STUDYING LIVENESS @OerolFestival2017

Team Members

Esther Hammelburg, Lauren Drakopulos, Andrea Benedetti and Gabriele Colombo

Introduction

This project focussed on the way in which images were used on Instagram to connect to the Oerol festival 2017 in the sense of ‘presencing’ (cf Meese et al., 2015). The case study is part of a PhD project that focuses on liveness in the mediatised experience of cultural events/festivals. Liveness is understood as being connected through media to events that matter to us as they unfold (cf. Vianello, 1985; Couldry, 2004; Auslander, 2012). Oerol is an ideal case to study how liveness is constituted during a cultural event.

Oerol is an annual festival for theatre, dance, street theatre, art and music that takes place during ten days in June on the island of Terschelling. Since performances are created or adapted specially for the unique natural locations where they are carried out (beaches, woods, dikes, barns, streets, etc) this case provides an interesting surrounding to explore the relation between presence and liveness. The festival attracts about 50.000 visitors each year.

Data Sets

The data set for this project contains 5784 unique images about Oerol 2017 collected from public Instagram profiles using DMI Visual Tagnet Explorer (DMI-VTE).

Data collection

Before collecting data, the scope of images around Oerol 2017 - both in time and in volume - and the most used hashtags were explored via Coosto and by downloading sets through Popsters. Coosto is especially handy to get an idea of the time frame of posting and often used hashtags. Using these hashtags data sets were downloaded with Popsters just to get a grip on how big the sets would be. The amount of iterations in collecting was based on this exploration and proved to work to get the whole data set.

Data collection using DMI Visual Tagnet Explorer (DMI-VTE):

1. Three location searches were done to cover the island of Terschelling during the festival, dates June 9 – June 18 2017:

   1. 53.347114, 5.175368, 5000 meter radius

   2. 53.375048, 5.263879, 5000 meter radius

   3. 53.401886, 5.350061, 5000 meter radius

2. Six tag searches were done using the most used hashtags for Oerol 2017:

   1. #oerol for 500 iterations.

   2. #festivaleiland for 70 iterations.

   3. #oerol2017 for 80 iterations

   4. #oerol17 for 40 iterations

   5. #oerolfestival for 20 iterations

   6. #oerolterschelling for 30 iterations.

Data selection and limitations of the data set

In Excel the 9 spreadsheets from the DMI-VTE searches were combined, and subsequently all posts before January 1st 2017 and all duplicates were removed. The result is a data set that contains 5784 unique images about Oerol posted on Instagram between January 1st 2017 and June 28th 2017: 4511 posts made during the festival, 833 before and 440 after the festival.

The aim was to collect all images that were posted about Oerol on public Instagram profiles in 2017 until the date of collection. However, images to which the user manually added the Instagram location ‘Oerol, Terschelling’ and that were not gps-located or posted with an Oerol-connected hashtag are not in the data set. When interviewing Oerol visitors during the festival many mentioned that they added the Instagram location and didn’t use hashtags. Using the tool 4K Stogram the ‘missing’ images were explored and estimated to be approximately 1000 images. At a first glance these images look similar to the collected ones, but is good to note the possibility of significant differences and therefore the limitation of the existing data set.

Research Questions

How do users visually place themselves in the here and now of the event-sphere of Oerol 2017?

1. What types of images - content, point-of-view, composition - do users share about Oerol?

2. How do visitors use, repurpose and shape the platform vernacular (Gibbs et al., 2015) of Instagram in sharing their Oerol experience?
   
   

By exploring the use of Instagram we also get an idea of which methods and tools are most useful to capture liveness in the mediated event-sphere of Oerol.

Methodology

For exploration of the dataset we chose to organise the images along three axes:

1. Time: when are posts made; how do images show the temporality of experiencing the event?

2. Content: what is shown in the images; what content is created in the interplay between user and platform vernacular?

3. Place: where are posts placed; how is presencing done geographically?

This results in three different outcomes that can be used for further visual content analysis: an image time plot, an image network and an image map.

Image time plot

An image time plot can help study how users (re)purpose the platform vernacular of Instagram for the expression of temporality during Oerol 2017. Using the ImagePlot visualization software images can be organized in several ways. We decided to plot the images in stacks using days on the horizontal axis and hours on the vertical axis. Questions that underlie this choice are:

• When during the day do people post?

• What do people post in which moments of the festival?

• Or more specific: are there certain arrival, being there and farewell images?

Including the day before and the day after the festival, 12 days were selected to visualize: June 8 until June 19 2017. This gave us a data set containing 4779 images.

