Sending or not? A multimodal framework for Danmaku comment prediction

作者：

Highlights：

• We propose a multimodal framework to predict the behavior of users sending Danmaku comments.

• We offer personalized analysis of multimodal data to precisely capture users’ points of interest.

• Our model combines multimodal fusion part and attention mechanism to ensure high accuracy.

• Our research contributes to understanding the emerging behavior of users watching online videos.

摘要

•We propose a multimodal framework to predict the behavior of users sending Danmaku comments.•We offer personalized analysis of multimodal data to precisely capture users’ points of interest.•Our model combines multimodal fusion part and attention mechanism to ensure high accuracy.•Our research contributes to understanding the emerging behavior of users watching online videos.

论文关键词：Live streaming commerce,Danmaku comment,Behavior prediction,Multimodal learning,Deep neural network

论文评审过程：Received 9 February 2021, Revised 4 May 2021, Accepted 5 July 2021, Available online 27 July 2021, Version of Record 27 July 2021.

论文官网地址：https://doi.org/10.1016/j.ipm.2021.102687