Paper Title: Platform-invariant Topic Modeling via Contrastive Learning to Mitigate Platform-induced Bias

Cross-platform topic dissemination is one of the research subjects that delved into MEDIA ANALYSIS; sometimes it fails to grasp the authentic topics due to platform-induced biases, which may be caused by aggregating documents from multiple platforms and running them on an existing topic model. My recent work, dealing with the impact of unique platform characteristics on the performance of topic models and proposing a new approach to enhance the effectiveness of topic modeling, has been accepted as Findings at EMNLP 2024, one of the top-tier AI conferences in natural language processing (NLP). The devised model reduces platform influence in topic models by developing a platform-invariant contrastive learning algorithm and removing platform-specific jargon word sets. I hope this method mitigates biases arising from platform influences when modeling topics from texts collected across various platforms and helps media researchers better understand general trends across social media.

This work will be presented in November 2024 in Miami, the US.

[Figure] The illustration of the proposed model. The model consists of two main components in order to reduce platform bias in topic modeling: platform-invariant CL (contrastive learning) and filtering jargon for BoW (bag of words) reconstruction.