This is the preface for the joint workshop between KaRS and ComplexRec: two workshops co-located with the 15th ACM RecSys 2021 conference. In this volume, we include the contributions presented at the Joint KaRS & ComplexRec Workshop, co-located with the 15ℎ edition of the ACM Conference on Recommender Systems (RecSys) in Amsterdam.
This joint workshop adopted a hybrid format aligned with the goal of this year’s main conference – congregating to continue to build community around recommeder systems research and development. In this joint workshop, we merged the main objectives envisioned for the 3 Edition of the KaRS Workshop and the 5ℎ edition of the Workshop on Recommendation in Complex Environments: • Providing an interactive venue for discussing approaches to recommendation in complex envi3rd Edition of Knowledge-aware and Conversational Recommender Systems (KaRS) & 5th Edition of Recommendation in Complex Environments (ComplexRec) Joint Workshop co-located with the 15th ACM Conference on Recommender Systems (RecSys 2021) © 2021 Copyright for this paper by its authors. Use permitted under Creative CPWrEooUrckReshdoinpgs IhStpN:/c1e6u1r3-w-0s.o7r3g CCoEmmUoRns LWiceonsrekAstthribouptionP4r.0oIncteerenadtiionnagl s(CC(CBYE4U.0)R.-WS.org) ronments that have no simple one-size-fits-all solution. In particular, we envisioned deepening the community understanding on complex inputs– e.g., active user inputs (interaction), implicit user inputs (task, context, preferences), item inputs (features or attributes), domain inputs (eligibility, availability)–and complex outputs–e.g., package recommendation, composite items, interface complexity, constraint-based recommendation. • Providing a meeting forum for stimulating and disseminating research in Knowledge-aware and Conversational Recommender Systems, where researchers can network and discuss their research results in an informal way. In particular, we aimed to expand community understanding on knowledge-aware recommenders–from models and feature engineering issues to beyond accuracy recommendation quality with a particular focus on real-world applications–and conversational recommenders– from the design of a conversational agent and its interface to the user modelling problems and evaluation issues.
Overall, we accepted 17 contributions: 11 long papers, 3 short papers, and 3 position papers. Each presentation was peer-reviewed by at least 3 program committee (PC) members. The presentations of the accepted contributions, along with the two keynote addressed by Edward C. Malthouse (during the virtual component) and Gerard de Melo (during the in person workshop component), sparked interactions among attendees and fostered ideas to continue to advance research focused around the topics of the joint workshop.
The 3rd Knowledge-aware and Conversational Recommender Systems (KaRS) Workshop focuses on all aspects related to the exploitation of external and explicit knowledge sources to feed and build a recommendation engine, and on the adoption of interactions based on the conver
As KaRS & ComplexRec co-organizer, we want to sational paradigm. The aim is to go beyond the traditional thank the RecSys 2021 workshop co-chairs, for their sup- accuracy goal and to start a new generation of algorithms port regarding hybrid workshop organization. Last, but and approaches with the help of the methodological dinot least, we would like to thank all authors and presen- versity embodied in fields such as Human–Computer ters, as well as the members of the program committee Interaction, Conversational Recommender Systems, Sewho selflessly shared their time and expertise in provid- mantic Web, and Knowledge Graphs. Consequently the ing feedback to workshop authors. Finally, the workshop focus lies on works improving the user experience and proceedings shall be submitted to CEUR-WS.org for on- following goals such as user engagement and satisfaction line publication. or customer value.
The aim of this third edition of KaRS is to bring together researchers and practitioners around the topics of designing and evaluating novel approaches for recommender systems in order to:
In the last few years, a renewed interest of the research community on conversational recommender systems (CRSs) is emerging. This is probably due to the great difusion of Digital Assistants (DAs) such as Amazon Alexa, Siri, or Google Assistant that are revolutionizing the way users interact with machines. DAs allow users to execute a wide range of actions through an interaction mostly based on natural language messages. However, although DAs are able to complete tasks such as sending texts, making phone calls, or playing songs, they are still at an early stage on ofering recommendation capabilities by using the conversational paradigm.
In addition, we have been witnessing the advent of more and more precise and powerful recommendation algorithms and techniques able to efectively assess users’ tastes and predict information that would probably be of interest to them. Most of these approaches rely on the collaborative paradigm (often exploiting machine learning techniques) and do not take into account the huge amount of knowledge, both structured and non-structured ones, describing the domain of interest of the recommendation engine. Although very efective in predicting relevant items, collaborative approaches miss some very interesting features that go beyond the accuracy of results and move in the direction of providing novel and diverse results as well as generating an explanation for the recommended items. Furthermore, this side information becomes crucial when a conversational interaction is implemented, in particular for the preference elicitation, explanation, and critiquing steps.
