Special issue on innovative techniques and applications of artificial intelligence

Abstract

This special issue of Expert Systems: The Journal of Knowledge Engineering includes the expanded versions of the best papers presented at the conference AI-2011, the 31st Specialist Group on Artificial Intelligence (SGAI) International Conference on Artificial Intelligence, which was held in Cambridge, England, from 13th to 15th December 2011. The original conference papers were selected by the AI-2011 programme committees as the top three best papers for each of the two streams: the technical stream and the application stream. The expanded versions of the papers were invited to be submitted to and peer reviewed independently for the inclusion in this journal special issue. The AI-2011 conference was organized by SGAI (http://www.bcs-sgai.org). SGAI is a specialist group of the British Computer Society and a member of ECCAI, the European Coordinating Committee on Artificial Intelligence. The British Computer Society SGAI AI-2011 is part of the series of conferences that has run annually without a break since 1981. All conference papers are reviewed by an international panel of expert referees. This paper presents a multi-level speaker verification system that uses 64 discrete Fourier transform spectrum components as input feature vectors. A speech activity detection technique is used as a pre-processing stage to identify vowel phoneme boundaries within a speech sample. A modified self-organizing map (SOM) is then used to filter the speech data using cluster information extracted from three vowels for a claimed speaker. This SOM filtering stage also provides coarse speaker verification. Finally, a second speaker verification level of three multilayer perceptron networks classifies the filtered frames provided by the SOMs. These multilayer perceptrons work as fine-grained vowel-based speaker verifiers. The proposed verification algorithm shows a performance of 94.54% when evaluated by using 50 speakers from the Centre for Spoken Language Understanding (CSLU2002) speaker verification database. In addition, it is shown that the novel discrete Fourier transform spectrum-based linear correlation pre-processing technique, presented here, provides the system with greater robustness against changes in speech volume levels when compared with an equivalent energy frame analysis. DenGraph-HO is an extension of the density-based graph clustering algorithm DenGraph. It is able to detect dense groups of nodes in a given graph and to produce a hierarchy of clusters that can be efficiently computed. The generated hierarchy can be used to investigate the structure and the characteristics of social networks. Each hierarchy level provides a different level of detail and can be used as the basis for interactive visual social network analysis. After a short introduction of the original DenGraph algorithm, we present DenGraph-HO and its top-down and bottom-up approaches. We describe the data structures and memory requirements and analyse the run-time complexity. Finally, we apply the DenGraph-HO algorithm to real-world datasets obtained from the online music platform Last.fm and from the former US company Enron. Daniel Neagu is a Professor of Computing and the leader of the Artificial Intelligence Research (AIRe) Group at the University of Bradford. His expertise covers data governance, data mining, knowledge representation and information processing topics with applications in healthcare, vulnerability in online social networks, product safety. The main theme throughout his research work (funded by EC and national institutions) is to develop models of multidisciplinary information systems by the fusion of experts' knowledge and digital information. Daniel has published over 100 papers in peer-reviewed journals, conferences and book chapters. He is a committee member for the BCS Specialist Group on Artificial Intelligence, Deputy Chair of the BCS SGAI Conference in AI, Fellow of the HEA and member of BCS, ACM and IEEE CS.