Claudia d’Amato

The Data Mining OPtimization Ontology

The Data Mining OPtimization Ontology (DMOP) has been developed to support informed decision-making at various choice points of the data mining process. The ontology can be used by data miners and deployed in ontology-driven information systems. The primary purpose for which DMOP has been developed is the automation of algorithm and model selection through semantic meta-mining that makes use of an ontology-based meta-analysis of complete data mining processes in view of extracting patterns associated with mining performance. To this end, DMOP contains detailed descriptions of data mining tasks (e.g., learning, feature selection), data, algorithms, hypotheses such as mined models or patterns, and workflows. A development methodology was used for DMOP, including items such as competency questions and foundational ontology reuse. Several non-trivial modeling problems were encountered and due to the complexity of the data mining details, the ontology requires the use of the OWL 2 DL profile. DMOP was successfully evaluated for semantic meta-mining and used in constructing the Intelligent Discovery Assistant, deployed at the popular data mining environment RapidMiner.

Evolutionary Conceptual Clustering Based on Induced Pseudo-Metrics

We present a method based on clustering techniques to detect possible/probable novel concepts or concept drift in a Description Logics knowledge base. The method exploits a semi-distance measure defined for individuals, that is based on a finite number of dimensions corresponding to a committee of discriminating features (concept descriptions). A maximally discriminating group of features is obtained with a randomized optimization method. In the algorithm, the possible clusterings are represented as medoids (w.r.t. the given metric) of variable length. The number of clusters is not required as a parameter, the method is able to find an optimal choice by means of evolutionary operators and a proper fitness function. An experimentation proves the feasibility of our method and its effectiveness in terms of clustering validity indices. With a supervised learning phase, each cluster can be assigned with a refined or newly constructed intensional definition expressed in the adopted language.

Towards Evidence-Based Terminological Decision Trees

We propose a method that combines terminological decision trees and the Dempster-Shafer Theory, to support tasks like ontology completion. The goal is to build a predictive model that can cope with the epistemological uncertainty due to the Open World Assumption when reasoning with Web ontologies. With such models not only one can predict new (non derivable) assertions for completing the ontology but by assessing the quality of the induced axioms.

Inductive Classification of Semantically Annotated Resources through Reduced Coulomb Energy Networks

The tasks of resource classification and retrieval from knowledge bases in the Semantic Web are the basis for a lot of important applications. In order to overcome the limitations of purely deductive approaches to deal with these tasks, inductive (instance-based) methods have been introduced as efficient and noise-tolerant alternatives. In this paper we propose an original method based on a non-parametric learning scheme: the Reduced Coulomb Energy (RCE) Network. The method requires a limited training effort but it turns out to be very effective during the classification phase. Casting retrieval as the problem of assessing the classmembership of individuals w.r.t. the query concepts, we propose an extension of a classification algorithm using RCE networks based on an entropic similarity measure for OWL. Experimentally we show that the performance of the resulting inductive classifier is comparable with the one of a standard reasoner and often more efficient than with other inductive approaches. Moreover, we show that new knowledge (not logically derivable) is induced and the likelihood of the answers may be provided.

Assertion Prediction with Ontologies through Evidence Combination

Following previous works on inductive methods for ABox reasoning, we propose an alternative method for predicting assertions based on the available evidence and the analogical criterion. Once neighbors of a test individual are selected through some distance measures, a combination rule descending from the Dempster-Shafer theory can join together the evidence provided by the various neighbor individuals in order to predict unknown values in a learning problem. We show how to exploit the procedure in the problems of determining unknown class- and role-memberships or fillers for datatype properties which may be the basis for many further ABox inductive reasoning algorithms. This work presents also an empirical evaluation of the method on real ontologies.

Tackling the Class-Imbalance Learning Problem in Semantic Web Knowledge Bases

In the Semantic Web context, procedures for deciding the class-membership of an individual to a target concept in a knowledge base are generally based on automated reasoning. However, frequent cases of incompleteness/inconsistency due to distributed, heterogeneous nature and the Web-scale dimension of the knowledge bases. It has been shown that resorting to models induced from the data may offer comparably effective and efficient solutions for these cases, although skewness in the instance distribution may affect the quality of such models. This is known as class-imbalance problem. We propose a machine learning approach, based on the induction of Terminological Random Forests, that is an extension of the notion of Random Forest to cope with this problem in case of knowledge bases expressed through the standard Web ontology languages. Experimentally we show the feasibility of our approach and its effectiveness w.r.t. related methods, especially with imbalanced datasets.

Transductive Inference for Class-Membership Propagation in Web Ontologies

The increasing availability of structured machine-processable knowledge in the context of the Semantic Web, allows for inductive methods to back and complement purely deductive reasoning in tasks where the latter may fall short. This work proposes a new method for similarity-based class-membership prediction in this context. The underlying idea is the propagation of class-membership information among similar individuals. The resulting method is essentially non-parametric and it is characterized by interesting complexity properties, that make it a candidate for the application of transductive inference to large-scale contexts. We also show an empirical evaluation of the method with respect to other approaches based on inductive inference in the related literature.

Comparing Rule Evaluation Metrics for the Evolutionary Discovery of Multi-Relational Association Rules in the Semantic Web

We carry out a comparison of popular asymmetric metrics, originally proposed for scoring association rules, as building blocks for a fitness function for evolutionary inductive programming. In particular, we use them to score candidate multi-relational association rules in an evolutionary approach to the enrichment of populated knowledge bases in the context of the Semantic Web. The evolutionary algorithm searches for hidden knowledge patterns, in the form of SWRL rules, in assertional data, while exploiting the deductive capabilities of ontologies.

Approximating Numeric Role Fillers via Predictive Clustering Trees for Knowledge Base Enrichment in the Web of Data

In the context of the Web of Data, plenty of properties may be used for linking resources to other resources but also to literals that specify their attributes. However the scale and inherent nature of the setting is also characterized by a large amount of missing and incorrect information. To tackle these problems, learning models and rules for predicting unknown values of numeric features can be used for approximating the values and enriching the schema of a knowledge base yielding an increase of the expressiveness, e.g. by eliciting SWRL rules. In this work, we tackle the problem of predicting unknown values and deriving rules concerning numeric features expressed as datatype properties. The task can be cast as a regression problem for which suitable solutions have been devised, for instance, in the related context of RDBs. To this purpose, we adapted learning predictive clustering trees for solving multi-target regression problems in the context of knowledge bases of the Web of Data. The approach has been experimentally evaluated showing interesting results.

Discovering Similarity and Dissimilarity Relations for Knowledge Propagation in Web Ontologies

We focus on the problem of predicting missing class memberships and property assertions in Web Ontologies. We start from the assumption that related entities influence each other, and they may be either similar or dissimilar with respect to a given set of properties: the former case is referred to as homophily, and the latter as heterophily. We present an efficient method for predicting missing class and property assertions for a set of individuals within an ontology by: identifying relations that are likely to encode influence relations between individuals (learning phase) and Leveraging such relations for propagating property information across related entities (inference phase). We show that the complexity of both inference and learning is nearly linear in the number of edges in the influence graph, and we provide an empirical evaluation of the proposed method.