Maximilian Schich | Meaning Spaces in Cultural Data Analytics
Вставка
- Опубліковано 26 вер 2024
- Talk kindly contributed by Maximilian Schich in SEMF's 2022 Spacious Spatiality
semf.org.es/sp...
TALK ABSTRACT
Cultural data analysis is an increasingly systematic multidisciplinary science that combines qualitative, quantitative, computational, and aesthetic expertise. While starting from traditional substance, such as images, objects, audiovisual media, music, or language, both topological and geometric multidimensional meaning spaces have begun to take center stage as subjects within the field. Topological aspects have gained attention due to the growing availability of large cultural knowledge graphs and the rise of multidisciplinary network science over the last two decades. Geometric aspects have become conspicuous more recently, not least due to advances in machine learning, where so-called latent embedding spaces are constituted between stimulus and response. Looking at relevant theory, we will close the circle to Peter Gärdenfors, the first speaker in this conference, while further highlighting useful connections to Eigen, Cassirer, and Leibniz. Focusing on practical results, I will present a collaboration that disambiguates polymorphic visual family resemblance via the harnessing of a fully explainable multidimensional embedding space, to make sense of art history over centuries and near-real-time contemporary art market dynamics. The aim of the talk is to spark a discussion, including questions such as: How are cultural meaning spaces constituted? How can we capture the structure and dynamics of such spaces? How do given datasets occupy such spaces? How are such spaces negotiated by cultural agents? And what is their relation to "reality"?
TALK MATERIALS
· Cultural Analysis Situs: archiv.ub.uni-...
MAXIMILIAN SCHICH
CUDAN Open Lab, Tallinn University: cudan.tlu.ee/
Tallinn University profile: cudan.tlu.ee/t...
Personal website: www.schich.info/
Google Scholar: scholar.google...
LinkedIn: / maximilianschich
Twitter: / schichmax
SEMF NETWORKS
Website: semf.org.es
Twitter: / semf_nexus
LinkedIn: / semf-nexus
Instagram: / semf.nexus
Facebook: / semf.nexus
MUSIC
/ baroquenoise
I really wish the visuals of this were clearer. I cannot resolve many of the graphs and figures due to the blurry text and relatively small size. I personally don't like the channel specific frame used here, (nor do I really like it for any content as it effectively shrinks my screen for no good reason), I think the video quality would be substantially better if figures could be resolved by viewers here.
Thank you very much for the feedback. We are working to get higher-resolution uploads in the future.
Interesting work, but bad title. The neural network only analyzed the visual properties of the works and some of their related data (author, date, style, selling price, etc...). At no point where their meanings analyzed.
That's a fair point, what would be a further analysis of their meanings?
@@SEMF Perhaps a natural language model that can be trained on written texts regarding these artworks, correlate that with the visual properties of the paintings, and then use that information in order to classify any artwork it comes across, including ones for which no textual data was provided. For example, someone could ask it to interpret the meaning of a painting it never saw before and the AI would then output a text essay describing what it thinks the artwork symbolizes.
@@SandroAerogen Great comment! Would you apply to to the same data set?
@@SEMF Sure thing. I suppose that there would a be a lot of interesting things to get even from the data the model was trained on. I'm sure a data scientist could have a field day trying to create a sort of "semantic tree" and trying to map out the history of the different symbolisms present in the history of the artworks.
@@SandroAerogen Sounds very cool! Perhaps you would like to share your thoughts with our online community: you can join following this link semf.org.es/participate/join.html