KtoAPP: Compiling Knowledge into Applications

Computer software is playing an increasingly important role in the modern world. In particular, data-centric applications that help collect, store, share and analyze information are of critical importance to society. Unfortunately, the development of software is challenging: it is time-consuming, error-prone and requires highly-skilled personnel. To make things worse, the fast evolution of the world requires developers to capture in computer code increasingly complex knowledge about problem domains, as well as to integrate an increasing number of diverse sources of information. Due to this complexity, reliable software often takes years to develop. Defiantly, today`s world needs technology to rapidly create highly reliable yet very customized applications. For example, when coping with a natural disaster or some other unexpected crisis, we would like to have reliable and situation-specific tools for the responders and the general public to report and analyze the crisis-related information. Another example are one-off clinical trials, in which researchers need reliable tools to collect and analyze data reported by trial participants, all in a very knowledge-intensive and highly regulated domain. To ensure the quality of data, the applications for data reporting should be tailored to the specific trial requirements, and the medical conditions and needs of the trial participants. These examples share a challenging combination of requirements, which current technologies do not allow to fulfill without very costly development efforts: (i) a very complex problem domain, (ii) the need for customization, and (iii) the need for accessibility to users with little or no training. This challenge calls for a major breakthrough in automating software creation, which is now possible due to the recent advances in Knowledge Representation and other areas of Artificial Intelligence. The main goal of this project is to obtain automated- reasoning techniques that exploit formally represented domain knowledge in order to raise the quality of data-centric applications, and to dramatically reduce their development efforts. Specifically, we will develop techniques that exploit domain knowledge given as two components: (1) a knowledge base that captures in a machine-readable language the general knowledge about a given problem domain, and (2) a focus specification that captures the requirements of a specific application. Using automated reasoning, these two components will be compiled into a reliable yet very accessible application, tailored for collecting and analyzing data about the entities described in the focus specification. The objectives of this project will involve work that connects several areas of computer science: Knowledge Representation, Artificial Intelligence, Database Theory and Software Verification.

This project provided a collection of new methods and techniques for building applications that use the graph data model to store information. This form of data representation is increasingly popular, especially in application areas involving complex domain knowledge where traditional relational databases are insufficient. For example, graph-based representations are particularly suitable when we need to manage data incompleteness or inconsistency. In this project we investigated novel methods for automated reasoning in ontologies written in Description Logics (DLs), which in turn paved the way for new, trustworthy reasoning services for ontology-based data management systems. DLs are classic knowledge representation languages that underpin the W3C OWL standard for Web ontologies, and our results are expected to support the development of applications involving complex domain knowledge. For instance, we developed methods for automatically deriving database schemas from ontologies written in DLs. Furthermore, our results on bounded predicates in DLs enabled new query-answering services in ontology-based data access (OBDA). In the context of OBDA, we established techniques that connect ontology-based data access with more traditional data access methods, such as rule-based languages like DATALOG and Answer Set Programming. Another research line focused on DLs that support numeric constraints, which are essential for modeling complex, real-world domains. Our research contributed new reasoning services for solution optimization, building a bridge between DLs and the field of Constraint Programming. In this direction, we developed new algorithms and obtained results on decidability and computational complexity. Our research was also timely in relation to developments in SHACL, a new W3C standard for managing graph-structured data on the Web. We provided significant insights into the foundations of SHACL, establishing connections between SHACL and other well-known languages for knowledge representation and data management. These results have the potential to influence future revisions of the SHACL standard. For example, using ideas from Logic Programming, we proposed new ways to handle self-referential (or "cyclic") SHACL documents. Our work also established strong links between SHACL and other constraint languages for graph-structured data, such as ShEx and PG-Schema. Moreover, we applied ideas from classical artificial intelligence to enable intelligent explanation services in cases where graph-structured data violates a given set of SHACL constraints. Furthermore, we developed tools based on Logic Programming to automatically repair such constraint violations. We also developed methods for automatically checking the quality of program code that operates on graph-structured data: when code passes our checks, it is guaranteed to comply with the specified SHACL constraints. While initially we were mainly using our expertise in DLs to advance SHACL, later some of our results on SHACL were transferred back into the domain of DLs-demonstrating a bidirectional flow of knowledge between two distinct research areas and communities.