Tagging is endemic on social media sites such as Twitter and for some time it has been an important feature of photo sites such as Instagram, Flickr and Delicious. A body of research on tagging has developed, and the existing lit-erature examines how tags contribute to ‘folksonomies’, that is taxonomies created by combining the tags of many users. Typically these studies have examined large data sets of tags on particular sites and this work has been mainly from a web designer’s and information scientist’s approach, where tagging is used as a hyperlink bringing together all uses of a specific tag. In this chapter I am interested in complementing this approach with a user’s view of tagging. This will be addressed by investigating in detail how people use tags on one particular site, Flickr, by drawing on a social practice view of language use online. This emphasises users’ practices and how tags provide a writing space with particular affordances which users build upon. This study enables us to see people’s purposes when tagging, how they design their sites in deliberate acts of curation, how tags are used more than just as parts of a folksonomy, and what is lost when discussing tags away from the pages where they are being used.
First, I will begin by explaining briefly a social practice view of language and how it is appropriate for studying tagging. This is followed by an over-view of the methodology. The literature on tagging which relates to Flickr is then reviewed and this then leads on to the main part of this chapter, analyses of the tagging practices of two sets of users. The aim here is to understand people’s practices, both in creating tags and in using them.
This chapter utilises a social practice approach to language online, developed from literacy studies (Barton and Lee 2013). This is an approach which can provide a way of examining texts and practices online. It starts out from what people do with language in their lives, locating this in broader social practices. What people are doing online can be described in terms of their language and literacy practices, in the sense of the commonly recog-nised patterns of activity where people bring their cultural knowledge to an activity. Practices are realised in specific events, such as uploading, tagging, searching and browsing. Emphasising the importance of language, a social
practice approach sees online spaces as part of a textually mediated social world. Even an intensely visual site such as Flickr has a large amount of lan-guage on it and each photo can be surrounded by lanlan-guage serving many functions. For instance, when uploading a photo to Flickr the user is invited to add written titles, descriptions, tags, and more.
Methodology for investigating tagging as a social practice
The idea of online activity as consisting of a set of practices that are inferred from what goes on in events, which themselves are mediated by texts demands a research methodology that pays close attention to the detail of particular instances. This is an ecological approach which keeps to situated examples and is aware that what users do both affects and is affected by the perceived possibilities. There are not ‘effects’ of technology – rather there is the complex interplay of affordances in people’s purposeful activities.
This approach has the very strong idea of an active user, and not someone passively responding to the design of a site.
It is also important to stress that online life is essentially social and that the role of other people, online and offline, is crucial. Therefore a social practice approach is interested in the networks and other groupings people participate in, noting the fluidity and flows in such online participation.
This is the approach which this paper is embedded in. The specific meth-odology involves revisiting existing data and locating tags in the context of their use. A related issue, the importance of how we talk about, visualise and represent tags, is a thread woven throughout the chapter, examining how different discourses and visualisations of tags represent or misrepresent tag-ging practices.
In this study evidence about tagging practices comes initially from exam-ining the texts, that is the web pages contaexam-ining the tags, and then from online interviews with some of the creators of the web pages. The web pages are treated as multimodal texts with distinct ‘writing spaces’ such as the spaces for titles, for tags and for comments. Each writing space has its own dynamics and has to be seen in relation to the other writing spaces on the page, to images and to overall layout. The tagging space is analysed in terms of content analysis and by drawing on linguistic analyses of discourse including examining cohesion and coherence, along with stance analysis.
After initial analysis of the web pages, some of the individual creators of the web pages were then contacted and asked to answer a generic online survey about their practices on Flickr. So for example, in a study about mul-tilingualism they were asked general questions about which languages they used and what affected their language choice. This was followed up by indi-vidually tailored online interviews which asked questions about specific web pages, such as why they chose a particular language in relation to a specific
image. The aim was to get the users to reflect on their practices by looking in detail at specific examples of their own pages. The analysis combined textual data with interview data.
This literacy studies methodology resonates well with approaches to discourse analysis which combine textual analysis with broader cultural knowledge, such as Wodak and colleagues’ discourse-historical approach to critical discourse analysis (Reisigl and Wodak 2009) and mediated dis-course analysis inspired by the work of Ron Scollon (Jones and Hafner 2012;
Norris and Jones 2005). The methodology also links up with current work in Linguistic Ethnography (Tusting 2013). While not ‘an ethnography’ we would argue that this is an ‘ethnographic approach’ in that people’s per-spectives are highlighted, the analysis is situated in a broader cultural con-text and that it aims to be naturalistic with low levels of intervention. In this particular study there was also reanalysis of existing data meaning that we already knew a great deal about these users.
The study is multi-method and combines qualitative and quantitative approaches. The analysis of websites and the interviews led to a focus on individual cases, as will be seen in this paper. There is a repeated going back and forth between individual cases and a wider perspective as a way of see-ing how individual interpretations hold up across broader data.
