that using a spatially distributed training dataset for our content-based image retrieval (CBIR) annotation process improves the performance of the content-based image labeling (to the level of ~75%). 

Tuesday July 12, 2016 10:31 - 12:00
PSH (Professor Stuart Hall Building) - LG01 Goldsmiths University, Building 2

10:31

Embedded Metadata and the Circulation of Images: Tracking, Storing and Stripping
Location: PSH (Professor Stuart Hall Building) - LG01, 
Goldsmiths, University of London, Building 2
Campus Map 

Contributor: Nathalie Casemajor, Université du Québec en Outaouais, Canada

Background:

Metadata is a set of descriptive, technical and administrative information that plays a key role in image storage, processing and circulation. Recent literature in media studies highlights the roles metadata plays in domains such as digital economies and informational infrastructures. Many of these studies focus on the music sector: in particular, they tackle the automated or manual classification of songs, as well as the algorithmic systems of recommendation (Beer, 2013; Morris, 2012, 2015). In the field of visual culture, various publications emphasize the role that metadata plays in the classification of images by amateur and professional photographers (Van Dijck, 2010; Boullier and Crépel, 2013). But few broach the distinction between embedded metadata and platform-specific metadata. The former refers to data directly stored within the image file, which allows the data to travel with the picture on its journey across platforms, whereas the latter refers to data separately stored on proprietary web servers, including keywords, geotags and other folksonomies, which are lost when the picture is copied from one platform to another.

Objective:

This paper focuses on image metadata (in particular, photographic images) to illustrate how web platforms handle images, and how these technical choices are tied to different economic models of content and audience retention. Its aim is to challenge the assumption that the more an image circulates and is appropriated on social media, the more metadata it subsequently accumulates. This paper suggests instead that there is a critical distinction between the way embedded and platform-specific metadata accumulate.

Methods:

This study is based on a set of experiments conducted on a small corpus of photographs. In collaboration with a Canadian visual artist, five images of artworks were marked with embedded metadata and steganographic information before being posted on three different platforms (Wordpress, Facebook and Instagram). Six months later, all the copies in circulation were collected through Google Image reverse search and TinEye, and their metadata were extracted via an application named Exiftool. A quantitative and qualitative analysis was conducted on the metadata to compare 1) how the transit through each platform affected the embedded information and 2) what kind of platform-specific metadata was attached to these images. A complementary analysis was conducted on Flickr and Twitter with a random set of images.

Results:

The preliminary results suggest that contrary to platform-specific metadata that stably accumulate on web servers, embedded metadata is shaped by a complex dynamic of accumulation and degradation. On the one hand, social media platforms tend to strip embedded metadata out of their users’ images (this is especially the case with platforms designed for non-professional image-sharing practices, such as Facebook), while on the other hand, social media platforms encourage users to recreate this data in a proprietary format tied to the platform. Therefore, the more an image circulates beyond the thresholds of proprietary platforms, the more its metadata becomes degraded. The images that cross boundaries between platforms and travel through various social media datascapes are the most portable (see Sterne, 2006), but the quality of their metadata is poorer. In terms of audience retention, metadata stripping increases content captivity, as it makes it more difficult for users to move their archives from one platform to another, knowing that metadata (re)creation is a time- consuming operation.

Future Work:

This paper argues that paying thorough attention to the specificities of image metadata lays the groundwork for an understanding of the broader ecology of social media. Further work on larger datasets could foster insights regarding the power and economic dynamics of data streams within and across social media (Manovich, 2012; Hochman, 2014), all the while reflecting on the politics of web platforms (Gillespie, 2010; Helmond, 2015).

References:

Boullier D. and M. Crépel (2013). Biographie d’une photo numérique et pouvoir des tags : classer/circuler. Revue d’Anthropologie des Connaissances, 7(4), 785-813.

Gillespie, T. (2010). The politics of ‘platforms’. New Media & Society, 12(3), 347-364.

Helmond, A. (2015). The Platformization of the Web: Making Web Data Platform Ready. Social Media + Society 1(2).

Hochman, N. (2014). The social media image. Big Data & Society, 1(2). Available at: http://bds.sagepub.com/content/1/2/2053951714546645 (accessed 12 January 2016). 