Using the whole data set would not lead to a readable visualization, therefore we used a random sample of 10% of the images of each selected day:

June 8	115 total > 12 in sample
June 9	337 total > 34 in sample
June 10	488 total > 49 in sample
June 11	476 total > 48 in sample
June 12	418 total > 42 in sample
June 13	402 total > 40 in sample
June 14	436 total > 44 in sample
June 15	478 total > 48 in sample
June 16	472 total > 47 in sample
June 17	500 total > 50 in sample
June 18	504 total > 50 in sample
June 19	153 total > 15 in sample

The sample was made in Excel following these steps:

1. Create a worksheet per day and cut and paste data from that day.

2. Add column to every worksheet and fill with random number using ‘=RAND()’

3. Arrange data per worksheet by random number.

4. Select first [amount for sample] items and paste these into a new worksheet

For a full-sized visualization see attachments

Since our questions focus on the user in the sense of the visitor of Oerol, we excluded posts from a couple of professional accounts that were prominent in the data set. Photo’s posted by the official Instagram account of the festival - oerolterschelling - were excluded from the sample from the start. After the sample was made 50 items of the total sample of 479 were selected based on username to examine more closely by opening them in Google Sheets and adding a column with the image. Of these items 15 were manually replaced by random post from approximately same date and time. Users excluded from sample:

Oerolterschelling, kaatjemup, blikopfestivals, teteateteportret, theatercollectiefgarp, yearoftherebel, wintertuin, thehillbillymoonshiners, stayokayterschelling, sir.duke_, opiumopoerol.

Image network

An image network helps to explore what content is created around Oerol in the interplay between user and platform vernacular. We created a visual network based on content labeling by the Google Vision API:

1. Run the data set of 5784 images through Google Vision API using Memespector.

2. Copy processed file and work from copy (to keep the original clean)

3. Create new spreadsheet for Gephi: end file needs a row per image per label and 3 columns: column 1 = Source (image file name); column 2 = Target (label); column 3= Weight (confidence in label).

• Paste in columns image_id and file_ext > combine them into new column and name that column Source > delete the pasted columns

• Paste in column gv_labels > separate text to columns by comma and ( to create separate columns per label and weight > find ) and replace with to delete all ) > copy paste to create three columns and a row per image per label.

4. Import data as node table in Gephi.

• Run ForceAtlas2 algorithm for network spatialization.

• Apply LinLog Mode and increase gravity to make clusters more visible.

• Explore labels and remove those that are not meaningful: vision care, vertebrate, product.

5. Export as .svg.

6. In the svg code change string of text into image tag.

7. Open in Adobe Illustrator.

Image map

Explore the geospatial patterns of images through an image map made in Carto. Mapping the images allows for analysis of when and where particular kinds of content are shared.

1. Create new data file for mapping with the full data set of 5784 images.

2. In Excel arrange data by date.

3. Create two additional columns in which posts are coded in numbers and in text:
   1 Oerol
   2 two weeks from Oerol (before or after)
   3 further from Oerol (in time)

4. Import coded data spreadsheet into Carto and georeference using lat/long coordinates.

5. Style map symbology to color code data point symbols according to time of photo post using the coding schema.

6. Include popup with relevant metadata: image url to display image when clicked, user hashtags and captions, number of likes.

Findings

Time

Plotting images along axes of days and hours enables us to have an overview of how users (re)purpose the platform vernacular of Instagram for the expression of temporality during Oerol 2017. Reading the stacks gives us some insight in what kind of images are posted in a and during the day. It can give us an idea of the most prominent features of a day as it is experienced and shared on Instagram.

What immediately shows is that the visual language of the experience of Oerol is very much connected to (the landscape of) the island of Terschelling. The second element that is prominent in these images is that presencing - as positioning oneself in the context of the festival - is often done with a selfie or a first person perspective including items, people or scenery that are meaningfully connected to that moment in the festival. These meaningful elements will be further explored under content.

When comparing stacks, exploring the variations between days, we can notice ‘on my way there’ and ‘farewell’ practices more prominent in the first and last days of the festival and of the weekends, and ‘I’m here’ practices more visible in the middle. Typical images used to portray the travel to the festival are selfies on the boat, pictures from a car window on the Afsluitdijk and selfies in the car. Farewell is often said by a picture from the boat, preferably with feet or waving hand visible, of the island of Terschelling. Oerol visitors use the platform vernacular of Instagram to position themselves either on the island of Terschelling and in the festival or at a meaningful close distance to it: on their way there or saying goodbye to Oerol. Both positions - in and close to Oerol - are often taken by the prominent selfie or first person perspective as a presencing practice. The images do differ however in sentiment and content, where the ‘on my way there’ and ‘farewell’ practices tend to take on the visual language from road movies. This gives us an angle to further explore.

