As the number of recommendations grows to tens or hundreds, it becomes increasingly difficult for the user to explore and analyse them. That is why we have put a lot of effort into developing tools that leverage the capacity of human vision to absorb and grasp a large amount of information at once. These visualisations of search results are accessible via the recommendation dashboard, which is integrated in the different applications, e.g. the Chrome extension or the WordPress plug-in.
The Recommendation Dashboard
The recommendation dashboard is accessible in the Chrome extension via this button on the left It provides several interactive visualisations that allow the user to analyse the recommendations along different perspectives (e.g. time, space, topics…). Currently, the recommendation dashboard provides the following visual representations of the list of recommendations:
- Keyword-based ranking refinement (uRank)
- Topical cluster analysis (“landscape representation”)
- Geographic analysis (GeoView)
- Timeline navigation
- Viewing distribution of different categories (e.g. source or language)
- Ranked recommendation list
Additionally, the recommendation dashboard provides features for bookmarking items and for setting multiple filters.
1. Recommendation List: Shows the relevance-ranked list of the recommended data set.
2. Visualisation Panel: Shows a visualisation of the currently selected recommendation set.
3. Tools Panel: Provides buttons to 1) adjust the mapping data onto colours, 2) reset the recommendation dashboard to its initial configuration, 3) to provide feedback and 4) to add a set of selected recommendations to a user-defined collection (bookmarking)
4. View Selection Panel: Lets the user choose between five different visualisations of the recommendation list.
5. Filter Panel: Shows the current set of filters using micro-visualisations.
You can open any recommended item by clicking on the respective list entry. To analyse a set of recommendations, the different visualisations that can be selected at 4. may be useful. The filter panel at 5. shows you a respective micro-visualisation and gives you the opportunity to lock or undo the filters you have selected within the different visualisations.
The timeline visualisation shows the temporal distribution of time-references included in the recommendation metadata.
GeoView shows the distribution of places taken from the recommendation metadata. To avoid clutter and overdraw when there is a large number of geo-references, we employ overlaid “donut” charts to aggregate the data elements. When you zoom into the map, the aggregation is split and dynamically recomputed to reveal more details.
Distribution of recommendations over the categorical metadata, such as language or the data provider, is shown using the bar chart.
The topical landscape visualisation is a visual metaphor, which shows the distribution of major topical clusters. Each recommendation is represented as a coloured icon.Topically similar recommendations are placed close to each other while dissimilar ones are positioned further apart, resulting in “islands” and “hills” that represent clusters of topically similar recommendations. Topical clusters are labelled with high frequency keywords extracted from the contents of the corresponding recommendations.
The user can select an area of interest in the landscape. This triggers the computation of the keywords for the selected recommendations.
uRank assists the user in picking interesting recommendations by a) providing an outline of the major keywords covered by the recommendations in the form of a tag cloud and b) enabling the user to select and weight the interesting keywords. Select the keywords by dragging it from the tag cloud (on the right) to the tag box (on top). The importance of the keywords can be modified by using a slider. The recommendations are then re-ranked depending on the user´s choices and visualised using stacked bars.
Are you becoming curious? You can explore the various features and opportunities of the visualisations by playing around with them. We would be glad to get your feedback. Get the EEXCESS Chrome Extension here!