In interactive systems, like the Zoom Illustrator, figure captions describe images exposed to changes. In practice, nearly all interaction on images lead to incomplete or even invalid figure captions. Hence, invalid parts must be detected automatically. Furthermore, these parts must either be updated or their status has to be communicated by means of an appropriate layout (see Figure 2). This raises the question of when the caption is to be updated and who initiates the update process. We utilize three approaches:
Update on explicit request only. The caption is updated only upon the user's request. The system indicates invalid parts.
Automatic update. Compared to the other variants, automatic updates by the system require fast generation of captions. This may confuse users because of the high frequency of changes in the corresponding area of the screen. However, the advantage of this approach is that captions remain consistent with the image at all times.
Hybrid variant. The caption is updated automatically only if radical changes in the image occur. Examples are considerable changes in the viewing specification (e.g.via a rotation) leading to a change in the visibility of a large number of objects or the removal of a class of objects due to filtering operations. Another example is the incorporation of an additional view which requires the figure caption to describe both views.
As there is no optimal variant for all users, however, leaving the decision on the initiation of an update up to the user seems to be the preferable option (see the update frequency option the bottom line in Figure 3).
Figure 2: A caption has sensitive parts (with an underlying rectangle) and invalid parts (light grey and italic). The interactive handling of the visualization with sensitive parts of figure captions is described in . Bold parts represent fixed phrases which are based on the templates.
Another aspect of variation concerns the fact that figure captions in interactive systems may change in length over time. The incremental completion of an existing figure caption where constant parts remain at their positions eases the identification of changing parts. This strategy may, however, conflict with a well-balanced layout where captions are centered below an image.