• Share research and techniques, including new design technologies and evaluation methodologies • Identify next key challenges in the area • Identify emerging topics in the field
The program of the half-day workshop (part virtual, part in-person) consists of: • An invited keynote by Professor Gerard de Melo from the Hasso Plattner Institute for Digital Engineering and the University of Potsdam, Germany. • The presentation of the selected research papers,
All workshop material including schedule and news
We thank the members of the PC for their thorough reviews and their detailed feedback they gave to the authors. The PC consisted of the following international experts. • Vito Walter Anelli, POLITECNICO DI BARI • Azzurra Ragone, EY BUSINESS AND TECHNOL
OGY SOLUTIONS • Giovanni Semeraro, UNIVERSITY OF BARI • Nourah Alrossais, UNIVERSITY OF YORK • Nicola Ferro, UNIVERSITY OF PADOVA • Yashar Deldjoo, POLITECNICO DI BARI • Claudio Pomo, POLITECNICO DI BARI • Antonio Ferrara, POLITECNICO DI BARI
in Complex Environments
• Paolo Rosso, UNIVERSITAT POLITÈCNICA DE
VALÈNCIA • Andrea Iovine, UNIVERSITÀ DEGLI STUDI DI
BARI ALDO MORO • Fedelucio Narducci, POLITECNICO DI BARI • Adir Solomon, BEN-GURION UNIVERSITY • Maurizio Ferrari Dacrema, POLITECNICO DI MI
LANO • Diego Antognini, ECOLE POLYTECHNIQUE
FÉDÉRALE DE LAUSANNE • Marco Polignano, UNIVERSITÀ DEGLI STUDI
DI BARI ALDO MORO • Tommaso Di Noia, POLYTECHNIC UNIVERSITY
OF BARI • Iván Cantador, UNIVERSIDAD AUTÓNOMA DE
MADRID • Marco de Gemmis, UNIVERSITY OF BARI ALDO
MORO • Rafaele Perego, ISTI-CNR • Claudio Gennaro, ISTI-CNR • Gianmaria Silvello, UNIVERSITY OF PADUA
During the past decade, recommender systems have rapidly become an indispensable element of websites, apps, and other platforms that seek to provide personalized interactions to their users. As recommendation technologies are applied to an ever-growing array of non-standard problems and scenarios, researchers and practitioners are also increasingly faced with challenges of dealing with greater variety and complexity in the inputs to those recommender systems. For example, • Cataldo Musto, DIPARTIMENTO DI INFORMAT- there has been more reliance on fine-grained user
ICA - UNIVERSITY OF BARI signals as inputs rather than simple ratings or likes. • Davide Di Ruscio, UNIVERSITÀ DEGLI STUDI Applications require more complex domain-specific
DELL’AQUILA constraints on inputs to the recommender systems. • Nicola Tonellotto, UNIVERSITY OF PISA Likewise, the outputs of recommender systems are • Pierpaolo Basile, UNIVERSITY OF BARI moving towards more complex composite items, such as • Alejandro Bellogin, UNIVERSIDAD AU- package or sequence recommendations. This increasing
TONOMA DE MADRID complexity requires smarter recommender algorithms • Chiara Renso, ISTI-CNR, PISA, ITALY that can deal with this diversity in inputs and outputs. • Pablo Sánchez, UNIVERSIDAD AUTÓNOMA DE
MADRID For the past four years, the ComplexRec workshop • Benjamin Heitmann, RWTH AACHEN UNIVER- series has ofered an interactive venue for discussing
SITY approaches to recommendation in complex scenarios • Maria Maistro, UNIVERSITY OF COPENHAGEN that have no simple one-size-fits-all solution. For the • Olga Marino, UNIVERSIDAD DE LOS ANDES iffth edition of ComplexRec we have narrowed the focus • Francesco M. Donini, UNIVERSITA’ DELLA TUS- of the workshop and contributions to the workshop
CIA about topics related to one of the two main themes on • Dietmar Jannach, UNIVERSITY OF KLAGEN- complex recommendation: complex inputs and complex
FURT outputs. • Cristina Gena, UNIVERSITY OF TORINO • Giorgio Maria Di Nunzio, UNIVERSITY OF
PADUA • Federica Cena, UNIVERSITY OF TORINO
An important source of complexity comes from the vari- The ComplexRec 2021 organizers would like to thank the ous types of inputs to the system beyond users and items, members of the program committee for their time and such as features, queries and constraints. There are active efort to provide timely and constructive reviews of the user inputs (interaction), implicit user inputs (task, con- submitted papers. text, preferences), item inputs (features or attributes) and domain inputs (eligibility, availability). In group-based recommendation, the user input can be a combination of inputs for multiple individual users as well as group aspects such as the composition of the group and how well they know each other. An additional challenge is providing users with ways to have control over the inputs. For instance by selecting and weighting or ranking user and item features, providing interactive queries to steer the recommendation, or deal with longer narrative statements that require natural language understanding.
• Panos Adamopoulos, EMORY UNIVERSITY • Ludovico Boratto, EURECAT • Christine Bauer, UNIVERSITY OF UTRECHT • Alejandro Bellogin, UNIVERSIDAD
AUTÓNOMA DE MADRID • Iván Cantador, UNIVERSIDAD AUTÓNOMA DE
MADRID • Tommaso Di Noia, POLITECNICO DI BARI • Mehdi Elahi, UNIVERSITY OF BERGEN • Fabio Gasparetti, ROMA TRE UNIVERSITY • Pasquale Lops, UNIVERSITY OF BARI “ALDO
MORO” • Mirko Marras, EPFL • Cataldo Musto, UNIVERSITY OF BARI “ALDO
MORO” • Fedelucio Narducci, UNIVERSITY OF BARI • Markus Schedl, JOHANNES KEPLER UNIVER
SITY • Peter Dolog, AALBORG UNIVERSITY • Cristina Gena, UNIVERSITA’ DEGLI STUDI DI
TORINO • Hanna Schäfer, UNIVERSITÄT KONSTANZ • Marco De Gemmis, UNIVERSITY OF BARI • Bei Yu, SYRACUSE UNIVERSITY