This is the preface for the Proceedings of the Fourth Workshop on Knowledge-Aware and Conversational Recommender Systems (KaRS 2022), co-located with the 16th ACM RecSys 2022 conference.
to advance research focused around the topics of the workshop.
the Fourth Workshop on Knowledge-aware and
Conversational Recommender Systems (KaRS), co-located with 2. Background and Goals
the 16th ACM Conference on Recommender Systems
(RecSys 2022) [
This workshop provides a meeting forum for stimu- identify latent connections among users and items [9, 10]. lating and disseminating research in Knowledge-aware Despite their excellent results in terms of accuracy in and Conversational Recommender Systems, where re- recommending new items, such approaches very ofsearchers can network and discuss their research results ten miss a fundamental actor in the loop: the endin an informal way. In particular, we aimed to expand user. For this reason, current research is focusing on community understanding on knowledge-aware recom- new challenges such as privacy [11], emotion awaremenders — from models and feature engineering issues ness [12], and new paradigms such as federated learnto beyond accuracy recommendation quality with a par- ing [13, 14]. The exploitation of the knowledge about ticular focus on real-world applications — and conver- the domain of interest of a catalog via automated reasational recommenders — from the design of a conver- soning as well as critiquing approaches are very comsational agent and its interface to the user modelling mon in the normal behavior of a human user, but they problems and evaluation issues. are not well codified in recommendation engine behav
Overall, we accepted 12 contributions: 7 long papers, iors. Knowledge-based approaches began to appear two
and 5 short papers. Each presentation was peer-reviewed decades ago [15, 16, 17, 18, 19]. Nonetheless, they
beby at least 3 program committee (PC) members. The pre- came more widely used with the advent of the Linking
sentations of the accepted contributions, along with the Open Data1 initiative when a huge number of
knowledgekeynote addressed by Xin Luna Dong, sparked interac- graphs started being released and were made freely
availtions among attendees and fostered ideas to continue able. These include encyclopedic datasets such as
DBpedia2 and Wikidata3, where semantics-aware
information is available on diferent knowledge domains and
applications [20]. The exploitation of such datasets to- of KaRS [
Furthermore, this side information associated with items becomes crucial when the interaction requires con- Topics of interests include, but are not limited to: tent features. This is the case of Conversational Recommender Systems (CRSs) [39]. CRSs are characterized by a multi-turn dialogue between the user and the system [40] and are exploited in several domains [41]. Note that “conversational” as defined here is not restricted to CRSs that conduct dialogues in natural language. A CRS might converse in natural language, but it may allow more constrained modes of user interaction as well [42]. This kind of interaction introduces new challenges, since it blurs the diference between recommendation and retrieval. A CRS ought to be able to exploit both short- and long-term preferences, for example. Furthermore, a CRS should be able to adapt its behaviour in a timely manner when user feedback is provided. These are just some peculiarities of this kind of interaction. As we can imagine, another sensitive issue is the evaluation of CRSs [43], since also in this case we need to go beyond simple accuracy metrics. The limited availability of datasets is an additional obstacle to the evaluation of these systems [44]. While research and development into CRSs has never gone away, it has certainly been less prominent for a while. Only recently has the literature on this topic been growing again quite notably [40]. • Models and Feature Engineering: Data models based on structured knowledge sources (e.g., Linked Open Data, Wikidata, BabelNet, etc.), Semantics-aware approaches exploiting the analysis of textual sources (e.g., Wikipedia, Social Web, etc.), Knowledge-aware user modeling, Methodological aspects (evaluation protocols, metrics, and datasets), Logic-based modeling of a recommendation process, Knowledge Representation and Automated Reasoning for recommendation engines, Deep learning methods to model semantic features • Beyond-Accuracy Recommendation Quality:
Using knowledge bases and knowledge graphs to increase recommendation quality (e.g., in terms of novelty, diversity, serendipity, or explainability), Explainable Recommender Systems, Knowledgeaware explanations (compliant with the General
Data Protection Regulation) • Online Studies: Knowledge sources for cross-lingual recommendations, Applications of knowledge-aware recommenders (e.g., music or news recommendation, of-mainstream application areas), User studies (e.g., on the user’s perception of knowledge-based recommendations), field studies • Design of a Conversational Agent: Design and implementation methodologies, Dialogue management (end-to-end, dialogue-state-tracker models), UX design, Dialogue protocol design • User Modeling and Interfaces: Critiquing and user’s feedback exploitation, Short- and Longterm user profiling and modeling, Preference elicitation, Natural language, multimodal, and voicebased interfaces, Next-question problem • Methodological and Theoretical aspects:
Evaluation and metrics, Datasets, Theoretical aspects of conversational recommender systems 2.1. Objectives The Fourth 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 conversational paradigm. The aim is to go beyond the traditional accuracy goal [8] and to start a new generation of algorithms and approaches with the help of the methodological diversity embodied in fields such as Machine Learning (ML), Human–Computer Interaction (HCI), Information Retrieval (IR), and Information Systems (IS).
