users’ interaction with tags and tagging, and information behaviour in general on a certain website, Cancer.dk. I studied a diversified material and thus had a chance to study and describe the case from different angels. In a user-oriented study, the user is seen as a part of the system (Järvelin & Ingwersen, 2010). With tags, the user also provide metadata into the system, and thus user behaviour is crucial in understanding the whole system.
2.2.1 CONTACT AND COOPERATION WITH THE DANISH CANCER SOCIETY
Tor Øyan was our contact person in the Danish Cancer Society, as the chief editor of Cancer.dk. After the initial contact, the FIRE team had a workshop with representatives from Cancer.dk and their vendors of content management system and search engine, respectively ProActive and Ankiro. We shared our thoughts about tagging and discussed how to introduce tagging to Cancer.dk.
A period of prototype building and testing followed the workshop. This included the preliminary studies reported in chapter 5, where I analysed Blogomkraeft.dk tags and compared them with the browsing structure of Cancer.dk. I also conducted a usability study based on a tagging feature prototype. After this, ProActive produced a finalized tagging feature for Cancer.dk, including a logging feature.
Øyan and his co-editors wanted there to be tags in the system from the beginning.
Thus, the first week only employees of the Danish Cancer Society had access to, and were encouraged to apply tags. Then Øyan reported to be live with the tagging feature available to all users. He wrote, “It looks as if the first external tag was; selleribøf = celerysteak”. This tag has a timestamp valued: “2011-11-30 10:36:49.297” in the transaction log.
2.2.2 MIXED METHODS
To study the tags gives an insight into what people actually do. In this case, the activity was high, but it is not obvious what really happened: Why and how did users apply tags? The interviewed users in this study are too few to give a general view on taggers purposes when they apply tags. A single tag can also serve several purposes. Thus, one cannot always expect to make conclusions on the purposes of a single tag, even if I had a complete list of possible purposes users have when they apply tags at Cancer.dk. On the other hand, the quantitative and the qualitative data do shed light on each other’s, which is why I chose these varied methods in the first place.
A combination of qualitative and quantitative methods give holistic view and thus a clearer picture of the case Cancer.dk (Bergman, 2008). For short, the quantitative data, from log files, give data on what is going on, and make it possible to find out whether a phenomenon is frequent or not. In this thesis, I have counted tags of various types, and then compared the number of tags in different categories. On the other hand, with the qualitative data I wanted to find explanations. For instance, when I see tags as
“wrong” according to my own or the editors’ expectations, or according to indexing standards, the qualitative interviews can give explanations to what role such tags can play.
The following chapters (2.2.3-2.2.6) give short descriptions the various studies conducted for this thesis.
2.2.3 PRELIMINARY STUDIES
The purpose of the preliminary studies was to explore whether and how social tagging could support user interaction and information retrieval on an information website like Cancer.dk, and how to implement social tagging in a way that supported this purpose.
2.2.3.1 Blogomkraeft.dk tags and Cancer.dk site structure
Blogomkraeft.dk was a blog site and a part of the Cancer.dk web. The tags applied to postings formed an extended narrow folksonomy (see chapter 3.4). There were 650 tags in total, and 344 unique tags, applied to 318 blog postings. The blog tags originated from selected users, but still users in the target group of Cancer.dk, and it covered the field of cancer and cancer treatment. Because of these similarities between Blogomkraeft.dk and a future tagging feature on Cancer.dk, I used Blogomkraeft.dk as an indication on what to expect from Cancer.dk. The content of the blog tags on Blogomkraeft.dk were analysed by categorizing all tags according to their meaning.
Cancer.dk also had a site structure, available as a sitemap on the site. The items in the structure can be compared to terms in a controlled vocabulary, where every menu item and link anchor gives information about the aboutness of the connected article. I compared the two, assuming that the result would indicate what tags could add to Cancer.dk. If it showed that tags only repeated the structure and metadata that was already there, this would indicate that there was no need for tags at all.
I did not compare individual tags and sitemap items but grouped them into categories based on the aboutness of each tag and site structure item. The number of tags in different categories gives a good picture of the important features of the collection of tags. The conclusion was that, based on experience from Blogomkraeft.dk and the sitemap of Cancer.dk, that tags have potential to support user Cancer.dk. For more details, see chapter 5.2 and Ådland & Lykke (2012).
2.2.3.2 Usability study
From the study of Blogomkraeft.dk and Cancer.dk site structure, I learned that a tagging feature could be useful on Cancer.dk. A usability study was set up to find out how to implement a tagging feature, using a prototype developed after the workshop and discussions about such a feature.
The usability test was conducted in June 2010. Five participants used the prototype. I observed them and communicated with them during the test. Pre- and post-test questionnaires gave data about personal background, Internet experience and the participants’ understanding of and opinions about the prototype. With five participants, I see all data, both in the prototype sessions and the questionnaires as qualitative data. Together, this gave an impression on users’ views on and opinions on tags.
Our participants were able to operate the feature and liked its design and functionality.
Thus, when implementing tagging at Cancer.dk, there should not be big changes compared to this prototype. For more details, see chapter 5.3 and Ådland & Lykke (2012).
2.2.4 TAGGER STUDY
After the launch of the tagging feature on Cancer.dk, conducted an empirical study of user behaviour on Cancer.dk, with a focus on the role of tags. Like in the preliminary usability study, I collected data in diverse ways, in order to obtain a view on the research questions from different angels. This time, eight participants solved tasks using the newly launched tagging feature on Cancer.dk, they filled out pre- and post-test questionnaires, and I interviewed them when the tasks were completed. The goal was to find out more about users’ thoughts when faced with a tagging feature and tags: their understanding and opinions of tags and tagging, and their purposes when applying tags. The study resulted in qualitative data, structured through questionnaires and semi-structured interviews. For details, see chapter 8.
2.2.5 EDITOR STUDY
When the tagging feature had been live for about a year, I interviewed three editors about their experience and opinions on tags and tagging. The interview was semi-structured. I asked questions about background and experience in their job, but we spent most of the time talking about their opinions about and experience with tags and the tagging feature on Cancer.dk. The goal was to find out about their opinions, and to be able to compare this to the users’ opinions. Both the tagger study and the editor study gave a better understanding of the tags logged for the tag study (below). For details, see chapter 9
2.2.6 TAG STUDY
The transaction log is the largest data material in this project, and I have focused on the tags. I took account of all tags in this study. Many researchers use most frequent tags or do other types of selection before they conduct their studies. Examples are Pera, Lund and Ng (2009), Munk and Mørk (2007), and Morrison (2008), who all study selections of tags. This can be a good thing, but in this project, I found it interesting to study the whole collection of tags, to obtain a complete picture of what tags can be like.
The log file includes more than 25,000 tags, a huge amount of tags for a small system like Cancer.dk. As regard functionality and interface design the tagging feature remained the same for the whole period, but the location and visability at the cancer.dk website changed in September 2012. Thus, the collected data were produced in two
slightly different settings, with the tagging feature less visible for the users in the second setting and time period. These changes influenced the tags that users applied.
I mainly analysed these quantitative data through categorizing the tags into categories:
1. Internal and external tags – indication on who applied the tag
2. Lay or professional – do the tag belong to a lay or professional vocabulary 3. Topical facets – what is the tag about
4. Aboutness – the relationship between the aboutness of the tag and the aboutness of the article
I use these categories to explore what types of tags the users applied to Cancer.dk.
Together with the qualitative data, it was also possible to explain tags and tag categories. I can also indicate how extensive phenomenas found in the qualitative data are, based on the quantitative data. For details, see chapter 7.