End of February not only saw the second JVMG Lab taking place at Ritsumeikan University, but we also held a JVMG Symposium there subsequently to present and discuss the results of the second phase of the project as well as look towards its future.
Magnus Pfeffer and Martin Roth, the two original creators of the Japanese Visual Media Graph project, opened the event by introducing the three upcoming sessions starting with the spotlight on participating online communities, followed by presentations on the results from the second phase of the project, and closing with a look at future directions and ongoing collaborations.
The first session, showcasing the presentations of the participating online communities, started with Maria Pino from AnimeClick walking us through the many features of the now 27 year old website that has become Italy’s leading anime-manga news source and community portal, boasting around 156.000 registered users. Founded in 1999, it is managed by a non-profit association since 2011, with all contents and features being created and developed by a group of roughly eighty volunteers. AnimeClick now also enjoys media partnerships with Lucca Comics & Games, Napoli Comicon, Far East Film Festival in Udine, and the Japan Matsuri in Bellinzona (CH). Furthermore, they themselves also actively foster their community through regular live streams on Twitch as well as larger events like the online Hanami Manga Festival in 2021 and the two-day Anime Mirai Festival held in 2024. Finally, to promote and celebrate anime-manga culture in Italy they have also created the Animeclick Awards, which are awarded every year since 2018 at Lucca Comics & Games with the participation of the Italian publishers of the nominated works.
Next, Stephen Goral, the founder of Anime Characters Database (ACDB) took to the stage to tell us about many of the new and sometimes impulsively built features of his website that was started in 2007 and which now numbers around 90.000 members with six to ten active moderators helping to keep new data checked and added all the time. The now over 145.000 characters featured on the site are the focus and building blocks of the many fun features and mini-games that make ACDB special. One such mini-activity that users seem to enjoy a lot is building custom family trees from the available characters on the site. Beyond the visual traits recorded for characters as well as character and genre tags the site features many more details. For example, birthdays have proven a very popular topic of interest, adding information on this facet of characters, however, also needs to go through the stringent quality control implemented by ACDB admins, thus, unless an official source can be provided for a character’s date of birth, it will not be added. Another major undertaking of the site is trying to collect quotes with the corresponding soundbites for all the characters, with the grand goal of amassing one million quotes.
Third, Yoran Heling the creator of The Visual Novel Database (VNDB), who is since then tragically no longer with us, explained the many changes and developments the website has seen since he last presented his plans for the upcoming years. One very interesting trend that had already started around 2019, but has only become even more pronounced over the past three years, is the increasing and now overwhelming number of English language visual novels being added to the database, along with a sizable count of other language works, resulting in VNDB no longer being a database of mostly only Japanese language games. Since 2023 he had completely reworked the backend framework of the site, as well as rewrote the frontend in Javascript, which used to be written in Elm. Among the many new developments, there is now a publicly visible changelog documenting all the updates to and new features of the site; a new data field for gender identity has been added to the character data model allowing for a wider range of sex/gender identities to be represented; and high-resolution package scans and artwork are being added to the release pages. Finally, there is now also a sister site to VNDB, VNDB.moe, which has all mature only content filtered out, offering a more safe for work window into the world of visual novels.
The second session, starting off the afternoon, featured presentations on the results from the second phase of the project. The session was moderated by Jan-Noël Thon, whose work on transmedia characters (see 2019, 2022) has proven to be key in addressing the complex challenges we face in the JVMG project when integrating data on characters.
First up, Magnus Pfeffer and Saskia Dreßler provided an overview of the many aspects of infrastructure and ontology development that had received emphasis in the second phase of the project. One key element for the JVMG project is its data-agnostic developmental approach, as we work with heterogeneous data sources, which also change their respective data structures over time. As a result all tools are built to discover features from the RDF data itself, and ontologies, which are managed in Protégé, should contain all the data that is required for the tools to work. A big ontology “spring cleaning” was also conducted to make sure that there are consistent paths and spelling across all sources as well as consistent labels for all entities and properties, with domain and range information provided for all of the latter. The JVMG ontology for the consolidated data was also further refined. The frontend received a new overview page inspired by the DraCor and GOLEM projects. A completely new search system was implemented that now relies on Fuseki‘s internal indexing features only, providing fast and reliable query results that are displayed ranked and according to data facets. A unique query-builder has also been developed by our former colleague, Tobias Malmsheimer, that allows for complex querying of the knowledge graph without any prior knowledge of the data ontologies or the SPARQL query language. As a finishing touch a self-contained Docker instance of the JVMG knowledge graph will be completed this summer allowing for local deployment and querying of the graph.