Tagging systems as folksonomies
Before examining the existing literature, and to introduce the topic of this chapter, Figure 4.1 provides an example of a set of tags related to one photo on a Flickr page. As can be surmised from the tags, the photo being described on this Flickr page consists of an Egyptian sculpture in a New York museum.
The photo page was created by a multilingual Spanish speaker from South America who was visiting New York. She was one of our informants on a study of how multilingual users of Flickr deploy their languages online
Figure 4.1 An example of tags (from http://www.flickr.com/photos/46836654@
N00/460544898/in/photolist-GGpXh-21NQnZ-5bb14d, last accessed 17 March 2014)
(Lee and Barton 2011). The page is also discussed in Barton and Lee (2013) and I will return to the example later in this chapter after reviewing some of the existing literature on tagging.
The idea of tags as being part of some form of folksonomy has been the prevalent view in research on tagging. A folksonomy is a taxonomy cre-ated by collating tags from a large number of people’s tags and it was seen originally, when the term was invented, as an unintended consequence of individual people’s actions (Vander Wal 2005). There was a cluster of stud-ies around 2005–6, often collating tags from a large number of users into a folksonomy. These studies were often carried out from a librarian and information science perspective (such as Heckner et al. 2007; Winget 2006) or by the site designers themselves (as in Marlow et al. 2006 whose research was done in association with Yahoo).
In their useful early study of tagging, Marlow et al. (2006) point out that tagging systems work very differently on different sites and this must be borne in mind when discussing the particular case of Flickr. They compare Delicious (formerly del.icio.us) and Flickr, providing a set of seven aspects of design which affect the tagging. These are useful in demonstrating the variety of forms of tagging online and can be discussed here in relation to Flickr. Sites vary on what can be tagged and who can tag. By default any member of Flickr can add a tag to a photo. Sites vary in what support there is for tagging, whether there is a limited set of possible tags and how the tags are presented. Some tags may be provided by the site, such as date and make of camera. Tags then provide links to other photos with the same tag and they enable users to link with each other. In comparing Delicious and Flickr, Marlow et al. show how the two sites differ on all these dimensions.
The idea that tagging differs from site to site can be seen by examining other sites such as YouTube, Twitter, Pinterest and Instagram. Looking at Flickr, not all the possibilities offered on the site are taken up by users. For example, although by default anyone can add tags, in practice most tags are created by the photographer and although they can add tags at any time they tend to create the tags when they initially upload the photo.
One study which has investigated people’s motivations for tagging is a small qualitative study by Ames and Naaman (2007) which interviewed Flickr users. This work was also associated with Yahoo and the aim was to encourage users to tag more. They found that the motivations for tagging were many and varied and they identified a set of categories for motivations to tag. These included social reasons such as to make photos searchable to others, as well as personal reasons such as to help the user organise and find their own photos.
The designers provide the space which people act within. The list of dif-ferences between sites, discussed earlier, is a description of the possibilities they envisage for the site. People act within these possibilities, taking up some opportunities, ignoring others and creating new activities which the
designers never dreamed of. It is this creative space between the designer and the user where the unexpected can happen which constitutes the affor-dances of tagging. Understanding the space between what the designers make available and what people then do within these spaces is the topic of this paper. In fact there is a remarkable range of possibilities on Flickr, many of which are hardly taken up, but at the same time there are strong constraints on what is possible.
To give an idea of what tags get used on Flickr, it may be useful to also show a visualisation of some of the most common tags on the site, as in Figure 4.2.
This is a tag cloud where size indicates number of instances of the tag.
These are gathered together into an alphabetical list where the frequency is shown by the size of the word. This constitutes a folksonomy. It is an important visualisation of tags, and differs, for example, from a vertical jus-tified alphabetical list of words all in the same font size. The form of these visualisations is created and made available by the designers of the site, not by the users. Visualisations of language are important and they can help and hinder us; here we see only the most common tags, and something else may be going on with the less common ones. Crucially it is in people’s own words, and that is the important distinction from a scientist’s or a linguist’s taxonomy, or the anthropologist’s folk taxonomy (which is in the anthro-pologist’s words).
The list shows a certain range of topics as representing the discourse world of Flickr. There is international travel, holidays, family rituals and festivals. ‘Wedding’ is the most common tag (apart from the machine tags),
Figure 4.2 The most common tags on Flickr (http://www.flickr.com/photos/tags/, last accessed 12 February 2014)
with around 17 million instances. In terms of parts of speech, it is mostly nouns including places, along with adjectives of colour. Some are machine tags which are automatically added, such as the names of cameras Canon and Nikon. Other parts of speech are there, including the deictic pronoun
‘me’ which is used nearly 4 million times.
