Conception et evaluation de methodes de navigation dans des formulaires presentes sur des assistants numeriques personnels

Despite the remarkable technological progress of PDAs, the small size of the screens compels major constraints for information layout, hence the need to optimise the utilization of available space on screen. The main stakes are the provision of a general view of the interface content to users (especially in the context of e-forms) and the improvement of navigation in order to ease information search. Adapting and transforming paper forms of different sizes too easy-to-use e-forms on PDAs are difficult tasks. Currently, the two presentation models used in e-form design on PDA are the scrolling and the menu. They involve serious drawbacks with respect to information search and navigation, and do not offer a general view of the form content. In order to come up with new solutions for presentation models on PDAs, we analysed 37 paper forms used by inspectors and 72 forms used by students of Ecole Polytechnique de Montreal and identified their structure, their components, their common features and their differences. Main results show that the forms differ in terms of length as well as number and types of components, and led us to define different levels of length; nonetheless, their structure is always hierarchical. Thus, we propose to use two presentation techniques which could fit to this sort of structure: flip zooming and tree structure. We design and develop two new presentation models for e-forms on PDAs while adapting these two techniques. Two questions were raised then: what is the best performing presentation model for e-forms on PDAs, and to which extent the performance of the model depends on the level of length of the form? We contend that there exists a relationship between the level of length of the form and the performance of the presentation model and we test it. We carried out an experiment to test and compare four presentation models of e-forms on PDAs: scrolling, menu, flip zooming and tree structure, depending on the level of length of the form. The comparison is based on information search time and users' satisfaction. We retained three levels of e-form length: short, intermediary and long, given the number of sections and components. We designed the four presentation models of e-forms on a PDA for each type of form and developed a system to manage the experiment. Each of the thirty-six subjects, which participated to the experiment, tested the four models for one type of form. For each model, the subject achieved three tasks consisting of searching information through the interface. Also, we asked an expert to repeat the exercise and test the four models for the three e-forms in order to obtain a benchmark. Moreover, we calculated the raw time that corresponds to the required minimum number of stylus clicks to execute the tasks, then we multiplied it by the average motion time of the stylus on the screen (4 inches), which has been obtained from the experimental study that we had carried out with 24 subjects for that purpose and which equals 0.924 seconds. It consists of a second reference to subjects' search time. Main results show that there exists a relationship between the level of length of the form and the performance of the presentation model, for three models out of the four tested. The relative performance of the scrolling and menu models decreases as the length of the e-form increases, while the relative performance of the tree structure model increases as the length of the e-form increases. The flip zooming model relative performance is unrelated to the level of length of the e-form. Besides, the tree structure model is the fastest, followed by the menu, the scrolling and the flip zooming models. The tree structure model is 1.5 times faster than the menu model, 2.5 times faster than the scrolling model and 2.75 times faster than the flip zooming model. The latter is proven to be the slowest; nonetheless, the results of the experiment with the 36 subjects do not reveal its real potential since the expert search time and the raw time show that this model can be ranked as the second among the four tested. The outcome of the experiment led to the elaboration of a number of guidelines for ergonomic design of e-forms on PDAs. (Abstract shortened by UMI.)