A Consideration to Estimate Spoiling Pages in Comics

>100 Views

September 27, 18

スライド概要

ISASE2018で発表したスライドです

profile-image

Nakamura Laboratory (Meiji University)

@nkmr-lab

スライド一覧

明治大学 総合数理学部 先端メディアサイエンス学科 中村聡史研究室

シェア

埋め込む »CMSなどでJSが使えない場合

各ページのテキスト
1.

A Consideration to Estimate Spoiling Pages in Comics Yoshiki Maki Yuji Shiratori Kenta Sato Satoshi Nakamura Graduate School of Advanced Mathematical Sciences, Meiji University

2.

Do you have experience to get disappointed by encountering the comic spoiler?

3.

Ms.△△ ? This is funny lol She is dead !! Have you read 〇〇○? Yes!!! Oh boy... I love Ms.△△ !!!

4.

Character A Character B

5.

Oh NO!!! I got spoiler!!

6.

Ms.△△ ? This is funny lol She is dead !! Conversation Tweet We want to stop these Spoiler Websites Images

7.

Ms.△△ ? This is funny lol She is dead !! Conversation Tweet Difficult to stop this spoiler Websites Images

8.

Ms.△△ ? This is funny lol She is dead !! Conversation The Twitter Mute Button: A Web Tweet Filtering Challenge [Golbeck 12] Enable to block TV program spoilers on Twitter by making datasets Websites Using this method, we can stop comic spoilers too Images

9.

Ms.△△ ? This is funny lol She is dead !! Conversation Proposal of deleting plots from the reviews to Tweet the items with stories [Ikeda 10] Enable to block plot spoilers of story contents on review Websites Using this method, we can stop comic spoilers too Images

10.

Ms.△△ ? This is funny lol She is dead !! No study to block image spoiler Conversation Tweet Focus on these spoiler Websites Images

11.

Purpose Stop images spoiler of comic contents

12.

Related Work Story Spoilers Don’t Spoil Stories [Leavitt, 2011] In novel, Spoiler make more interesting Only clarify novel spoiler,not comic spoiler

13.

Purpose Stop the images spoiler Investigating the influence comiccontents spoiler ofofcomic Generate a pre-datasets of spoilers Make an pre-experiment whether the spoiler effect depending on spoiler timing or not

14.

Generate Pre-Datasets

15.

Generate Datasets Why “Generate datasets?” In related work, writer make a spoiler and we don’t think these spoiler is real spoiler. We want to select real spoiler but there are no study to clarify real spoiler. So we generate datasets by asking collaborators to select spoiler page Which page is the spoiler?

16.

Datasets(Used Manga) Battle ©️Yoshihiro Tomigashi ©️Takeshi Obata Sports ©️Tsujitomo ©️Taiyo Matsumoto Mystery ©️Aki Shimizu ©️Tetsuya Tsutsu Romance ©️Sotara Akiduki ©️Toko Minami

17.

Datasets(Design) Genre×2 Comic = 8 Comics Amount Manga Participants Question a4 people who already read each comic Select and Rank(1st to 3rd) pages with spoilers

18.

Datasets(Selected Scenes) Battle Sports Mystery Romance Dead or Alive Win / Lose Detect mystery Confess love ©️Yoshihiro Tomigashi Dead or Alive ©️Takeshi Obata ©️Tsujitomo Ending ©️Taiyo Matsumoto ©️Aki Shimizu Ending ©️Tetsuya Tsutsu ©️Sotara Akiduki Confess love ©️Toko Minami

19.

PreExperiment

20.

Pre-Experiment(Design) 4 Genre×2 Comic = 8 Comics Amount Manga Participants a120 people who not read each manga Enjoyment value Question Interest of the next episode

21.

Experiment(Design) Without Spoiler Outline Part 2 Story Part 1 Comics Part 3 Spoiled After Outline Outline Part 1 Comics Part 2 Story Part 3 Spoiled After Part 1 Outline Part 2 Part 1 Comics Story Part 3 Spoiled After Part 2 Outline Part 1 Comics Story Part 2 Part 3 Spoiler Timing Questionnaire Timing

22.

Experiment(Design) Without Spoiler Outline Part 2 Story Part 1 Manga Spoiled After Outline Outline 2 2 PartPart 3 3 PartPart 1 1 PartPart Spoiled After Part 1 Outline Part 1 3 3 PartPart 2 2 PartPart Spoiled After Part 2 Outline Part 1 Part 2 Spoiler Timing Part 3 PartPart 3 3 Questionnaire Timing

23.

Enjoyment Value at last in each group Enjoyment value is decreasing depending on timing 2.0 1.5 1.0 0.5 0.0 Without Spoiler -0.5 After Outline After Part.1 After Part.2 No significant difference

24.