Hence, the focus lies on research improving the user experience and following goals such as user engagement and satisfaction or customer value as has also been advocated by Zanker et al. [45]. The aim of this fourth edition
• an invited keynote by Xin Luna Dong (Meta) on “Next-Generation Intelligent Assistants for AR/VR Devices”; • the presentation of the selected research papers.
KaRS 2022 for their thorough reviews and their detailed feedback they gave to the authors. The PC consisted of the following international experts: Aris Anagnostopoulos (Sapienza University of Rome), Vito Walter Anelli (Politecnico di Bari), Marco Angelini (Sapienza University of Rome), Pierpaolo Basile (Dipartimento di Informatica - University of Bari), Roberto Basili (Dept. of Enterprise Engineering - Univ. of Roma Tor Vergata), Alejandro Bellogin (Universidad Autonoma de Madrid), Ludovico Boratto (University of Cagliari), Eric Charton (Banque Nationale du Canada), Giandomenico Cornacchia (Politecnico di Bari), Fabio Crestani (Università della Svizzera Italiana) (USI), Danilo Croce (Dept. of Enterprise Engineering - Univ. of Roma Tor Vergata), Marco de Gemmis (University of Bari Aldo Moro, Dept. of Computer Science), Tommaso Di Noia (Politecnico di Bari), Davide Di Ruscio (Università degli Studi dell’Aquila), Fabrizio Falchi (ISTI-CNR), Antonio Ferrara (Politecnico di Bari), Maurizio Ferrari Dacrema (Politecnico di Milano), Andrea Iovine (Università degli Studi di Bari Aldo Moro), Dietmar Jannach (University of Klagenfurt), Daniele Malitesta (Polytechnic University of Bari), Rubén Francisco Manrique (Universidad de los Andes), Olga Marino (Universidad de los Andes), David Massimo (Free University of Bolzano), Franco Maria Nardini (ISTI-CNR), Fedelucio Narducci (Politecnico di Bari), Rafaele Perego (ISTI-CNR), Marco Polignano (Università degli Studi di Bari Aldo Moro), Claudio Pomo (Politecnico di Bari), Yongli Ren (RMIT University), Gaetano Rossiello (IBM Research AI), Pablo Sánchez (Universidad Autónoma de Madrid), Giovanni Semeraro (University of Bari), Damiano Spina (RMIT University), Alain Starke (Wageningen University & Research) September 2021 - 1 October 2021, ACM, 2021, pp. ence reasoning with soft constraints in constraint806–809. based recommender systems, Constraints 15 (2010) [7] V. W. Anelli, P. Basile, T. D. Noia, F. M. Donini, 574–595.