Looking more broadly at all the tags used on Flickr, rather than just the most common, there are all sorts of words including technical terms, dialect words, obscure words, abstractions, abbreviations, and many written languages are represented. All parts of speech are there. There is consider-able deixis – ‘here’ has 80,000 instances and ‘there’ 46,000 instances. Even
‘whence’ has 213 instances and ‘whenever’ 492. There are several misspell-ings so, for example, ‘writting’ gets 8744 instances, some of which may be deliberate misspellings. Some aspects of the image are automatically added elsewhere on the page, such as when the photo was taken, and increasingly what make of camera, the camera settings and geotags of where it has been taken.
In the work of the mid-2000s comparing Flickr and Delicious, Guy and Tonkin (2006) review some of the literature on folksonomies, looking at what makes them work and listing some of the problems which other researchers identify. This is a useful review of the work at that particular time. They begin by pointing out that
the number one gripe for those happier with more formal classification systems–is that the tagging terms used in those systems are imprecise. It is the users of a folksonomy system who add the tags, which means that the tags are often ambiguous, overly personalised and inexact.
They point out how many studies are critical of tags which are only used once as well as the use of ‘nonsense’ tags designed as unique markers that are shared between a group of friends or co-workers. The result is seen as an uncontrolled and chaotic set of tagging terms that do not support search-ing as effectively as more controlled vocabularies do. What they refer to as
‘sloppy’ tags included misspellings, compound words, single use tags and personal tags.
In their own study of tags on Flickr and Delicious, Guy and Tonkin found many such ‘flaws’. ‘By testing against multilingual dictionary software,’
they write, ‘we found that 40% of Flickr tags and 28% of Delicious tags were either misspelt, from a language not available via the software used, encoded in a manner that was not understood by the dictionary software, or compound words consisting of more than two words or a mixture of languages.’ Overall, they state that ‘Somewhere around a third of tags were indeed “malformed”, in that they were beyond the grasp of a multilingual spell-checker for one reason or another.’ Ironically, given the more recent ubiquity of hashtags on Twitter and many other sites, one of the complaints
of that period was the use of symbols in tags and they observe that: ‘Symbols such as # were used at the beginning of tags, probably for an incidental effect such as forcing the del.icio.us interface to list the tags at the top of an alphabetical listing.’
As a solution, Guy and Tonkin point out how researchers at that time were identifying the need to improve ‘Tag Literacy’ which would include
‘Educating users to add “better” tags’ and ‘Improving the systems to allow
“better” tags to be added’, and they point to various lists of tag selection
‘best practices’ which could lead to general guidelines for users. These would include: ‘using plurals rather than singulars; using lower case; group-ing words usgroup-ing an underscore; followgroup-ing tag conventions started by others and adding synonyms’. In reviewing this area Guy and Tonkin are careful not to identify with this critique, which I would refer to as a ‘deficit theory’, very similar to other areas of public denigration of areas of literacy. This is seen in the terms used such as ‘chaotic’, ‘nonsense’, ‘sloppy’, ‘flawed’ and
‘malformed’. (Elsewhere Marlow et al. 2006 mention another author refer-ring to ‘feral’ tags.) Nevertheless, Guy and Tonkin are more even-handed.
They see the problem for tagging systems as being the way they ‘are trying to serve two masters at once; the personal collection, and the collective col-lection’. And that tags are not necessarily ‘sloppy’ or ‘bad’. They point out that ‘revisiting the data with another aim in mind might reveal usefulness in some categories of “sloppy” tag’. They raise questions about whether tags have a use beyond being search items. I would push this further and say that starting from people’s tagging practices, a quite different view is apparent.
Tagging on Flickr
Turning to the present study and details of Flickr, it is a distinctly multi-modal site, where the central focus is on images but, nevertheless, a great deal of language is involved. On any online site there are distinct writing spaces, each with their own affordances. On a Flickr photo page these include space for a title, a description, tags and comments, and there is a link to the person’s profile page. Photos are surrounded by writing. As part of the fluidity of the online world, this layout has been changed by the designers several times over the past few years, and the user has little con-trol over the overall layout. As of May 2013, the title is in a larger bold font and superimposed over the bottom left hand side of the image. It is limited in length (apparently to 155 characters). Below this in a smaller font is the description space which can be empty or can contain several pages of text.
Tags appear as left to right text with a space between each tag (and, apparently, there can be up to seventy-five tags although in practice people rarely have more than twenty). Writing is also involved in describing the sets (themed groupings of one’s own photos) and groups (themed groupings with other people’s photos). Another important writing space, which will
not be discussed in detail here, is the space for comments, which is available below the image.
The focus of this chapter is people’s practices of tagging. What are users endeavouring to do? What are their purposes, and how do they use this writing space in relation to other writing spaces? The aim is to investigate how the act of tagging is situated in people’s broader practices and what
The focus of this chapter is people’s practices of tagging. What are users endeavouring to do? What are their purposes, and how do they use this writing space in relation to other writing spaces? The aim is to investigate how the act of tagging is situated in people’s broader practices and what