May 1, 2017
Members of the MultiComp had 3 papers (2 oral and 1 poster) at Association for Computational Linguistics (ACL 2017, https://acl2017.org) in Vancouver, Canada.
One of the papers “Context-Dependent Sentiment Analysis in User-Generated Videos” (Poria et. al.) analyses sentiment based on a context window in videos.
Second, “Affect-LM: A Neural Language Model for Customizable Affective Text Generation” (Ghosh et. al.) generates text with different degrees of emotional content which is conditioned on an additional design parameter.
And third, “Combating Human Trafficking with Deep Multimodal Models” (Tong et. al.) discusses an approach to detect human trafficking in advertisements using text and images. This paper also presents a dataset Trafficking-10k which consists of 10k annotated samples.
For EMNLP’17 we had “Tensor Fusion Network for Multimodal Sentiment Analysis” which looks at intermodality and intramodality dynamics for sentiment analysis.