Manovich, L. (2012). Data stream, database, timeline: the forms of social media. Software Studies Initiative. Available at: http://lab.softwarestudies.com/2012/10/data-stream-database-timeline-new.html (accessed 12 January 2016)

Morris, J. W. (2012). Making music behave: Metadata and the digital music commodity. New Media & Society, 14(5), 850-866.

Morris, J. W. (2015). Selling Digital Music, Formatting Culture. Berkeley: University of California Press.

Sterne J. (2006.) The mp3 as cultural artifact. New Media & Society, 8(5), 825–842.

Van Dijck, J. (2010). Flickr and the Culture of Connectivity: Sharing Views, Experiences, Memories, Memory Studies, 4(4), 401-415. 


Tuesday July 12, 2016 10:31 - 12:00
PSH (Professor Stuart Hall Building) - LG01 Goldsmiths University, Building 2

10:31

Everyday Socio-Political Talk in Twitter Timelines: A Longitudinal Approach to Social Media Analytics
Location: PSH (Professor Stuart Hall Building) - LG01, 
Goldsmiths, University of London, Building 2
Campus Map 

Contributors:
  • Phillip Brooker, University of Bath, United Kingdom
  • John Vines, Newcastle University, United Kingdom
  • Julie Barnett, University of Bath, United Kingdom
  • Tom Feltwell, Northumbria University, United Kingdom
  • Shaun Lawson, Northumbria University, United Kingdom

Background

Increasingly, social media spaces are understood by researchers to be a valuable site of everyday politically-relevant discussions. However, qualitative usages of social media data are typically undertaken with existing tools and methods developed for more ‘traditional’ tools and methods (i.e. thematic analysis and content analysis and so on). This, we argue, misses an opportunity to develop new methods which may be more tightly attuned to the idiosyncrasies of such data. Accordingly this paper aims to provide a means of drawing on and working with such idiosyncrasies, demonstrating the value of doing with reference to an empirical case.

Objective 

Building on previous work looking at how everyday discussions around social welfare issues arise on Twitter around the broadcast of a TV show (Benefits Street) (Brooker et al, 2015), we seek to explore the possibilities arising from capturing an atypical slice of Twitter data (i.e. whole timelines) and treating those data with an atypical analytic approach (i.e. investigating timeline narratives longitudinally). Hence, this paper will dually comment on both the empirical case at hand, and the methods requirements worked out through the course of undertaking this work.

Methods

We captured timeline data of 2581 Twitter users tweeting using the ‘official’ #BenefitsStreet hashtag during the broadcasts of both series of the programme (January 2014 and May 2015), amounting to 6,260,444 tweets in total. We undertake an exploratory analysis of an arbitrarily selected subset of user timelines within the master dataset, concentrating on the pervasion of welfare discussion throughout these users’ timelines between the two series’ of Benefits Street, as well as drawing out other themes and topics that motivate these tweeters to tweet.

Results

The study elaborates on how socio-political talk on Twitter fits in with tweeters’ everyday talk around a range of different interests and topics. This study also demonstrates the potential for longitudinal analysis of timeline narratives as an innovative qualitative approach to social media data, which can tap into the depth of meaning that such data may hold for those who produce it.

Future Work

The present study stands as a first exploratory step in the analysis of the full dataset of 6,260,444 tweets, also providing more generalizable methodological grounding on which to base the idea of longitudinal research with social media timeline data. However, the complexity of applying this approach to data of this kind also requires further thought and discussion around the development of scalable computational tools for assisting qualitative researchers in the handling of such large-scale data. The present paper points the way towards solutions for both of these problems.