Next, Martin Hennig and Christopher Zysik highlighted some of the main findings in their ongoing exploration of genre through community data. While traditional genre research is mostly based on qualitative approaches, this undertaking — working with data from TVTropes — aimed to bridge media theory, enthusiast communities and industry approaches. Tropes are recurring patterns related to rhetorical figures, stereotypical role models/narrative elements/production and performance strategies. Although it is commonly accepted knowledge both in and outside of academia that tropes can be and are often connected to genres, the large-scale systematic exploration of this relationship is still an open problem. This work focused on examining the kaiju genre, which has a rich transcultural history and is also one of the foundational pillars of modern popular Japanese visual media. By examining the way tropes are linked to the kaiju genre and kaiju works on TVTropes, we gained insights into the bottom-up approach of genre crystallization, uncovering how genres are actively and collaboratively constructed, negotiated, as well as aesthetically and discursively expanded — one example of this is the way the sixteenth century Carta Marina is tagged as kaiju — through such a collective engagement. The inclusive perspective on genres offered by the dynamics of social tagging also foreground the impact of incomplete data and at times contradicting perspectives, as well as their constant process of transformation as the meaning of tropes are negotiated on forums. Aspects of power asymmetry and the effects of distortion/bias can all be traced in the community data, while at the same time the contours of a strong challenge to an established genre canon become visible in the way the scope and meaning of kaiju is reconfigured through the lens of the interconnections of the network of tropes and genre tags in the TVTropes data. For lots more details and data see their series of excellent Tiny Use Cases (1, 2, 3, 4) on their ongoing exploration.
Third, Bryan Hartzheim, Stevie Suan, and Martin Roth presented some of the most striking results in our pilot study of the anime credits information found in Anime News Network (ANN) conducted to assess the viability of this type of data source for anime research, and laying the foundations for the future integration of more creator credits information in the JVMG knowledge graph expected from future updates to the Media Arts Database (importantly, ANN data itself cannot be integrated into the JVMG due to licence issues). Although creators’ careers can never be fully captured in credits information alone, it nevertheless offers a certain window into the types of roles that creators go through over the course of their professional pathways. By aggregating and mapping the typical role transitions that many thousands of creators’ careers — as reflected in credits information — make visible we gain an unparalleled view of the profession based structure of the Japanese anime industry. This blueprint of the industry can serve as both a starting point for new questions and hypotheses, as well as help re-assess or confirm certain commonly held assumptions about the distribution of particular types of labor in anime production.
Last in this session, Federico Pianzola and Zoltan Kacsuk looked back on the work we conducted in cooperation with the Graphs and Ontologies for Literary Evolution Models (GOLEM) project and the next steps to come in this collaboration. The GOLEM project, led by Federico Pianzola at the University of Groningen, works with various fanfiction corpora and their corresponding metadata to provide a cross-cultural analysis of the development and variance found in fiction, and thereby test some of the tenets of the theory of cultural evolution. Our ongoing dialogue with our colleagues at the GOLEM project has proven influential in our work on the JVMG ontology (see their foundational paper on the GOLEM ontology here). Furthermore, connecting data between the two projects has been one of the key interests for us, exploring how fanfiction metadata can be linked to the JVMG knowledge graph. Creating bridges on the level of works was the first successful attempt at this, and we hope to continue our joint efforts by exploring the way character information in the JVMG knowledge graph can contribute to stratifying fanfiction data for a more structured yet granular analysis.
The final session of the day, dealing with future directions and ongoing collaborations was very fittingly moderated by Jaqueline Berndt, who has not only been an important ongoing supporter of our project from its very inception, but who is also one of the most curious and forward looking scholars in the field of comics and manga studies well-known for fostering connections and collaborations across language domains, disciplines and generations alike.
