Multi-view Story Characterization from Movie Plot Synopses and Reviews

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 5629–5646, November 16–20, 2020. c⃝2020 Association for Computational Linguistics 5629 Multi-view Story Characterization from Movie Plot Synopses and Reviews Sudipta Kar♣, Gustavo Aguilar♣, Mirella Lapata♠, Thamar Solorio♣ ♣ University of Houston ♠ ILCC, University of Edinburgh {skar3, gaguilaralas}@uh.edu, mlap@inf.ed.ac.uk, tsolorio@uh.edu Abstract This paper considers the problem of charac- terizing stories by inferring properties such as theme and style using written synopses and reviews of movies. We experiment with a multi-label dataset of movie synopses and a tagset representing various attributes of sto- ries (e.g., genre, type of events). Our pro- posed multi-view model encodes the synopses and reviews using hierarchical attention and shows improvement over methods that only use synopses. Finally, we demonstrate how can we take advantage of such a model to ex- tract a complementary set of story-attributes from reviews without direct supervision. We have made our dataset and source code pub- licly available at https://ritual.uh.edu/ multiview-tag-2020. 1 Introduction A high-level description of stories represented by a tagset can assist consumers of story-based me- dia (e.g., movies, books) during the selection pro- cess. Although collecting tags from users is time- consuming and often suffers from coverage issues (Katakis et al., 2008), NLP techniques like those in Kar et al. (2018b) and Gorinski and Lapata (2018) can be employed to generate tags automatically from written narratives such as synopses. However, existing supervised approaches suffer from two sig- nificant weaknesses. Firstly, the accuracy of the extracted tags is subject to the quality of the syn- opses. Secondly, the tagset is predefined by what was present in the training and development sets and thus is brittle; story attributes are unbounded in principle and grow with the underlying vocabulary. To address the weaknesses presented above, we propose to exploit user reviews. We have found that movie reviews often discuss many aspects of the story. For example, in Figure 1, a reviewer writes that The Godfather is about family, rela- Plot Synopsis In late summer 1945, guests are gathered for the wedding reception of Don Vito Corleone’s daughter Connie (Talia Shire) and Carlo Rizzi (Gianni Russo). ... ... The film ends with Clemenza and new ca- poregimes Rocco Lampone and Al Neri arriving and paying their re- spects to Michael. Clemenza kisses Michael’s hand and greets him as ”Don Corleone.” As Kay watches, the office door is closed.” Review Even if the viewer does not like mafia type of movies, he or she will watch the entire film ... ... Its about family, loyalty, greed, relationships, and real life. This is a great mix, and the artistic style make the film memorable. violence action murder atmospheric revenge mafia family loyalty greed relationship artistic Figure 1: Example snippets from plot synopsis and review of The Godfather and tags that can be generated from these. tions, loyalty, greed, and mafia, whereas the gold standard tags from the plot are violence, murder, atmospheric, action, and revenge. In this paper, we show that such …

Content truncated to 500 words