C. Musto, F. Narducci, M. Zanker, H. Abdollahpouri, [18] V. W. Anelli, T. D. Noia, E. D. Sciascio, A. Ragone, T. Bogers, B. Mobasher, C. Petersen, M. S. Pera J. Trotta, How to make latent factors inter(Eds.), Joint Workshop Proceedings of the 3rd Edi- pretable by feeding factorization machines with tion of Knowledge-aware and Conversational Rec- knowledge graphs, in: C. Ghidini, O. Hartig, ommender Systems (KaRS) and the 5th Edition of M. Maleshkova, V. Svátek, I. F. Cruz, A. Hogan, Recommendation in Complex Environments (Com- J. Song, M. Lefrançois, F. Gandon (Eds.), The SemanplexRec) co-located with 15th ACM Conference tic Web - ISWC 2019 - 18th International Semantic on Recommender Systems (RecSys 2021), Virtual Web Conference, Auckland, New Zealand, October Event, Amsterdam, The Netherlands, September 25, 26-30, 2019, Proceedings, Part I, volume 11778 of 2021, volume 2960 of CEUR Workshop Proceedings, Lecture Notes in Computer Science, Springer, 2019, CEUR-WS.org, 2021. pp. 38–56. [8] D. Jannach, P. Resnick, A. Tuzhilin, M. Zanker, [19] V. W. Anelli, T. D. Noia, E. D. Sciascio, A. Ferrara, Recommender systems—beyond matrix completion, A. C. M. Mancino, Sparse feature factorization for Communications of the ACM 59 (2016) 94–102. recommender systems with knowledge graphs, in: [9] I. Fernández-Tobías, I. Cantador, P. Tomeo, V. W. RecSys, ACM, 2021, pp. 154–165.
Anelli, T. D. Noia, Addressing the user cold start [20] V. W. Anelli, A. Bellogín, A. Ferrara, D. Malitesta, with cross-domain collaborative filtering: exploit- F. A. Merra, C. Pomo, F. M. Donini, T. D. Noia, ing item metadata in matrix factorization, User V-elliot: Design, evaluate and tune visual recomModel. User-Adapt. Interact. 29 (2019) 443–486. mender systems, in: RecSys, ACM, 2021, pp. [10] V. W. Anelli, A. Bellogín, A. Ferrara, D. Malitesta, 768–771.
F. A. Merra, C. Pomo, F. M. Donini, T. D. Noia, El- [21] E. Palumbo, D. Monti, G. Rizzo, R. Troncy, E. Barliot: A comprehensive and rigorous framework for alis, entity2rec: Property-specific knowledge graph reproducible recommender systems evaluation, in: embeddings for item recommendation, Expert Syst.
SIGIR, ACM, 2021, pp. 2405–2414. Appl. 151 (2020) 113235. [11] V. W. Anelli, L. Belli, Y. Deldjoo, T. D. Noia, A. Fer- [22] Y. Zhang, X. Xu, H. Zhou, Y. Zhang, Distilling rara, F. Narducci, C. Pomo, Pursuing privacy in structured knowledge into embeddings for explainrecommender systems: the view of users and re- able and accurate recommendation, in: J. Caverlee, searchers from regulations to applications, in: Rec- X. B. Hu, M. Lalmas, W. Wang (Eds.), WSDM ’20: Sys, ACM, 2021, pp. 838–841. The Thirteenth ACM International Conference on [12] M. Polignano, F. Narducci, M. de Gemmis, G. Se- Web Search and Data Mining, Houston, TX, USA, meraro, Towards emotion-aware recommender February 3-7, 2020, ACM, 2020, pp. 735–743. systems: an afective coherence model based on [23] C. Ni, K. S. Liu, N. Torzec, Layered graph embedding emotion-driven behaviors, Expert Systems with for entity recommendation using wikipedia in the Applications 170 (2021) 114382. yahoo! knowledge graph, in: A. E. F. Seghrouchni, [13] V. W. Anelli, Y. Deldjoo, T. Di Noia, A. Ferrara, G. Sukthankar, T. Liu, M. van Steen (Eds.), ComF. Narducci, Federank: User controlled feedback panion of The 2020 Web Conference 2020, Taipei, with federated recommender systems, in: D. Hiem- Taiwan, April 20-24, 2020, ACM / IW3C2, 2020, pp. stra, M.-F. Moens, J. Mothe, R. Perego, M. Potthast, 811–818.
