Abstract Our paper introduces an innovative automated system designed to extract logical rules using the $\textbf{T}^{\mathsf{\tiny CL}}$ logic from various datasets, with a particular emphasis on tabular data. Our starting point is the CN2 algorithm. Typically employed for classification tasks, we have adapted this algorithm to suit our descriptive objectives. We consider well-known datasets (such as Iris and Zoo) to illustrate our approach. Furthermore, we extend this analysis to a complex dataset, notably the GTZAN musical dataset. We have then tested our system by reclassifying the songs available in the GTZAN database with respect to the newly generated musical genres, obtaining encouraging results. This example showcases the algorithm’s efficacy in generating descriptive rules across different data domains. We discuss the adaptability of the proposed approach across various data types, including images, sounds and various heterogeneous structures.

In recent years, people have been spending more and more time on social media. Within the realm of multimedia contents used by platforms, the quantity of visuals is certainly growing in significance. Interaction data enables to know the users’ favourite images. This information could be exploited to gain a deeper insight into their psychological profile, since the literature on automatic personality recognition suggests that personality traits may correlate with aesthetics. In this paper we explore the use of personal preference on multiple images to predict personality traits of users. Unlike previous works, we propose a model that exploits ResNet50, a Convolutional Neural Network, to automatically extract features from the images in the PsychoFlickr dataset. We then fit five independent linear regressors on these features to detect personality. In order to determine whether using more than one image leads to better results, we train the model multiple times, using one to five images as input, and we compare the performances. Our method seems to outperform the related state-of-the-art works.

Abstract In this paper we present our final solution to the problem of designing an efficient theorem prover for Conditional Logics with the selection function semantics. Conditional Logics recently have received a renewed attention and have found several applications in knowledge representation and artificial intelligence. In order to provide an efficient theorem prover for Conditional Logics, we introduce labelled sequent calculi for the logics characterized by well-established axioms systems including the axiom of strong centering CS, the axiom of conditional identity ID, the axiom of conditional modus ponens MP, as well as the conditional third excluded middle CEM, rejected by Lewis but endorsed by Stalnaker, as well as for the whole cube of extensions. The proposed calculi revise and improve the calculi SeqS introduced in Olivetti et al. (2007, ACM Trans. Comput. Logics, 8). We also present an implementation of these calculi in SWI Prolog, including a graphical interface in Python as well as standard heuristics and refinements that allow us to obtain an efficient theorem prover for the logics under consideration. Moreover, we present some statistics about the performances of the theorem prover, which are promising and significantly better than those of its predecessor CondLean, an implementation of the calculi SeqS.