The map tool supports selecting subsets of folktale data based on search queries. It visualises the folktales on a map in combination with a timeline (see Figure 2). This dynamic visualisation utilises a faceted navigation technique [English 2002], which can be used to obtain subsets of folktales meeting specific selection criteria and narrow down on folktales of interest. The facets are shown as pie charts on the right side of the interface (see Figure 2) and can be clicked to update the results on the map. The facets are updated as well to show the contents of the subset. The visualisation starts with a full dataset overview from which cross sections can be made. At the moment, the tool only visualises the locations where the folktales were narrated and not, for example, the locations mentioned in the folktales. Still, the completeness of overview of the collection and therefore the accessibility of all corners of the collection is apparent.

The interface relies on shapes and colours to represent several properties of the data. A circle on the map represents a location where at least one folktale was narrated. The size of a circle represents the number of tales in a location. Transparency can be set for the circles to create an effect similar to a heat map, which can give a clear picture of the concentration of folktales in certain areas. To make individual circles more noticeable, they can all be made the same size. This way, small circles will not be hidden behind large ones. Folktales that have not been annotated with a location are currently shown in the North Sea; this is a temporary solution to keep them from being overlooked.

Geographical mapping as a principle can serve as a starting point for international cooperation between folklore researchers by linking their national collections, as advocated by Meder [Meder 2010]. Due to various factors such as ownership, differences in metadata and opinions about content, it is very hard to establish a single international database of folktales. It seems more feasible to let folktale databases exist in their own context, and promote the “donation” of content to a central system. A logical assumption that can be made about folktales is that, with the exception of Internet based tales, they were recorded at a certain time and place. This means that all folktale collections are expected to contain at least time and place as metadata, which makes a map-based interface very suitable as a collective interface.

Similarity between folktales is of great interest to researchers who are searching for the origins, kinship, variability and stability of tales [Tehrani 2013]. However, most empirical investigations of narrative similarity have only been carried out on a small scale so far [Nguyen et al. 2014]. In our visualisation tool, the similarity of large numbers of folktales can be determined based on different configurations of metadata such as folktale type, keywords, named entities, word count, genre, etcetera. This comparison yields a score that can be used to determine their degree of similarity. In line with Abello et al., our system visualises the similarity between tales as a dynamic network graph (see Figure 3) [Abello et al. 2012].

The tool can be used from different starting points. A single folktale can be loaded to which neighbours, folktales with similar metadata, can be added. These neighbours can subsequently be interconnected and added upon as well. This way, an entire network can be built up from a single item. Another starting point is to do a keyword search to obtain multiple folktales as nodes. These can then be interconnected and added upon as well.

Once a network has been built, it is possible to start exploring grouped nodes and links by clicking on them. There is also a possibility to select multiple nodes to show the composition of this subset, using the same pie chart shaped facets as are available in the map tool. By doing so, the user is invited to find new perspectives to compare folktale stories. The tool gives the user the freedom to hypothesise and test which metadata fields are valuable to compare in order to gain new insights into the relations between folktales, and how these relations evolved over place and time. It could be used to verify existing theories and classifications that have been based on qualitative research. For instance, when connecting folktale nodes purely based on keywords, we found that clusters are formed that show good overlap with the frequently used Aarne-Thompson-Uther (ATU) folktale type-index [Uther 2004]. In other tests with the network tool, users have already reported finding unexpected connections between tales. This is because the tool connects texts outside traditional classification systems that experts commonly use, such as ATU. These early findings suggest that in the future, the tool may open up new avenues for folktale research that go beyond what was previously possible.

A simplified network visualisation has been developed based on the network software. It is included on each folktale page on the Dutch Folktale Database website, see Figure 4. Around the folktale of interest a “flower” of similar documents appears when such documents are available. Users can inspect these similar documents by clicking on them. We believe that both experts and casual users will appreciate this addition, as Elbedweihy et al. found that both groups like their search results to be augmented with related information [Elbedweihy et al. 2012].

Both the expert Map tool and the expert Network tool are still under development. We expect their current interfaces to be somewhat hard to understand, because the focus thus far has not been on user friendliness but on functionality. Thorough user evaluations with expert users still need to be carried out to investigate the usability of both interfaces, similar to what was done with the interfaces we designed for casual users, as discussed in the next sections.

The main access point to the Dutch Folktale Database used to be a simple search box. As noted by Whitelaw, the trouble with this is that it requires a query from the user, even though many users are not seeking specific information [Whitelaw 2015]. Our goal was therefore to design a new search interface that would allow the users to make use of metadata to search the collection, like in the advanced search (see Figure 1), but with an easy-to-use faceted search interface instead of a list of search fields. Such an interface should provide the users with a good overview of the collection and of the different dimensions it can be searched or browsed on. We expected this type of interaction with the collection to be intuitive, as most users will already be familiar with faceted search interfaces from popular web shops such as Amazon.

