Semantics in Adaptive and Personalised Systems : Methods, Tools and Applications

The human desire to make the machine always smarter has been the driving force for all the research in the Artificial Intelligence (AI) area. Generally speaking, what makes a system intelligent is the capability of understanding signals coming from the environment and of correctly adapting its behavior accordingly. Such a capability is strictly related to the definition and the design of specific techniques for interpreting messages generated by the users. Some years ago, when we typed on Google the query How tall is the Eiffel Tower?, the system answered with a set of documents, some of them including the information we were seeking for, but without a precise identification of the correct answer. Today, this is no longer the case since intelligent assistants like Siri, Alexa, or Google assistant, and the Google search engine itself, are able to provide the exact answer the user is looking for, that is, in the case of the Eiffel Tower, 300 m. Without any doubt, we can state that the semantics represents the theoretical foundation to implement models and technologies that allow the machines to interpret and understand information provided in natural language. Indeed, thanks to the semantics, it is possible to give meaning to documents, sentences, and questions expressed in natural language and to create a bridge between the information needs of a user and the answers to those needs. Such an intuition is currently implemented in several tools and platforms as search engines, recommender systems, digital assistants, and contributes to their tangible improvement in accuracy and effectiveness we are recently witnessing. We hope this book could become a reference point in the panorama of adaptive and personalized systems exploiting semantics. The book is organized into three main parts. First, we motivate the need to exploit textual content in intelligent information access systems, and then we give an overview of the basic methodologies to process and represent content-based features. Next, we thoroughly describe state-of-the-art methodologies and techniques to enrich textual content representation by introducing semantics. Finally, the last part of the book provides a more practical perspective and discusses several applications that exploit the techniques introduced and described in the previous chapters. We would like to sincerely thank everyone who contributed to this book, and the various people who provided us with comments and suggestions and encouraged usto summarize years of work in a single book. We thank, in particular, Nancy Wade-Jones from Springer, who supported us throughout the editorial process. We are very grateful to the people of the Semantic Web Access and Personalization—SWAP research group,1 who contributed to most of the work cited and described in this book. We would like to thank Marco de Gemmis, who started to investigate how Natural Language Processing techniques could be adopted to devise a new generation of content-based recommender systems, Pierpaolo Basile, who made available his great expertise related to Word Sense Disambiguation and Distributional Semantics Models, which were successfully used in complex recommendation environments, Annalina Caputo, a former member of the research group working on semantic information retrieval methods. We would also like to thank all the other collaborators, Ph.D. students, and research fellows of the SWAP research group, in particular, Leo Iaquinta, Andrea Iovine, Piero Molino, Marco Polignano, Gaetano Rossiello, Lucia Siciliani, and Vincenzo Tamburrano, each giving a specific contribution to the ideas, systems, and research presented in this book.