F. Sebastiani (Eds.), Advances in Information Re- [24] T. Dettmers, P. Minervini, P. Stenetorp, S. Riedel, trieval, Springer International Publishing, Cham, Convolutional 2d knowledge graph embeddings, in: 2021, pp. 32–47. S. A. McIlraith, K. Q. Weinberger (Eds.), Proceedings [14] V. W. Anelli, Y. Deldjoo, T. D. Noia, A. Ferrara, of the Thirty-Second AAAI Conference on Artificial F. Narducci, How to put users in control of their Intelligence, (AAAI-18), the 30th innovative Applidata in federated top-n recommendation with learn- cations of Artificial Intelligence (IAAI-18), and the ing to rank, in: SAC, ACM, 2021, pp. 1359–1362. 8th AAAI Symposium on Educational Advances [15] R. Burke, Knowledge-based recommender systems. in Artificial Intelligence (EAAI-18), New Orleans, encyclopedia of library and information systems: Louisiana, USA, February 2-7, 2018, AAAI Press, Vol. 6.(supplement 32), 2000. 2018, pp. 1811–1818. [16] A. Felfernig, G. Friedrich, D. Jannach, M. Zanker, [25] M. Nayyeri, S. Vahdati, X. Zhou, H. S. Yazdi, Constraint-based recommender systems, in: Rec- J. Lehmann, Embedding-based recommendations ommender Systems Handbook, 2015, pp. 161–190. on scholarly knowledge graphs, in: A. Harth, S. Kir[17] M. Zanker, M. Jessenitschnig, W. Schmid, Prefer- rane, A. N. Ngomo, H. Paulheim, A. Rula, A. L. Gentile, P. Haase, M. Cochez (Eds.), The Semantic Web - cational Advances in Artificial Intelligence, EAAI 17th International Conference, ESWC 2020, Herak- 2019, Honolulu, Hawaii, USA, January 27 - February lion, Crete, Greece, May 31-June 4, 2020, Proceed- 1, 2019, AAAI Press, 2019, pp. 3060–3067. ings, volume 12123 of Lecture Notes in Computer [32] X. Wang, X. He, Y. Cao, M. Liu, T. Chua, KGAT: Science, Springer, 2020, pp. 255–270. knowledge graph attention network for recommen[26] V. Bellini, A. Schiavone, T. D. Noia, A. Ragone, E. D. dation, in: A. Teredesai, V. Kumar, Y. Li, R. RosSciascio, Computing recommendations via a knowl- ales, E. Terzi, G. Karypis (Eds.), Proceedings of the edge graph-aware autoencoder, in: V. W. Anelli, 25th ACM SIGKDD International Conference on T. D. Noia, P. Lops, C. Musto, M. Zanker, P. Basile, Knowledge Discovery & Data Mining, KDD 2019, D. G. Bridge, F. Narducci (Eds.), Proceedings of Anchorage, AK, USA, August 4-8, 2019, ACM, 2019, the Workshop on Knowledge-aware and Conversa- pp. 950–958. tional Recommender Systems 2018 co-located with [33] X. Wang, D. Wang, C. Xu, X. He, Y. Cao, T.-S. Chua, 12th ACM Conference on Recommender Systems, Explainable reasoning over knowledge graphs for KaRS@RecSys 2018, Vancouver, Canada, October 7, recommendation, in: Proceedings of the AAAI 2018, volume 2290 of CEUR Workshop Proceedings, Conference on Artificial Intelligence, volume 33, CEUR-WS.org, 2018, pp. 9–15. 2019, pp. 5329–5336. [27] G. He, J. Li, W. X. Zhao, P. Liu, J. Wen, Mining im- [34] V. W. Anelli, T. D. Noia, E. D. Sciascio, A. Ragone, plicit entity preference from user-item interaction J. Trotta, Semantic interpretation of top-n recomdata for knowledge graph completion via adversar- mendations, IEEE Trans. Knowl. Data Eng. 34 (2022) ial learning, in: Y. Huang, I. King, T. Liu, M. van 2416–2428.
