In this paper we present the PRiSMHA project, whose main goal is to demonstrate that a rich semantic representation of the content of historical documents is useful - since it can significantly improve the access to archival resources - and sustainable - thanks to a crowdsourcing approach. This goal poses interesting research challenges, both for the semantic model definition and the user interaction. Such challenges range from the dialog between computer scientists and historians, to the design of an effective ontology-driven user interface; from the strategies to ensure the quality of the semantic metadata produced, to the application of Information Extraction techniques to support user annotation.
For more than a decade, archive specialists have been aware of the huge gap between the resources contained in archives and the communities of users interested in them: "Organizations, even small, possess information worth millions, but cannot get it to the right people" (Making Information Work: the Dublin Core Way, 2007, DCMI). To address this gap, research in metadata has proposed flexible and richer schemata (such as Dublin Core), but only the advent of semantic annotation has brought about a change of paradigm in the description of resources. However, we think that in order to be effective, semantic annotation has to rely on a rich semantic model, enabling metadata to provide a rich description of the resource content. In fact, although the relative simplicity of currently available semantic models benefits processing, interoperability and sharing, it often prevents metadata to be actually useful.
A major flaw, however, still affects approaches based on rich semantic models, namely the difficulty of collecting the semantic annotations. The crowdsourcing model can contribute to overcome this obstacle by leveraging the work of users who do not share timing and location, through the mediation of specialized annotation platforms.
The PRiSMHA (Providing Rich Semantic Metadata for Historical Archives)
project (lasting 2017-2019) was born in this perspective. It is a national project, funded
by Compagnia di San Paolo and Università di Torino, involving the Computer Science
and the Historical Studies Departments of the same university, and based on a close
collaboration with the Polo del ‘900 (www.polodel900.it), a cultural center
headquartered in Torino (Italy) and co-funded by Compagnia di San Paolo, Comune di
Torino and Regione Piemonte. It involves nineteen institutions engaged in research and
cultural initiatives about the XX Century social, economic and political issues in
Piemonte, and hosts a very rich set of documentary repositories. In particular,
PRiSMHA will mainly rely on resources from the archives and library of the Fond. Ist.
Piemontese A. Gramsci (www.gramscitorino.it), which with its 2.5 km of documents
represents the major “contributor” (25%) of the Polo del ‘900 archives [
The main objective of PRiSMHA is to demonstrate how a rich semantic representation of the content of historical archival resources can both: (a) provide a significant enhancement in the possibilities of exploitation of archival resources; (b) be sustainable, as far as the overload imposed by knowledge acquisition (and the consequent bottleneck in the process) is concerned.
By "a rich semantic representation" we mean a formal representation based on established standards (e.g., OWL, RDF, Linked Open Data), grounded on well-founded reference ontologies (e.g., DOLCE; see Section 2), and supporting the detailed representation of historical events, including their location, temporal information, how the involved entities participate in them, and relations among events themselves. Such a representation would enable users, for example, to query historical archives about an event (e.g., the general strike on April 18, 1945 in Turin) and getting a set of references to pictures, texts, letters and historical accounts of the event, together with links to people and organizations involved in the strike; or to query archives about a certain typology of events (e.g., strikes) and getting also events belonging to related typologies (e.g., protests, marches, demonstrations).
The sustainability of the approach will rely on a crowdsourcing collaborative model, where experts and trusted users participate in the semantic annotation process. The interaction model will be driven by the underlying ontologies and supported with suggestions provided by automatic Information Extraction techniques. The project aims at demonstrating that the proposed crowdsourcing model overcomes the knowledge acquisition bottleneck thus making the overall approach sustainable.
In order to build the semantic model, we will couple the study of existing ontologies (see Section 2), with the analysis of published material, mainly books containing memoirs that narrate events related to the same domain (period and place) the selected archival resources refer to. Such kind of texts can also be objects of user annotation in the crowdsourcing platform, since they typically offer detailed information about the events (e.g., they narrate a strike) and thus they represent a valuable source for building the semantic representations that can be connected to the archival resources (e.g., pictures illustrating the same strike).
The PRiSMHA semantic model will be centered around the notion of (historical) event,
which typically plays a major role in the historical domain [
The other relevant field whose state-of-the-art should be considered is represented
by crowdsourcing projects in the Cultural Heritage domain. Crowdsourcing is
becoming a fundamental approach for providing rich meta-data describing those
resources that cannot be automatically processed (e.g., old pictures, handwritten
documents), and many international Institution are starting crowdsourcing projects
e.g., The Library of Congress, The British National Archives, New York Public Library,
among many others [
PRiSMHA has been conceived as gathering cultural needs, objectives and technological efforts from different disciplines and approaches, in a multidisciplinary perspective. The research issues it will face include the definition of: (a) best practices, grounded in consolidated approaches of the historical research, aimed at supporting the specification of both the user requirements and the semantic model needed in the project; (b) a semantic model (ontology) tailored to the domain of historical events narrated by archival resources; (c) an innovative interaction model, grounded in collaborative approaches and Human-Computer Interaction best practices, in order to enable (selected) users to provide high-quality semantic metadata. These issues pose interesting challenges, which are presented in the following.
Definition of a rich and interoperable semantic model, emerging from the
dialog with historians. The definition of the semantic model should follow
the best practices of the applied ontology community [
Publication of good quality datasets. Semantically enriched metadata produced by PRiSMHA should be published on the Linked Open Data (LOD) cloud, in order to make them largely accessible and to promote their innovative use (e.g., in applications in the tourism or education domain).
Definition of user requirements of the crowdsourcing platform, based on the dialog with historians. Again, a fruitful interaction between computer scientists and historians plays a major role in the identification of target users and in the definition of requirements of the crowdsourcing platform. The latter include, for example, the specification of the roles involved in the annotation processes, and the workflow required to collect and align the annotations in order to reach an acceptable agreement.
crowdsourcing platform will be driven by the underlying ontology. Given the claim that such an ontology should be semantically rich, the design and implementation of the UI requires a careful study to merge the requirements imposed by the ontology with those emerging from the platform users. Quality of the semantic metadata produced. Although we expect users of the PRiSMHA crowdsourcing platform to be selected and trusted, specific mechanisms should be designed and implemented to guarantee the quality of the semantic metadata produced. First steps in this direction are the selection of a pilot corpus and a set of data for testing the suitability of the semantic model and the design of the annotation workflow. Moreover, users themselves could be involved in the metadata quality assessment process.
important support to the user annotation activity could be provided by
automatic techniques for Information Extraction, when full-text is available.
As widely recognized [
In this paper we presented the PRiSMHA project, by focusing on the main issues we are going to face, concerning the semantic model definition and the user interaction. Such challenges will be faced having in mind the goal of the project, i.e. demonstrating that a rich semantic representation of the content of historical documents can significantly improve the access to archival resources and can be sustainable, thanks to a crowdsourcing approach. We foresee that the results of the project will highlight the tight connection between the quality of the underlying ontology and the production of usable and useful semantic metadata, through user interfaces able to mirror the semantic model in an effective and user-friendly way. Moreover, the inclusion of an inference engine in the system architecture could be taken into account, in order to fully leverage the semantic metadata.