Faceted search and browsing is an intuitive and powerful way of traversing a structured knowledge base and has been applied with success in many contexts. The GramsciSource project is currently investigating how faceted navigation and Linked Data can be combined to help Humanities scholars in working with digital text corpora. In this short paper we focus on the "Quaderni dal carcere" by Antonio Gramsci, one of the most popular Italian philosophers and politicians, we present our ongoing work and we discuss our approach. This consists of rst building a RDF graph to encode di erent "levels" of knowledge about the texts and then extracting relevant graph paths to be used as navigation facets. We then built a rst prototype exploration tool with a two-fold objective: a) allow non experts to make sense of the extremely fragmented and multidisciplinary text corpus, and b) allow Gramsci scholars to easily select a subset of the corpus of interest as well as possibly discovering new insights or answer research questions.
text with references to speci c Gramsci's Notes from the Quaderni dal Carcere. We automatically processed such citations using regular expressions and producing RDF triples representing such connections, expressing relations among single notes and a number of dictionary topics they are relevant to.
Entity extraction and linking. Several approaches and tools for
extracting and disambiguating relevant entities mentioned in a text appeared in the
last years. Among them, DataTXT4 is, to our knowledge, one of the best tools
supporting Italian language. DataTXT derives from previous academic research
[
Scholars annotations. Pundit5 is a semantic web tool that enables users to produce machine readable data in the form of RDF by annotating web pages. Annotations from a single scholar are collected in so called "notebooks", which can be private or public. For the purpose of our proof of concept we created a set of sample annotations by manually linking texts to DBpedia and Freebase entities. At data representation level, such annotations are equivalent to those produced by DataTXT, once imported they are naturally captured by the facet queries (discussed in the next section).
Linked Data and facets to explore text corpora in the Humanities
Existing approaches to identify relevant facets to browse a RDF graph based on
quantitative measures such as predicate frequency, balance and objects
cardinality [
Our simple approach is to derive facets from SPARQL queries of the form:
select distinct ?url ?facet ?value where { CUSTOM_QUERY }
Where ?url is a resource of interest (notes in our case), ?facet is a facet name
and ?value is a possible value of such a facet. Such a simple approach is also
quite exible and allows, for example, to easily turn all the datatype properties
of a resource to facets, e.g with the following query:
select distinct ?uri ?facet ?value where {
?uri rdf:type gramsci:Note. ?uri ?facet ?facet.}
Deriving facets from SPARQL queries is an approach already explored in
literature [
To enable navigation of the corpus along the di erent "dimensions", we implemented a faceted browser based on Apache Solr6. Solr, along with its AjaxSolr7 frontend provides a relatively easy way to build a performant faceted browser on top of Lucene. We built the solr index by running the SPARQL queries (described in the previous section) and using results associated to the ?uri variable as document ID, ?facet as index eld and ?value as eld values. The prototype is available at http://purl.org/gramscisource/quaderni.
7 http://github.com/evolvingweb/ajax-solr/wiki
Some usage patterns have been individuated by scholars involved in the project: a) Using the Dictionary facet to intersect two or more topics from the vocabulary. This is a simple but useful "advanced search" feature; b) Choose one or more Dictionary topics (e.g. Storia), then use the facets on the right (DBpedia entities) to provide additional context (e.g. Hegel, Croce and Plechanov are the main persons related to History, "Teoria e storia della storiogra a" and "Misre de la philosophie" are two related books, etc.); c) Start from a full text search or from a DBPedia entity (e.g. "Conte di Montecristo") and discover related topics.
In this short paper we discussed preliminary results in leveraging Linked Data in
the GramsciSource project and we presented a proof of concept prototype.
Feedback from Humanities scholars involved in the project (and in related projects,
such as DM2E8) was positive and encouraged us to move further. End user
evaluation will be run in the next months. We are currently evaluating automatic
methods to derive entities and facets (e.g. based on language analysis tool such
as [
This work is supported by the GramsciSource project funded by the Italian Ministry of Education under the FIRB action.