.

Choices and limitations

Since the time included in the metadata is the time that an image is posted we have to account for the fact that it is not necessarily the time that the picture is taken. The insight we get is more about the rhythm of posting, about the platform vernacular, than about the temporality of photographing moments in the experience of Oerol.

A second point to consider is in which time frames to sample. For this image time plot we chose to randomly sample 10% images per day and therefore this shows us also which hours are busiest. However it does give us quite a lot of overlap in images and quite some blanks in the days. It might be interesting to explore what happens if you make sure that the images in the sample spread more along the day: fi x images per hour. The rhythm of posting is then less visible, but we do get more of a view of what kinds of images are posted when.

Content

Using Google Vision API labels for creating a network and then adding the actual images provides us with an overview of what content is created in the interplay between user and platform vernacular. As we have seen in the exploration of the Image time plot people seem to include specific items, people or scenery that are meaningfully connected to Oerol. In a first exploration of the network a couple of meaningful clusters have been found: boats, bikes, sunsets, selfies and wristbands.

The wristbands form an interesting cluster to explore as a form of visual presencing: users clearly position themselves in the Oerol festival, they even visually ‘chain’ themselves to it by photographing their own wrists with the wristband that is needed for entrance to Oerol activities. Often these images show a group of wrists, showing the experience of being in the festival together. From interviews held on the island and exploring Instagram in years before, we know that for many people this photo is a traditional one that they make to show that they are there again.

Choices and limitations

The network created this week has been made as a test drive. It could have been worked out better in Gephi to get clearer clusters. Now images are still somewhat shattered across the canvas. We have here chosen to go with the labels attributed by the Google Vision API and not clean up the data set to much. Next time we would take more time exploring the labels and possibly removing some of them or grouping some of them. For instance the labels recreation and fun were given often. Possibly removing these more general/abstract labels would have made the clusters more prominent. Also some of the labels were quite similar and it would’ve maybe make sense to combine these, for instance text and font.

Place

Mapping visitors posts allows for the visualization of the distinct geography of the Oerol experience, which extends beyond the bounded location of the event itself. The movement of people two and from the event can be tracked, highlighting a clear migration path. GIven the event’s unique island location, a significant aspect of the Oerol experience is the ferry ride to the island. Image posts support this and unsurprisingly, those georeferenced as on the ferry route contain content of water and or boats. Areas for further analysis could include symbolizing by hashtag themes, Google Vision API codes.

Conclusion/discussion

Exploring the original images organised by time, content and place, has provided us with an interesting way in for visual analysis to answer the question how users visually place themselves in the here and now of the event-sphere of Oerol 2017 on Instagram. Using selfies and first person perspectives, with specific meaningful items or scenery (eg wristbands and sunsets), visitors position themselves in the context of the festival and bring that position to a wider social network.

Bringing this visual data set together with collected ethnographic data (74 interviews held on the island during the festival) will expectantly lead to very interesting thick descriptions of these practices. Where the images show us what happens online on public profiles, people can tell us more about the experience of sharing images, about the choices they make, about their experience of what Instagram brings them, about how they live and re-live Oerol through their Instagram posting.

References

Auslander, P. (2012). Digital Liveness: A Historico-Philosophical Perspective. PAJ: A Journal of Performance and Art, 34(3), 3–11.

Couldry, N. (2004). Liveness, “Reality,” and the Mediated Habitus from Television to the

Mobile Phone. The Communication Review, 7(4), 353–361.

Gibbs, M., Meese, J., Arnold, M., Nansen, B., & Carter, M. (2015). #Funeral and Instagram:

death, social media, and platform vernacular. Information, Communication & Society, 18(3), 255–268. https://doi.org/10.1080/1369118X.2014.987152.

Meese, J., Gibbs, M., Carter, M., Arnold, M., Nansen, B., & Kohn, T. (2015). Selfies at

Funerals: Mourning and Presencing on Social Media Platforms. International Journal of Communication, 9(0), 14.

Vianello, R. (1985). The power politics of “live” television. Journal of Film and Video, 37(3),

26–40. https://doi.org/10.2307/20687672

-- NataliaSanchez - 18 Jul 2017