| dc.description.abstract | Data are the new oil and it is widely recognised the role of publishing them as Open Data to let data consumers freely access and
exploit them. Data providers are not only encouraged to publish
data but to ensure that available datasets are fit-for-use, meaning
that data users can directly exploit them without investing effort,
time, and money in performing data cleansing. The situation
becomes even more complex when data publishers deal with
data concerning individuals. Data in their raw form may contain
personal and sensitive information about people and publishing
them as are violate individual privacy. Hence, data publishers
need to apply privacy-preserving data publishing procedures by
publishing (sensitive) data without violating individual privacy.
Thus, data publishers before publishing data or data consumers before exploiting them require privacy-aware data cleansing approaches. Data publishers mainly opt for publishing data
in tabular format. Hence, data cleansing approaches should be
compatible with this format. As assessing and improving data
quality cleansing is time-consuming and expensive, the proposed
approaches should simplify as much as possible the procedures
to guarantee high-quality data by proposing (semi-)automatic
procedures. Moreover, data cleansing approaches usually require
specific expertise that limits the applicability of the proposed
mechanism. To ameliorate competencies requirements, novel
proposals should limit the required skills to favorite wider exploitation of data and their cleaning methodologies.
In this context, the first pillar of my research is placed: proposing (semi-)automatic privacy-aware data cleansing approaches
dealing with tabular data to make data users able to improve
Open Data quality while preserving individual privacy. It resulted in a series of approaches and prototypes, mainly integrated into a Social Platform for Open Data (SPOD) used by Public Administrations, such as the Campania Region, associations,
such as Hetor, and citizens, such as students joining activities to
familiarise themselves with the Open Data directive. While data providers mainly publish tabular data, data consumers might be interested in semantic reach data format, such
as graph-like structures, as they can be easily navigated and
explored thanks to their interlinked properties. However, directly
querying Knowledge Graphs requires expertise in query languages and awareness in the conceptualised data, which are
considered too challenging for lay users.
Hence, data consumers require Knowledge Graph exploitation
means being able to mask underlying technical challenges. Moreover, data users may require to consume data according to their
expertise, background, application contexts, needs, interests, capabilities. It requires designing data exploitation approaches that
deal with specific requirements according to the targeted stakeholders. This dissertation mainly focuses on people with data table
manipulation and visualisations experiences, to guide them to move
from tabular data to Knowledge Graph exploitation means, education to guide pupils in implicitly exploiting Knowledge Graphs in
knowledge management and information retrieval tasks, and the
cultural heritage community, for their wide interest in publishing
their data according to the Semantic Web technologies.
It results in the second pillar of this dissertation, the effort in
designing and implementing Knowledge Graph exploitation
means. As a general approach, users are guided in querying
Knowledge Graphs by (controlled) natural language interfaces
and organising results as data tables, manually or automatically
perform data manipulation, and exploit results in dynamic artifacts. According to target-oriented requirements, experts in data
table manipulation are provided with a mechanism to author
dynamic and exportable data visualisation components; pupils
are guided to navigate word clouds while implicitly consuming
Knowledge Graphs; cultural heritage lovers are guided to author
virtual reality-based virtual exhibitions or ready-to-use virtual
assistant extensions behaving as virtual guides. The generated
artifacts demonstrate our interest in letting data consumers play
the role of an active user of available data and exploiting them
in concrete, dynamic, reusable and shareable artifacts taking
advantage of (Linked) Open Data. [edited by Author] | it_IT |
