Exploratory computing : designing engines for discovery driven user experiences

The research work described in the present thesis focuses on exploratory user experiences over information spaces, with the aim of devising a comprehensive framework of design patterns for supporting information discovery and sense making in large datasets, knowledge bases and corpora of multimedia content. The proposed model extends current faceted navigation mechanisms and encompasses the following issues: (i) semantic multifaceted knowledge representation, allowing to describe and integrate multiple data sources and different kind of classification (controlled vocabularies, semantic ontologies, folksonomies and data extracted from text and data mining), (ii) adaptive interface widgets for formulating highly expressive queries by navigation, and (iii) information visualization via advanced multivariate graphs (e.g. thematic maps, mosaic plots, histograms, etc.) for exploratory analysis of significant statistical correlations between the different facets (properties) of datasets. In particular, concerning the latter, a set of metrics and corresponding graphical representations is proposed to help the user inspect large datasets, suggesting what is possibly interesting according to the previous selections. \end{inparaenum} Experimental evidence comes from real case studies: a prototype for the Directorate-General for Antiquities depending from the Ministry of Culture of Italy, as well as portals for Politecnico di Milano educational initiatives. The proposed exploration paradigm is much more than the combination of search and hypertext browsing, being a different, complementary approach. It goes beyond the hypertext interaction model, since at each stage of interaction we have no longer just pointers to interconnected pages, but a representation of the whole information space for global understanding and sense-making. In such a context, the discovery process takes place by reshaping such a space, drilling down, zooming on or out from a specific "region" of it, and such elements may require more responsive and complex interface elements that "plain" hyperlinks, i.e. sliders, chart controls, tooltips and contextual menus. The user learns not just by observing each specific "screenshot" of the information space, but also from the sequence of applied moves and the consequent changes. The feedback the user receives is not just a matter of visualization and interaction strategies, but depends on the synergy of all the aforesaid aspects, and in particular on a strong underlying model and by proper relevance metrics. Finally, the proposed framework may pave the road the creation of a complete authoring tool for exploratory applications, aiding domain experts lacking programming skills in the whole process of design, prototyping and development of the user experience.

Il lavoro di ricerca descritto nella presente tesi è rivolto alla definiozine di un framework complessivo per le esperience utente esplorative in spazi d'informazione (es. grandi insiemi di dati o collezioni di contenuto mutimediale), allo scopo di favorire la scoperta della conoscenza e il sense-making. Il modello proposto estende i meccanismi attuali di faceted navigation, considerando sia gli aspetti relativi alla descrizione (semantica) della conoscenza, sia quelli relativi all'interazione (attraverso widget avanzati) e alla vusualizzazione avanzata (es. attraverso mosaic plots, tag clouds, mappe tematiche), supportata da adeguate metriche per rilevare la rilevanza dei concetti nello spazio informativo. Evidenze sperimentali provengono da casi di studio reali. quali un prototipo per la Direzione Generale delle Antichità (dipendente dal Ministero dei Beni Culturali) e una serie di portali a supporto delle esperienze educative organizzate dal Politecnico di Milano.