japanese visual media graph project introduction
TRANSCRIPT
Japanese Visual Media Graph–
Project Introduction
Prof. Magnus PfefferStuttgart Media University
JVMG Project Introduction 2
Overview Data-driven research in the MAG field Japanese Visual Media Graph project
Main project goals Data representation and integration
JVMG Project Introduction 4
A vast and complex field of study MAG are a research interest of both Japan and media
studies From large commercial franchises with complex media
mix to works by independent artists and producers Multitude of themes and topics Distinct visual styles and expressions Exponential growth of publications
JVMG Project Introduction 5
Lack of comprehensive resources Outdated encyclopedic works by individual scholars Small collections in libraries, with limited metadata
focussed on the media carrier instead of content
No larger bibliographic databases No standardized vocabulary to describe themes,
styles or topics
JVMG Project Introduction 6
Strong need for more research options Search possibilities for both carrier and content
metadata, e.g. characters, themes, and topics Network analysis: how are works connected by shared
contents or contributors? Time analysis: how have certain aspects of visual
media changed over time?
Completeness: what are all the works that share a certain aspect?
JVMG Project Introduction 7
A vibrant enthusiast community International and national enthusiast communities
exist on the web since the 1990s They address the “discovery and access” problems of
their peers by collecting information on Japanese Visual Media
Larger communities strive for complete documentation of their favoured type of media
Enthusiast communities focus on the contents / works first, and carriers second
Most communities use refined entity-relationship data models with strong quality control mechanisms
JVMG Project Introduction 8
A vibrant enthusiast community Examples:
anidb.net (anime) myanimelist.net (anime and manga) vndb.org (visual novel type computer games) animecharactersdatabase.com (characters from all
media) animexx.de (all media, German) animeclick.it (all media, Italian)
JVMG Project Introduction 9
Project partners Martin Roth
Ritsumeikan University Media / Japan Area Studies
Magnus Pfeffer Stuttgart Media University Information Sciences
Leander Seige University of Leipzig University Library
JVMG Project Introduction 10
Experience from prior projects diggr.link (Leipzig University)
Goal: building a data-driven research infrastructure for computer games research
Access to large local game collection Working with enthusiast communities
FID Judaica (Stuttgart Media University) Goal: building an expert information service for the
domain of Jewish studies Access to digitized archives and encyclopedic data Large scale metadata integration
JVMG Project Introduction 12
Main project goals Create a research database on Japanese visual
media, including, but not limited to, anime, manga, computer games and visual novels
Aimed at researchers in Japan Studies who focus on modern media and its expressions, themes, topics, characters and reception
Use a graph-based, highly interconnected database structure, similar to the Google knowledge graph, that is combined with a flexible search interface and analytic tools
Use the data on Japanese visual media that is being created and curated by the many enthusiast communities on the web
JVMG Project Introduction 13
Data model Needs to accommodate
the media itself the entities involved in its creation the contents the relationships between the individual works and their
elements.
Derived from the models used by enthusiast data The final model will not be described as a traditional
metadata schema, but as a data graph containing typed entities with a pre-described set of attributes and typed relations between these entities
JVMG Project Introduction 14
Data model: Methods Create a local RDF model for each data source
Separate namespaces with URIs for all entities, properties and relations
1:1 mapping of the existing data
Create a central model for the graph data and retrieval Learn from enthusiast data models Consider research questions of the media researchers
Map and merge entities, properties and relations
JVMG Project Introduction 15
Data model: Challenges Reuse of existing ontologies or metadata schemata
Pro: Data sharing with other projects; reuse in external apps
Con: Loss of precision in attributes; researchers not familiar with the existing ontologies
Also: mapping at a later stage always possible
Matching of entities Boundaries not always clear, so 1:N mappings are
possible Automatic matching can lead to errors, which must be
fixable manually
JVMG Project Introduction 16
Technological realisation Adapting graph data to a traditional database schema
often results in information loss or performance problems in storing and retrieval
Native graph databases have become more mature in the past years and it is viable for our project to store the data directly into such a database
No own solution, but implementation using an open-source software stack that is adapted from other similar projects
JVMG Project Introduction 17
Technological realisation: Methods Install and evaluation of multiple solutions
Fuseki triple store + Pubby web interface Wikibase software stack (BlazeGraph + web interface) Others?
Data pipelines for each source Download, storage, conversion to RDF (local model)
Central data process Matching entities and merging of properties (global
model)
JVMG Project Introduction 18
Technological realisation: Challenges Uncertain future of open source graph database
development Performance and indexing issues possible Costly commercial solutions not viable
Position at Leipzig University still open
Documenting the full development process Ideally, a new instance of our database can be
generated by “git pull” of our repository
JVMG Project Introduction 20
Stuttgart Media University Zoltán Kacsuk
Media researcher DA in Manga Studies
(Kyoto Seika University)
Luca Bruno Media researcher MA in Japan Studies
(University of Venice ca’ Foscari)
Senan Kiryakos Information Science researcher PhD in Information Science
(University of Tsukuba)
JVMG Project Introduction 21
Timeline Past
2018-01: Project proposal 2018-12: Grant approved 2019-05: Project start 2019-07: First community workshop in Leipzig 2019-08: First data sharing agreed
JVMG Project Introduction 22
Timeline Planned
2020 Q1: Choice of software for prototype 2020 Q2: Access to first prototype Mid 2020?: Follow-up workshop with communities Late 2020?: First workshop with researchers
Project funding ends 2022-06
JVMG Project Introduction 23
Project website https://jvmg.iuk.hdm-stuttgart.de/