PERCEIVE Benchmark Revolutionizes Personalized Emotion Analysis on Social Media
A team of researchers has launched PERCEIVE, an innovative large-scale benchmark designed for understanding personalized emotions and communication behaviors on social media in both English and Chinese. This benchmark uniquely combines five essential aspects: content generated by authors, authentic emotional reactions from readers' comments, communication behaviors, user characteristics, and the social graph. By focusing on reader perspectives rather than solely on authors, it effectively captures subjective emotional reactions through actual interactions. PERCEIVE facilitates a significant transition towards a more personalized analysis that reflects varied emotional responses from different readers engaging with the same content.
Key facts
- PERCEIVE is a bilingual (English and Chinese) benchmark
- It integrates five dimensions: author content, reader emotions, communication behavior, user attributes, social graph
- It enables reader-centric emotion analysis
- Emotions are annotated from reader comments
- The benchmark captures subjective emotional responses
- It is the first of its kind to integrate these dimensions
- The research is published on arXiv
- The paper is arXiv:2605.12525v1
Entities
Institutions
- arXiv