First, Akiko Sugawa-Shimada and Minori Ishida introduced the work and results of the Women In Japanese Animation (WIJA) project so far. One of the questions that the project has sought to answer has been: how can gender diversity in animation studios impact gender justice in the contents of anime? Although women alone cannot achieve gender justice, but having individuals sensitive to gender inequality issues in the production process in decision-making positions is crucial for more inclusive storylines to become more prevalent. Female scriptwriters and animators were found to incorporate or remove elements to avoid strengthening gender stereotypes as well as to increase inclusivity of representation in anime. Another strand of research pursued as part of the WIJA project has been conducting interview surveys with veteran female professionals working in the anime industry in positions with higher degrees of decision making power. These interviews focus on the career histories of the female creators from how they got interested in this field of work, their different positions along their way, as well as their future aspirations. One of the goals of these interviews is to help describe events and experiences that cannot normally be expressed or discussed in public settings concerning the experiences of female creators working in the Japanese anime industry and documenting these for researchers. These interviews are available online in both Japanese and English on the WIJA project website, for example, here is the English version of the interview with Noriko Matsumoto.
Next, Tetsuya Mihara presented, also on behalf of Kazufumi Fukuda and Hideyuki Ōtsubo (who were also present at the event), about new developments in relation to the Media Art Database (MADB). The only missing member of the MADB “fantastic four” from the list of presenters and the event was Ikki Ōmukai, with the “father of MADB” Shigeo Sugimoto also present at the symposium. After a brief overview of the current state of the manga, anime and video games data models in MADB as well as the MADB Lab initiative, Tetsuya Mihara turned to introducing new developments (such as the MADB dashboard that provides statistics on the available data) and upcoming challenges faced by their project. One of the major innovations coming to MADB hopefully soon is the OCR solution for anime credits information being developed by Hideyuki Ōtsubo. The MADB MCP server is both a challenge and a new development and is an attempt at giving generative AI access to use the data and have it answer questions, which could also help enhance SPARQL queries. Finally, the most important future challenge is the upcoming establishment of the Center of Media Arts, which is planned to be built as a part of the National Film Archive of Japan, which would be a physical location for the archiving of materials (e.g. intermediate production materials). At the same time the Japanese government is motivated to now also track the international activity in relation to Japanese productions, which will hopefully be a driving force towards the development of international data infrastructure for anime/manga/video games.
Martin Roth and Magnus Pfeffer once again took to the stage to finish off the symposium with their look towards the future of the JVMG project. First, and most important of all, the data will be available going forward, the mediagraph.link site will remain available, and once the Docker instance is completed researchers will also have the option to run the whole knowledge graph on their own servers or local machines. Second, they also provided a short demo of the SPARQL builder that will hopefully make the data even more accessible for a wider audience of researchers. Third, the connections that have been established during the project are the grounds for further exchange and developments. The project has been a proof of concept that we can work together in interdisciplinary ways, and hopefully it can continue to attract people working together from various perspectives. Fourth, the way we have been relying on communities to share data with us is becoming a very rare thing indeed, as we are increasingly seeing big corporations purchasing community data platforms. On a brighter note, the Japanese government has also recognized the value of this type of community data, and fortunately for now are working towards opening up this data to a wide public with a very permissive open license via, for example, the MADB, but there are no guarantees for how things will be in the future. Finally, the AI solutions that are helping to better enable our work on the one side, are at the same time also a threat to the very work that we are trying to pursue, thus we also need to defend the data to be able to keep the service running in the future.
The JVMG Symposium was a true celebration of the many results and collaborations that had been formed over the years thanks to this project. Thank you to everyone, not just those present at this event, but really everyone who has shared a part of this journey with us, we truly couldn’t have done it without you! The generous funding provided by the DFG has contributed to more than just the development of new research infrastructure, it has also enabled the forging of a community around it with a shared vision of harnessing community data for large-scale quantitative research. And the work will continue, data will be updated, new data sources will be added, research projects and Tiny Use Cases will continue to be pursued, publications will keep coming out, collaborations will be further developed. And thanks to the generous cooperation of our community partners, the CC BY-NC-SA 4.0 license of the JVMG knowledge graph will enable the further development and even forking of the consolidated data and its future iterations to come. So, instead of farewell, we would like to say: Godspeed!