The challenge of heterogeneity in recommender systems

• Název práce v češtině: Problém heterogenity v doporučovacích systémech
• Název v anglickém jazyce: The challenge of heterogeneity in recommender systems
• Klíčová slova: doporučovací systémy|heterogenita
• Klíčová slova anglicky: reommender systems|heterogeneity
• Akademický rok vypsání: 2021/2022
• Typ práce: disertační práce
• Jazyk práce: angličtina
• Ústav: Katedra softwarového inženýrství (32-KSI)
• Vedoucí / školitel: Mgr. Ladislav Peška, Ph.D.
• Řešitel: skrytý - zadáno a potvrzeno stud. odd.
• Datum přihlášení: 01.09.2021
• Datum zadání: 01.09.2021
• Datum potvrzení stud. oddělením: 14.09.2021

Zásady pro vypracování

The aim of this thesis is to provide novel insight on some axis of the heterogeneity challenges in recommender systems (RS). The topic is rather broad and should be narrowed down based on the initial discussions and particular interests of the enrolled student. Some options are examining the role of heterogeneous metrics in long-term sustainability of RS, the role of available data sources on overall RS performance or devising proper means for aggregation of heterogeneous RS methods. The work on this thesis should incorporate the usage of existing as well as development of new models, algorithms, tools, and benchmarks. Fundamental part of the work will be comparison on real data and real-world applications of proposed models.

Seznam odborné literatury

Abdollahpouri, H., Mansoury, M., Burke, R., Mobasher, B.: The connection between popularity bias, calibration, and fairness in recommendation. RecSys ’20, p. 726–731. ACM (2020).
Anelli, V.W., Bellini, V., Di Noia, T., La Bruna, W., Tomeo, P., Di Sciascio, E.: Ananalysis on time- and session-aware diversification in recommender systems. UMAP ’17, p. 270–274. ACM (2017).
Bertani, R.M., A. C. Bianchi, R., Costa, A.H.R.: Combining novelty and popularityon personalised recommendations via user profile learning. Expert Systems with Applications 146, 113149 (2020).
Fleder, D., Hosanagar, K.: Blockbuster culture’s next rise or fall: The impact of recommender systems on sales diversity. Manage. Sci. 55(5), 697–712 (2009).
Ge, Y., Zhao, S., Zhou, H., Pei, C., Sun, F., Ou, W., Zhang, Y.: Understanding Echo Chambers in E-Commerce Recommender Systems, p. 2261–2270. ACM (2020).
Garcin, F., Faltings, B., Jurca, R., Joswig, N.: Rating aggregation in collaborative filtering systems. RecSys’09, p. 349–352. ACM (2009).
Jannach, D., Ludewig, M., Lerche, L.: Session-based item recommendation in e-commerce: on short-term intents, reminders, trends and discounts. User Modeling and User-Adapted Interaction 27(3), 351–392 (2017).
T. Joachims, A. Swaminathan, and T. Schnabel. Unbiased learning-to-rank with biased feedback. In WSDM’17, pages 781–789. ACM, 2017
Kaminskas, M., Bridge, D.: Diversity, serendipity, novelty, and coverage: A survey and empirical analysis of beyond-accuracy objectives in recommender systems. ACM Trans. Interact. Intell. Syst. 7(1) (2016).
L. Peska and P. Vojtas. Using implicit preference relations to improve recommender systems. Journal on Data Semantics, 6(1):15–30, Mar 2017
Steck, H.: Calibrated recommendations. RecSys ’18, pp. 154–162. ACM (2018)
Starychfojtu, J., Peska, L.: SmartRecepies: Towards cooking and food shopping integration via mobile recipes recommender system. In: iiWAS ’20, ACM (2020)
X. Yi, L. Hong, E. Zhong, N. N. Liu, and S. Rajan. Beyond clicks: Dwell time for personalization. RecSys’14, page 113–120. ACM, 2014.

Předběžná náplň práce v anglickém jazyce

Recommender systems (RS) are an important and intensively studied branch of information retrieval. The main aim of RS is to proactively propose unknown yet relevant objects based on individual users’ feedback and preferences. Despite the years of RS research, there are still important unresolved challenges mostly connected with various notions of heterogeneity in RS. One possible axis is heterogeneous source data available for individual scenarios. Some examples are multimodal user feedback, raw text or audio-visual data or non-trivial contextual information. Such data may lead to novel hypotheses about user’s behavior followed by novel algorithms capable to process such data properly. The need for such solutions is fostered by continuously emerging novel domains of RS, which often provides rather unique conditions and challenges for RS applications. Goals of RS also moved beyond simple result’s relevance and nowadays aim to incorporate e.g. results diversity, novelty, user profile calibration, fairness w.r.t. different aspects or aiming to break the filtering bubble. Such heterogeneous goals are both interesting from the algorithmic perspective as well as provide several challenges for RS evaluation.