Our first step was to investigate which of the folktales’ metadata were of interest to casual users, as opposed to domain experts. To this end we recruited a test group of 9 potential users and presented them with a list of metadata that could potentially serve as facets. We asked the participants to indicate for each potential facet whether they would want it to be included in the search interface. The ages of the participants (4 male, 5 female) ranged from 21 to 78 years, with an average age of 43 years. This resembles the age distribution of the visitors of the Dutch Folktale Database. Of the participants, two were scouting leaders and one was a teacher on a primary school; these are the kinds of casual users that would occasionally visit the Dutch Folktale Database looking for stories to tell by the campfire or at school. None of the participants were familiar with the Dutch Folktale Database.

The list of facets a majority of the users showed interest in is shown in Table 2. Examples of potential facets that did not make the selection are “motif”, a term from folktale research that was not familiar to the participants, and “literary”, a field with a yes or no value indicating whether a collection item was a literary text or not.

The next step was the design of the search interface. We created several alternative designs and presented them to six of the participants who had judged the potential facets. In our design choices we mainly focused on various options to present the facets, e.g., listing selected facets next to each other or below each other; showing selected facets together with non-selected facets, or in a separate section; showing all possible values of each facet or only the most frequent ones, etcetera. Based on the users’ preferences, we created the final design shown in Figure 5. Selected facets are shown below the main search box, but also among the non-selected facets in the menu on the left. The list of shown values for each facet is limited to a maximum of the eight most frequent ones, but users can fold open the list for more values. After each value, the number of matching items in the current result list is shown. When no search terms have been entered and no facets have been selected, the numbers for all items in the database are shown, giving the user an overview of the entire collection. The new search interface was implemented using an Omeka plugin called SolrSearch, which implements SOLR to become the primary search engine on Omeka-based sites. SolrSearch was made by researchers of the ScholarsLab of the University of Virginia (https://github.com/scholarslab/SolrSearch).

We tested the usability of the new search interface using a Dutch translation of the Standard Usability Test (SUS) [Brooke 1996]. In this test, participants indicate their level of agreement with ten statements on a Likert scale ranging from 1 to 5. Example statements include, “I found the system unnecessarily complex” and “I thought the system was easy to use”. The new design was tested by nine participants recruited from a public library. We expected library visitors to be more likely than the average person to have an interest in folktales, and thus to more be representative of our target group of causal users. To compare the usability of the new search interface with that of the old version, we also tested the old search interface. The test of the old version involved a total of 14 participants: five library visitors, eight bachelor students from our university (study programme: Creative Technology), and one actual visitor of the Dutch Folktale Database who reacted to a call for participants on the website. None of the participants had used either of the tested interfaces before.

Participants were given a brief introduction to the Dutch Folktale Database, and were then asked to do a search using an initial query of their own choice. They were encouraged to explore the results page at their leisure, after which they filled in the SUS questionnaire.

The results of the SUS test are calculated as described by Brooke [Brooke 1996]. A total SUS score of 68 (out of 100) is seen as average. In our test, the old search interface received a SUS score of 60.89, which is below average (Figure 6: SUS score V1). The new faceted design received a SUS score of 81.39, which is well above average (Figure 6: SUS score V2). We can conclude that the new design is a marked improvement over the old search interface. However, a few critical remarks are in order. First, those who tested the new search interface were more evenly distributed in age than those testing the old version, which did not include the age groups 25-34 and 35-44. Second, the students who tested the old search interface might have been more critical of it due to their academic background, and less interested in the contents of the Folktale Database. This may have resulted in a lower SUS score for the old search interface than would have been given by users who were more representative of the target group. Still, Figure 6 shows similar trend lines for both interfaces, so we feel it is safe to say that the overall usability of the new search interface is higher than that of the old version.

In a short poll on the website of the Dutch Folktale Database (Spring 2015), 152 out of 240 respondents (63%) said they would like to make use of an interactive map in the future. Therefore we created a second interface, based on the map tool for folktale researchers that was described earlier in this paper. We used the same technology as in the expert map tool, but designed a new interface more suitable for casual users.

Given the complexity and size of the data that can be portrayed with maps, especially interactive ones, there is a strong need for simplicity in design so that the interface remains clear and overseeable for users [Harrower and Fabrikant 2008] [Jones et al. 2009]. We established the following requirements for the map-based user interface, based on the user interface guidelines specified by Brown [Brown 1999]:

These requirements can be contradicting; for instance, a minimalistic interface might emphasise the database content well, but could come at the cost of being unintuitive. It is hence very important to find the right balance between the requirements. For this reason, we carried out numerous informal user tests with different interface designs.

