A method to construct comics spoiler dataset and the analysis of comics spoilers

A method to construct comics spoiler dataset and the analysis of comics spoilers Takumi Takakua, Yoshiki Makia, Satoshi Nakamuraa a Meiji University, Nakano 4-21-1, Nakano-ku, Tokyo, 164-8525, Japan 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Q. Have you ever encountered a spoiler? 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Background • Spoiler is usually known as revealing key plot details or outcomes of contents before someone has had a chance to experience it themselves. • People might encounter spoilers even if you try to avoid them. 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Background • I had an experience being spoiled on Twitter. 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Background • Spoilers can happen unintentionally in many different places, such as SNS, review websites, thumbnail images, etc. • Various studies have been conducted to investigate the issue of spoilers. ➢ Spoilers in film affect consumer behavior and found that spoilers makes people more reluctant to watch movies. [Tsang et al. 2009] ➢ Clarifying the characteristics of sports spoilers on Twitter, and proposed a method for judging the spoilers. [Shiratori et al. 2018] • • Tsang. A. S. L. and Yan, D. (2009) “Reducing the Spoiler Effect in Experiential Consumption.” Advances in Consumer Research, 36: 708–709. Yuji, S. Maki, Y. Satoshi, N. and Takanori, K. (2018) “Detection of Football Spoilers on Twitter” The 10th International Conference on Collaboration Technologies (CollabTech 2018): 129–141. 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Our Previous Study • We conducted an experiment to find out the effects of spoilers on comics. [Maki et al. 2017, 2018] • We focused on comics because they are published in more extensive serialization periods than other content, such as novels, dramas, animations, and movies. • We divided the participants’ groups into those with spoilers and those without spoilers, and found no significant difference in enjoyment between them. • • Maki, Y., Shiratori, Y., Sato, K. and Nakamura, S. (2018) “A Consideration to Estimate Spoiling Pages in Comics” International Symposium on Affective Science and Engineering ISASE2018 4: 1–6. Maki, Y. and Nakamura, S. (2017) “Do Manga Spoilers Spoil Manga?” The Sixth Asian Conference on Information Systems: 258–262. 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Our Previous Study • We conducted an experiment to find out the effects of spoilers on comics. [Maki et al. 2017, 2018] • There was a issue on our previous studies. ➢ The definition of a spoiler was ambiguous, so there could be a possibility that participants had different criteria for spoilers. 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Our Previous Study • We conducted an experiment to find out the effects of spoilers on comics. [Maki et al. 2017, 2018] • There was a issue on our previous studies. ➢ The definition of a spoiler was ambiguous, so there could be a possibility that participants had different criteria for spoilers. It is necessary to define the definition of spoilers in comics, and reinvestigate the effects on comics. 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Purpose of our study 1. We first clarify the definition of spoilers. 2. Ask experimental collaborators to select comics book pages that match the defined spoilers. 3. By analyzing the spoiler data obtained from the collaborators, we will consider what kinds of places are likely to be avoided as spoilers. 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Defining the spoilers • Why defining Spoilers? ➢ Previous spoiler studies had imprecise and deficient spoilers. • Spoilers on short stories [Leavitt et al. , 2011] ➢ They only use punchlines as spoilers. • Spoilers on comics [Hassoun, 2013] ➢ The last panel of the facing page was the spoiler. • Spoilers are likely to be responded to differently depending on where readers are read and where spoilers are presented 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Defining the spoilers Readers reading Spoilers Presented 91 Volumes 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Defining the spoilers Readers reading Spoilers Presented 91 Volumes Readers do not notice spoilers since the place where they are reading and where the spoiler is presented are too far apart. 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Defining the spoilers Readers reading Spoilers Presented 9 Volumes 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Defining the spoilers Readers reading Spoilers Presented 9 Volumes Where readers are reading and spoiler is presented are close to each other, so they perceive it as a spoiler and feel unpleasant 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Definition of spoiler • Spoilers are likely to be responded to differently depending on where readers are read and where spoilers are presented • We define spoiler as : A page p in episode N+1 that would be objectionable to others who have read up to episode N of the comic when it is presented 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Experiment & Constructing Spoiler Dataset Using the definition we defined, we conducted the experiment to find out what kind of episodes are treated as spoilers. • Therefore, we construct spoiler datasets based on the experiments’ data, and compare the results with previous studies. • Battle ©️Takeshi Obata Sports ©️Tsujitomo Mystery ©️Tetsuya Tsutsu Romance ©️Toko Minami 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Flow of the experiment 1 We asked participants to read the Nth Story. N starts from 1 2 We listed the N-1th thumbnail pages and asked participants to browse them briefly. 3 Select a spoiler for each page of the Nth story based on the definition. They also select the 3 levels on degree of spoilers. Choosing Spoilers © Ken Akamatsu 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Flow of the experiment Choosing Spoilers • We asked participants to select the tags based on the description below ➢ Red ➢ Yellow : Probably a Spoiler ➢ Blue : Definitely a Spoiler : Not a Spoiler © Ken Akamatsu 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Experiment Summary • Experiment in which participants are asked to identify comic book spoilers. • Told participants to “When deciding the spoiler page, be aware that whether other people would find it offensive, not you.