Translate Bar of Interest Of The Next Value Graph 2.0 ● Spoiler Timing 1.5 1.0 0.5 0.0 Outline Spoiler Part 1 Spoiler Part 2 Spoiler Part 3 -0.5 Pattern 0 Pattern 1 Pattern 2 Pattern 3

25.

Translate Bar of Interest Of The Next Value Graph ● Spoiler Timing 2.0 1.5 Increasing at after outline group(10%) 1.0 0.5 0.0 Outline Spoiler Part 1 Spoiler Part 2 Spoiler Part 3 -0.5 Pattern 0 Pattern 1 Pattern 2 Pattern 3 Decreasing at after Part2 group(5%)

26.

Discussion Reading Last Part Even if spoiling, enjoyment value is not changing Spoiler contents is Immediately after spoiler bad contents! If spoiling next episode, reader’s interest of next episode is decreasing The spoiler contents has a possibility to decrease readers purchase motivation

27.

Discussion Reading Last Part Even if spoiling, enjoyment value is not changing Immediately after spoiler If spoiling next episode, reader’s interest of next episode is decreasing The spoiler contents has a possibility to decrease readers purchase motivation

28.

Discussion Think and redefine “What contents is spoiler” Immediately after spoiler If spoiling next episode, reader’s interest of next episode is decreasing

29.

N-2 What is Spoiler? Redefinition about the spoiler The spoiler contents is the contents which reader get disappointed in episode N+1 if reader had read until episode N. N-1 N N+1 N+2

30.

Purpose Investigating the influence SOLVED Of comic spoiler Find a characteristics of spoiler pages Stop the images spoiler Making estimating spoiler of method comicofcontents

31.

Characteristics Of Spoilers

32.

Characteristics Find A Special Features To create method of estimating spoiler images, we analysis spoiler pages in pre-datasets and find a special features in spoiler pages focus on frame size and words in spoiler pages. Words Frame

33.

Characteristics(Selected Spoiler) Battle Sports Mystery Romans

34.

Characteristics(Several Genres) Battle Frames Bigger frames than other pages Words “I won” and “Finished” Win/Lose words

35.

Characteristics(Several Genres) Sports Frames Big frame and small frames Words Little words

36.

Characteristics(Several Genres) Mystery Frame Bigger frame or smaller frames Words Many words or little words

37.

Characteristics(Several Genres) Romans Frame Bigger size than other pages Words “I love you”

38.

Characteristics(Selected Spoiler) Battle Sports Mystery Romans Collaborators can select only three pages in a comic Must make real spoiler datasets

39.

Method Of Create Datasets

40.

N-2 Method Question to the contributors Please select facing pages which you think that if you show the facing page in the story number N+1 to your friend who had read until the story number N, your friend will angry with you. N-1 N N+1 N+2

41.

Method Red card / yellow card Ask contributors to set two levels of spoiler to these pages. Yellow Card Red Card Spoiler Rating: 0~49% Spoiler Rating: 50~100%

42.

How To Estimate Spoiler Pages

43.

Estimate Spoiler Page Layout Analysis of Tree-Structured Scene Frames in Comic Images [Tanaka 2007] Frame Vision API[Google 18] Word 1000~2000 Spoiler Pages Page

44.

Summary Pre-Experiment Generate a pre-datasets of spoilers Make an pre-experiment whether the spoiler effect depending on spoiler timing or not Stop the spoiler Find a characteristics of spoiler pages Making method of estimating spoiler

45.

Translate Bar of Enjoyment Value Graph ● Spoiler Timing 2.0 1.5 1.0 0.5 0.0 Outline Spoiler1 Part.1 Spoiler2 Part.2 Spoiler3 Part.3 -0.5 Without Spoiler With Spoiler1 With Spoiler2 With Spoiler3

46.

Translate Bar of Enjoyment Value Graph ● Spoiler Timing 2.0 1.5 Decreasing at after Part3 spoiler(10%) 1.0 0.5 0.0 Outline Spoiler1 Part.1 Spoiler2 Part.2 Spoiler3 Part.3 -0.5 Without Spoiler With Spoiler1 With Spoiler2 With Spoiler3

47.

OCR ONLINE OCR LI いい音だを ゼンに会って 1織にい る嚇で ーフん ヴつと 加舶加加鵬 JJI且1 い叫痢 し」 Tesseract _鱒 慣のと 薔れ咆搬っ疋 泄ウに肩鱗う Vision API 音だな うん オロ 知って しまった ゼンに会って 一緒にいる中で ずっと 途切れなかった ゼンに向かう 気持ちの 中に まだ M+ f 見つけていない 感情があった ゼン ゼンのことが 好きです SAS