A Multi-source Graph Representation of the Movie Domain for Recommendation Dialogues Analysis

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 1297–1306 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 1297 A Multi-source Graph Representation of the Movie Domain for Recommendation Dialogues Analysis Antonio Origlia1, Martina Di Bratto1, Maria Di Maro1, Sabrina Mennella2 1University of Naples Federico II, 2University of Catania {antonio.origlia, martina.dibratto, maria.dimaro2}@unina.it, sabrina.mennella@phd.unict.it Abstract In dialogue analysis, characterising named entities in the domain of interest is relevant in order to understand how people are making use of them for argumentation purposes. The movie recommendation domain is a frequently considered case study for many applications and by linguistic studies and, since many different resources have been collected throughout the years to describe it, a single database combining all these data sources is a valuable asset for cross-disciplinary investigations. We propose an integrated graph-based structure of multiple resources, enriched with the results of the application of graph analytics approaches to provide an encompassing view of the domain and of the way people talk about it during the recommendation task. While we cannot distribute the final resource because of licensing issues, we share the code to assemble and process it once the reference data have been obtained from the original sources. Keywords: graph databases, movies domain, cross-resource analysis 1. Introduction Graph databases have gained popularity, in recent years, for their capability to efficiently represent com- plex data in an interpretable, flexible format. In particular, graph databases are well-suited in cross- referencing different resources so that information dis- tributed in multiple sources can be analysed in an in- tegrated way. Graph databases have found applica- tion in many different fields requiring the management of large datasets characterised by complex interactions among items. In the field of Natural Language Pro- cessing, Word Sense Disambiguation (Yao et al., 2017), Knowledge Base Collection (Yu et al., 2020) and Fraud Detection (Srivastava and Singh, 2018; Stray, 2019) are strongly investigated areas. Also, node embedding techniques have been significantly improved to capture latent network semantics to support machine learning tasks (Chanpuriya et al., 2020; Yang et al., 2020). In this sense, the movies domain represents a well stud- ied case of interest for a number of different tasks, from dialogue analysis in recommendation (Lee and Choi, 2017) to movie revenues prediction (Zhou et al., 2019) and movie rating prediction (Forouzandeh et al., 2021). As there are many different approaches and points of view about this specific domain, we present a data pro- cessing pipeline to integrate, in a single, graph-based, resource, the information contained in some of the most relevant resources that are available online. Also, we apply graph analysis techniques to enrich the obtained network with latent information extracted from the fi- nal graph structure. More in detail, the procedure can be summarised in three steps: • Data collection • Graph database assembling • Data enrichment The main goal of this work is to support investiga- tion in the field of argumentation based dialogue, for which a general theoretical …

Content truncated to 500 words