Relevance & Research Question: Social media platforms have evolved into significant ecosystems, where content creators derive income from their digital presence. Despite its prominence as one of the biggest social media platforms, TikTok remains understudied in social science literature, primarily due to the methodological challenges in analyzing video content.

Methods & Data: This paper presents a novel computational approach to evaluate content creators' performance through systematic video data analysis. By embedding the videos via VideoMAE, a state of the art model for embedding of visual data,I examine the performance variations across different content categories and conduct a comparative analysis of creator performance stratified by gender. Using a unique data set of 36,166 videos that have been sampled using a rigorous hourly sampling approach to access content performance over a 41 day period.

Added Value : My methodology demonstrates the feasibility of video content analysis on TikTok, contributing to both the theoretical understanding of digital creator economies and the methodological toolkit for social media research. It also demonstrates how embedding models can be leveraged for social scientific studies of visual data, which is still scarce in the field.

Results: I find differences in the distrubtion of content types across genders. The choice of content type is also related to the sucess of generating user engagement (views / likes). However, individual level intercepts also strongly predict performance of shared content.

Relevance & Research Question

Languages worldwide differ significantly in how they incorporate gender into grammar and phonetics. In the German language, the generic masculine form (e.g., saying “Lehrer” [teacher, male, sing.]) is used to refer to a group of people with unknown (or non-male) sex and has been criticized for rendering women and non-binary people invisible in language, thereby reinforcing gender biases and unequal power dynamics. Gender-inclusive language (GIL) has been proposed as an alternative to the generic masculine and involves various subtypes. Our study investigates the development of GIL on Twitter between 2018 and 2023. In addition, we study individual (gender) and contextual (regional) effects on the use of GIL.
Methods & Data

We rely on a unique dataset of over 1 Billion German language Tweets. We present a pipeline to detect three types of GIL, namely binary feminization, non-gendered GIL and non-binary inclusive language. We do this through a combination of using a fine-tuned German BERT model, regular expressions, and a corpus of German gender-inclusive language words. User names are analyzed based on lists of male, female and unisex names. By inferring the place of residence for the users of more than 300 million Tweets, we shed light on the correlations of socio-structural variables and use of gender-inclusive language across Germany.

Results

We find that GIL adoption increases slightly over the studied 5 year period and we identify different trends among GIL types in this adoption. Furthermore, profiles with female usernames use GIL more often than those with masculine or unisex usernames. In addition, we find regional patterns with more use of GIL in urban regions and regions with a higher share of users with young population.
Added Value

Our study makes several novel contributions to the understanding of gender-inclusive language adoption and digital socio-linguistics. First, it provides insights into the real-world uptake of gender-inclusive language through the largest-scale analysis of German social media communication to date. Second, by linking language use to regional socio-structural variables, we offer the first comprehensive geographic analysis of gender-inclusive language adoption patterns in Germany.

Relevance & Research Question

Online communities are widely acknowledged to provide new opportunities for meaningful interaction between individuals with similar tastes in cultural consumption. However, the flip side of this coin – that online communities likewise provide new social connection opportunities for producers of cultural content – has received much less attention so far.

This project, therefore, starts with the premise that cultural producers foster “competitive co-operation” among themselves for mutual benefits through collaboration. We attempt to explore the specific dynamics of how such collaborations fosters growth and success of individual producers as well as of their community as a whole.

Methods & Data

To investigate these and related claims empirically, we trace the development of an online community of cultural producers creating video content on a popular Japanese manga on YouTube. We develop a new interactive interface allowing for the systematic coding and preprocessing of Youtube data. The coding interface allows for the coding of collaboration cues: words and phrases indicating collaborations (e.g. “feat”, “w”, “@”). With this we construct a longitudinal collaboration network where cultural producers of this specific community share a directed tie if they have jointly published a video on YouTube, with the host as a source and the guest as destination.

Results

The results of negative binomial regression estimating view count indicate great benefits of being invited by other youtubers. A tie to a new host, i.e. an increase in indegree, heightens the yearly viewer yield of a youtuber by around 20%. Inviting new guests, i.e. an increase in outdegree, also increases the yearly viewer yield of a youtuber, however only by around 7%.
Added Value

This study contributes to the study of online producers in two way. First we show a method how to code and preprocess youtube data. Second, we explore specific effects taking place in a cultural producer community pointing out very different effects of in- and outdegree. Both give insight into the dynamics behind emergence of communities online.

Relevance & Research Question

Tremendous advances have been made in measuring the meaning of text data allowing us to quantify entire discourses within unified data structures such as a vector spaces or networks. What is often lost in the application of such methods is the perspectives that different types sources or actors may contribute to a discourse and how such perspectives may come to shape it. We demonstrate a method for evaluating the different positions text sources can have in co-occurrence networks, what we call the sources semantic position. We then create a typology of the networks that emerge based on the semantic position of the actors.

Methods & Data

We use online data from a sample of leading US sources collected with the NewsVibe platform, which allows access to web news articles and facebook posts based on search queries. We sample articles around the 2024 US election. The co-occurrence network for each source is computed first, and then all keyword networks are added to a global co-occurrence network. This allows us to determine the position of each source within the discourse according to the keyword. Finally we evaluate the different types of networks that emerge from such an analysis.

Results

We show that depending on the search query, networks as well as source contribution can be highly contextual. Discourses around “economics”, but also discourses around candidates like Trump and Kamala are highly polarized. The sources in each of these cases split into almost separate parts of the network displaying what one could call semantic polarization. However networks around topics such as NATO are governed by strong influence by a few actors and very little participation by other sources.

Added Value
We further the study of semantic networks by demonstrating how to measure the individual sources' position to a discourse. By keeping the computations rather simple, non-expert third parties can also evaluate discourses by themselves, improving transparency and access to complex interpretation of data around meaningful online issues and actors.