The main concept was to provide a map showing folktales in the form of bubbles (similar to the map tool for folktale researchers) and only a few minimal “filter buttons” corresponding to facets that can be used to select subsets from the collection. Figure 7 shows the initial design of the filter buttons. After clicking a filter button, the values of the facet are folded out for selection or deselection by the user. The folktales matching the user’s selection are shown on the map. This way, the users get to see only the stories they are interested in, while still being invited to explore the other folktales in the database.

We included the following filters: folktale subgenre (e.g., legends, jokes), source type (e.g., book, Internet), language/dialects (e.g., standard Dutch, Frisian), keywords (e.g., death, man), and tale collectors. We selected this last category instead of creators/tellers, because users looked more often for collectors than creators in the transaction logs analysed by [Trieschnigg et al. 2013a]. We made a conscious choice to strongly limit the number of facets, preferring minimalism and simplicity of the interface over completeness.

Based on informal feedback of several potential users on our initial design ideas, we created a first mock-up of the interface (Figures 7 and 8) with a few interaction possibilities, but no actual functionality. In a small-scale user test, four people (age range 20-50, two male, two female) solved a few simple tasks with the mock-up. This test showed that some of the icons we used for the buttons were unclear. However, the biggest problem with the interface design turned out to be the visualisation of multiple folktales at the same location on the map. In the mock-up we used a bookmark metaphor for this, using tabs to indicate multiple folktales per location (Figure 8). The test showed that users did not notice these tabs very quickly, and when they did, it was not clear to them that the tabs correlated with multiple folktales. In our final interface design, shown in Figure 9, we therefore used a fundamentally different solution, indicated by (4) in the figure. The folktales are now shown as individual bubbles, placed in a circle around their location. The subgenre of each folktale is indicated by an icon within its bubble. The same icons are also used in the fold-out menu of the subgenre filter button. This provides users with an instant indication of the kinds of stories that are available per location. If the number of folktales per location exceeds the number that can be presented in the circle, a “More” button is included in the circle. When clicked, it shows the next set of folktales at that location. This new circle includes a back button to return to the previous set, and another “More” button if even more folktales are available.

In the final design we changed the icons of most filter button icons to show a clearer link with the depicted facet; see (1) in Figure 9. At the top left of the interface (2) we added a switch between the map interface (Verhalenkaart in Dutch) and the regular interface of the Dutch Folktale Database (Verhalenbank). Since we expected users to have an interest in folktales at specific locations, e.g., local legends, a location search feature was included. This allows users to instantly go to a specific place on the map by typing in a location. We also included a “Locate me” button for directly zooming in on the user’s current location to show folktales in their vicinity (3). Finally, we customised the map, using a muted colour scheme to avoid distracting the users, and showing only those details which are important for finding folktales (e.g., including town names but leaving out the names of highways).

To determine the usability of the final design, we created a prototype with limited functionality for a final evaluation experiment. The prototype did not have access to the actual folktale database, but contained all essential elements needed to test the interface and its functionality. The evaluation was carried out with 22 participants, 19 of whom carried out the test online. The remaining three interacted with the prototype in the presence of an observer. Participants varied in terms of their gender (59% male, 41% female), work or study (ranging from healthcare to ICT), and age (between 17 and 50 years, with an average age of 24). Most had received higher education. Only one of them had used the Dutch Folktale Database before.

Participants were given three tasks that explored all the major elements and core functionality of the map-based interface, including using multiple filter options and navigating between several folktales at a single location. After performing the tasks they were asked to fill in the Usefulness, Satisfaction, and Ease of use (USE) questionnaire [Lund 2001], which requires the users to indicate their level of agreement with a number of statements concerning ease of use, ease of learning and satisfaction on 7-point Likert scale. We also included a few interface-specific statements on the effectiveness of the filter options and the navigation between multiple folktales. Finally, we asked them to provide feedback on the interface. The average results of the questionnaire are given in Table 3. The participants were positive on all USE dimensions as well as on the interface-specific statements.

The participants commented that the interface was user-friendly, clear, easy to learn and tranquil. Negative comments mostly pertained to limitations of the interactive mock-up. Other comments included suggestions such as adding hover tooltips and adding extra animations to provide feedback. The three participants we observed while they interacted with the prototype quickly started using the filter buttons, understanding their purpose within a few seconds. No navigation problems were observed; the participants seemed to grasp the concept of multiple folktales around a location quickly and well.

The results indicate that we succeeded in designing an interface for casual users that is minimalistic, simple, appealing and intuitive. To which extent the interface encourages exploration could not be tested given the limited functionality of the prototype. Investigating this requires a longer-term study using the fully functional interface, which has only recently been integrated with the Dutch Folktale Database.