References

Brooker, P., Vines, J., Sutton, S., Barnett, J., Feltwell, T. and Lawson, S. (2015). Debating poverty porn on Twitter: Social media as a place for everyday socio-political talk. CHI ’15 Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems, 3177-3186. http://dx.doi.org/10.1145/2702123.2702291 

Tuesday July 12, 2016 10:31 - 12:00
PSH (Professor Stuart Hall Building) - LG01 Goldsmiths University, Building 2

10:31

Temporalities of Personal Analytics: emerging patterns of engagement with temporal data about the self
Location: PSH (Professor Stuart Hall Building) - LG01, 
Goldsmiths, University of London, Building 2
Campus Map 

Contributors:
  • Martin Hand, Queen's University, Canada
  • Michelle Gorea, Queen's University, Canada

Background

the proliferation of ‘self-tracking’ devices has become a recent focus of research into ‘everyday’ or ‘personal’ analytics, including historical antecedents (Crawford et al. 2015), implications for citizenship, health and biopolitics (Lupton 2014), self-tracking markets (Pantzar and Ruckenstein 2015), surveillance (Whitson 2013), and the broader ‘quantified self’ movement (Nafus and Sherman 2014). There has been relatively little qualitative analysis of the contexts in which such devices are ordinarily used, how the data is interpreted, used, and shared by individuals, and how such data relates to broader practices of temporal scheduling and coordination in daily life. This paper makes a significant contribution to knowledge, showing how such devices are becoming integrated with established technologies of marking and making time (clocks, calendars), are being used to explicitly manage time, and are ambiently shaping ‘lived time’ in diverse ways (Wajcman 2015). 

Objective

The paper aims to provide detailed empirical data on how individuals do and do not adopt self-tracking devices and negotiate their own data in terms of the temporality of personal analytics. Drawing upon in-depth interviews, we show the different ways in which these devices presume, produce, mediate, manage and shape temporal practices. 

Methods

The empirical data was gathered over several months as part of a larger SSHRC funded program concerned with the contours of ‘iTime’. The data used here is in-depth interviews (N=30) selected by quota sample to reflect the overall demographic of the university. There are two dimensions to this group being explored. First, multiple device ownership within this demographic is very high, but we know very little about their understanding and management of temporality through digital mediation, and how this relates to the specific expectations of university life, friendships and maintaining a connected presence’ across multiplying social media platforms. Second, approximately half of the sample (N=15) was selected for their ownership and use of wearable fitness applications (i.e. ‘fitbit’). This is a focused effort to understand emerging practices of self-tracking in relation to the production of temporal data about the self. We have rich data on the connections between these practices and the broader expectations within this group. Participants reflected upon their own devices and social media data during interviews. 

Results

Preliminary analysis of our data reveals continuities between existing temporal practices, but also significant novel trajectories encouraging users to (a) rethink and reshape their conception and organization of time (b) share their data across social media platforms to regulate personal time, (c) meet new expectations about temporal management being produced through the tracked data.

Future Work

These findings will enable important insights into the normative temporal expectations of self-tracking devices, and how these are understood and negotiated both through social media and a range of integrative practices. How these devices become elements of people’s media ecologies or ‘manifolds’ is crucial to understanding their relative significance (Couldry 2012). These initial findings will also be revisited alongside interview data (N=100) from other populations being gathered during 2016, concerning differences in socio-economic resources, age, and urban proximity.

References

Couldry, N (2012). Media, Society, World. Cambridge: Polity.

Crawford, K., Lingel, J., and Karppi, T. (2015). Our metrics, ourselves: A hundred years of self-tracking from the weight scale to the wearable device. European Journal of Cultural Studies, 18 (4-5), 479-496.

Lupton, D. (2014). Quantified Sex: A Critical Analysis of Sexual and Reproductive Self- Tracking Apps. Culture, Health and Sexuality, 17(4), 440–53.

Nafus, D., Sherman, J. (2014). This one does not go up to 11: the Quantified Self movement as an alternative big data practice. International Journal of Communication, 8 (11), 1784-1794.

Pantzar, M., Ruckenstein, M. (2015). The heart of everyday analytics: emotional, material and practical extensions in self-tracking market. Consumption Markets & Culture, 18(1), 92–109. 

Ruckenstein, M. (2014). Visualized and Interacted Life: personal analytics and engagements with data doubles. Societies, 4, 68-84.

Wajcman, J. (2015) Pressed for Time. Chicago: Chicago University Press.

Whitson, J. (2013). Gaming the Quantified Self. Surveillance and Society, 11 (1/2), 163-167. 

 

Tuesday July 12, 2016 10:31 - 12:00
PSH (Professor Stuart Hall Building) - LG01 Goldsmiths University, Building 2