Integrating Defeasible Argumentation and Machine Learning Techniques

Abstract

The field of machine learning (ML) is concerned with the question of how to construct algorithms that automatically improve with experience. In recent years many successful ML applications have been developed, such as datamining programs, information-filtering systems, etc. Although ML algorithms allow the detection and extraction of interesting patterns of data for several kinds of problems, most of these algorithms are based on quantitative reasoning, as they rely on training data in order to infer so-called target functions. In the last years defeasible argumentation has proven to be a sound setting to formalize common-sense qualitative reasoning. This approach can be combined with other inference techniques, such as those provided by machine learning theory. In this paper we outline different alternatives for combining defeasible argumentation and machine learning techniques. We suggest how different aspects of a generic argument-based framework can be integrated with other ML-based approaches.