Defining conceptual representations along with their associated reasoning procedures required and still requires truly interdisciplinary efforts, involving Cognitive Psychology, Philosophy, Neuroscience and, of course, Computer Science. The ever-growing number of applications of semantic technologies demands for further investigation on concepts’ meaning, at the interface between lexicon and semantics: this intersection is where we target our research efforts in building usable and wide-coverage resources for devising NLP applications.
In order to provide artificial systems with human-level competence in understanding text documents, our lexical resources are aimed at merging the lexicographic richness and precision (which is proper to WordNet) with wide coverage traits (featuring, e.g., BabelNet) and common-sense knowledge (such as that available in ConceptNet).
Such resources are at the heart of the following applications: conceptual siamilarity and text similarity judgements; keywords extraction, question answering, conceptual categorization, computational explanation.
In this line of research we focus on conceptual categorization: our hybrid reasoning system (Dual PECCS) puts together common-sense knowledge and formal ontologies, and it relies on the dual process theory of the mind.
The Dual PECCS system has been integrated into a number of cognitive architectures (such as ACT-R, SOAR, CLARION, LIDA), thus proving to be compatible with different assumptions on the architecture of the mind.
This line of research, based on joint work with Dr. Roberto Esposito, has produced a novel algorithm, CarpeDiem, that significantly improves on the time complexity of Viterbi algorithm, preserving the optimality of the result. This fact has consequences on Machine Learning systems that use Viterbi algorithm during learning or classification, and in other application domains, as well.
The problem of tonal harmony analysis can be cast to a classification problem, the problem of associating a chord label to each time point in musical flow. The analysis problem can be solved via the HMPerceptron algorithm. The main concepts used in music analysis by human analysists are introduced in an automatic analysis environment. This line of research investigates their cognitive motivation, such that the resulting system is both effective on a computational viewpoint, and justifiable on musical and psychological accounts.