Steen (Eds.), WWW ’20: The Web Conference 2020, [35] H. Wang, F. Zhang, J. Wang, M. Zhao, W. Li, X. Xie, Taipei, Taiwan, April 20-24, 2020, ACM / IW3C2, M. Guo, Exploring high-order user preference on 2020, pp. 740–751. the knowledge graph for recommender systems, [28] Y. Cao, X. Wang, X. He, Z. Hu, T. Chua, Unifying ACM Trans. Inf. Syst. 37 (2019) 32:1–32:26. knowledge graph learning and recommendation: [36] V. W. Anelli, R. D. Leone, T. D. Noia, T. Lukasiewicz, Towards a better understanding of user preferences, J. Rosati, Combining RDF and SPARQL with cpin: L. Liu, R. W. White, A. Mantrach, F. Silvestri, J. J. theories to reason about preferences in a linked McAuley, R. Baeza-Yates, L. Zia (Eds.), The World data setting, Semantic Web 11 (2020) 391–419. Wide Web Conference, WWW 2019, San Francisco, [37] T. Di Noia, V. C. Ostuni, Recommender systems and CA, USA, May 13-17, 2019, ACM, 2019, pp. 151–161. linked open data, in: Reasoning Web Int. Summer [29] G. Piao, J. G. Breslin, Transfer learning for item School, Springer, 2015, pp. 88–113. recommendations and knowledge graph comple- [38] T. Di Noia, C. Magarelli, A. Maurino, M. Palmonari, tion in item related domains via a co-factorization A. Rula, Using ontology-based data summarizamodel, in: A. Gangemi, R. Navigli, M. Vidal, P. Hit- tion to develop semantics-aware recommender syszler, R. Troncy, L. Hollink, A. Tordai, M. Alam (Eds.), tems, in: The Semantic Web - 15th Int. Conf., ESWC The Semantic Web - 15th International Conference, 2018, Heraklion, Crete, Greece, June 3-7, 2018, Proc., ESWC 2018, Heraklion, Crete, Greece, June 3-7, Springer New York, 2018, pp. 128–144. 2018, Proceedings, volume 10843 of Lecture Notes [39] K. Zhou, W. X. Zhao, S. Bian, Y. Zhou, J.-R. Wen, in Computer Science, Springer, 2018, pp. 496–511. J. Yu, Improving conversational recommender sys[30] M. S. Schlichtkrull, T. N. Kipf, P. Bloem, R. van den tems via knowledge graph based semantic fusion, Berg, I. Titov, M. Welling, Modeling relational in: Proceedings of the 26th ACM SIGKDD Internadata with graph convolutional networks, in: tional Conference on Knowledge Discovery & Data A. Gangemi, R. Navigli, M. Vidal, P. Hitzler, Mining, 2020, pp. 1006–1014.
R. Troncy, L. Hollink, A. Tordai, M. Alam (Eds.), [40] D. Jannach, A. Manzoor, W. Cai, L. Chen, A survey The Semantic Web - 15th International Conference, on conversational recommender systems, arXiv ESWC 2018, Heraklion, Crete, Greece, June 3-7, preprint arXiv:2004.00646 (2020). 2018, Proceedings, volume 10843 of Lecture Notes [41] M. Polignano, F. Narducci, A. Iovine, C. Musto, in Computer Science, Springer, 2018, pp. 593–607. M. De Gemmis, G. Semeraro, Healthassistantbot: [31] C. Shang, Y. Tang, J. Huang, J. Bi, X. He, B. Zhou, A personal health assistant for the italian language, End-to-end structure-aware convolutional net- IEEE Access 8 (2020) 107479–107497. works for knowledge base completion, in: The [42] A. Iovine, F. Narducci, G. Semeraro, Conversational Thirty-Third AAAI Conference on Artificial Intel- recommender systems and natural language:: A ligence, AAAI 2019, The Thirty-First Innovative study through the converse framework, Decision Applications of Artificial Intelligence Conference, Support Systems 131 (2020) 113250. IAAI 2019, The Ninth AAAI Symposium on Edu- [43] S. Zhang, K. Balog, Evaluating conversational recommender systems via user simulation, in: Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2020, pp. 1512–1520. [44] A. Iovine, F. Narducci, M. de Gemmis, A dataset of real dialogues for conversational recommender systems., in: CLiC-it, 2019. [45] M. Zanker, L. Rook, D. Jannach, Measuring the impact of online personalisation: Past, present and future, International Journal of Human-Computer Studies 131 (2019) 160–168.