“ • 5 university students, aged between 21- 25 • We did not consider whether participants had read the comic oftenly. 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Results • We scored spoilers based on the degree of spoilers tagged. ➢ Red ➢ Yellow : 1 point ➢ Blue • For each page, higher points means high degree of spoilers, and lower points means low degree of spoilers. Choosing Spoilers : 2 points : 0 point © Ken Akamatsu 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Results (All You Need is Kill) Dataset for the current study Dataset for the Previous study Battle ©️Take shi 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Results (All You Need is Kill) Mystery of story solved Mystery of heroine solved Dataset for the current study Punchline of the Story Dataset for the Previous study Battle ©️Take shi 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Results (GIANT KILLING) Dataset for the current study Dataset for the Previous study Sports ©️Ts ujito 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Results (GIANT KILLING) Another Match Result Another Match Result Dataset for the current study Match Result Dataset for the Previous study Sports ©️Ts ujito 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Results (PROPHECY) Dataset for the current study Dataset for the Previous study Mystery ©️Tetsuya Tsutsu 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Results (PROPHECY) Motive of the criminal Background of the criminal Dataset for the current study Punchline of the Story Dataset for the Previous study Mystery ©️Tetsuya Tsutsu 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Results (ReRe:Hello!) Dataset for the current study Dataset for the Previous study Romance ©️Toko Minami 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Results (ReRe:Hello!) Outcome of the friend’s love Dataset for the current study Outcome of the love Dataset for the Previous study Romance ©️Toko Minami 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Discussion • For “All You Need Is Kill”, “GIANT KILLING”, and “PROPHECY”, we were able to collect more spoiler pages compare to previous studies. ➢ The main spoilers in previous studies were “punchline of stories”, while in this study, more detailed character stories and other information were also selected • For “ReRe:Hello!”, number of spoilers have not increased much. ➢ This may be due to the fact that the stories are often about everyday life, so there were fewer scenes that could be treated as spoilers. 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Discussion • From the results, we founded out that spoilers do not only occur at the end of the story but can exist in a wide range of stories and that more spoilers may be needed to reduce the interest in the work. 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Discussion • From the results, we founded out that spoilers do not only occur at the end of the story but can exist in a wide range of stories and that more spoilers may be needed to reduce the interest in the work. In spoiler studies, it is necessary to consider not only important spoilers in one part of the story, but also spoilers as a whole. 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Preliminary Investigation • In the future, spoiler detection should be automatic rather than manual. • For automatic spoiler detection, it is necessary to find a relationship between elements in the comic and spoilers. 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Preliminary Investigation • In the future, spoiler detection should be automatic rather than manual. • For automatic spoiler detection, it is necessary to find a relationship between elements in the comic and spoilers. As an element of the comics, we focus on the number of characters per page. 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Preliminary Investigation • We investigated what kind of relationships exists between the number of characters per page and spoilers. • There are various methods for OCR (Optical Character Recognition). Among these methods, we use Mokuro*, which allows users to retrieve images from e-comics and obtain text information through OCR analysis. • We investigate how the number of characters affects the degree of spoilers in the four works mentioned earlier. 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Preliminary Investigation Battle ©️Takeshi Obata Sports ©️Tsujito mo 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Preliminary Investigation Mystery ©️Tetsuya Tsutsu Romance ©️Toko Minami 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Preliminary Investigation • From the investigation, it can be seen that pages with a higher degree of spoilers tend to have fewer characters. ➢ This can be because the characteristic scenes of spoilers tend to be spoiled from comics’ images rather than texts. 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Preliminary Investigation • From the investigation, it can be seen that pages with a higher degree of spoilers tend to have fewer characters. ➢ This can be because the characteristic scenes of spoilers tend to be spoiled from comics’ images rather than texts. Although some scenes do not fit this tendency, there could be a possibility to automatically detect the spoiler scenes using the relationship between number of characters and spoilers. 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Future Prospects • For the spoiler dataset construction, we only asked 5 participants to participate the experiment and only using those data. ➢ • We plan to further study the appropriate amount of spoilers in the dataset We had done preliminary investigation for automatic detection of spoilers just focusing on number of characters. ➢ We plan to use onomatopoeia, the unique characteristics in comics, in addition to handle all spoiler patterns. 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023

Conclusion Purpose: It is necessary to define the definition of spoilers in comics, and reinvestigate the effect on comics. Experiment: Participants were asked to identify the spoiler pages in comics. Results: We were able to collect more spoiler pages than previous studies. We found out the relationship between number of characters and spoilers for auto- detection. Prospects: Finding the appropriate amount of spoilers & improve the auto-detection using onomatopoeia. 27th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Athens, Greece, 06-08 September 2023