HotComment: A Benchmark for Evaluating Online Comment Popularity
A team of researchers has launched HotComment, a new multimodal benchmark designed to assess the popularity of online comments. This benchmark combines video and text elements to measure popularity through three key dimensions: the quality of content, prediction of popularity, and simulation of user behavior. Content quality is evaluated by comparing semantic similarity to authentic human comments and four clear dimensions. Popularity predictions rely on models developed from actual interaction data. Additionally, user behavior simulation models the distribution of users across platforms to estimate engagement. This initiative tackles the issue that comment popularity is influenced by linguistic quality, originality, and emotional impact, while also differing among various platforms and user demographics.
Key facts
- HotComment is a multimodal benchmark integrating video and text modalities.
- It evaluates popularity from three aspects: content quality, popularity prediction, and user behavior simulation.
- Content quality assessment includes semantic similarity and four interpretable dimensions.
- Popularity prediction uses models trained on real-world interaction data.
- User behavior simulation models platform user distribution to approximate engagement.
- The benchmark addresses challenges of varying stylistic preferences across platforms and user groups.
- The work is published on arXiv with ID 2604.25614.
- The announcement type is new.
Entities
Institutions
- arXiv