On the Use of Modality-Specific Large-Scale Pre-Trained Encoders for Multimodal Sentiment Analysis.
Atsushi AndoRyo MasumuraAkihiko TakashimaSatoshi SuzukiNaoki MakishimaKeita SuzukiTakafumi MoriyaTakanori AshiharaHiroshi SatoPublished in: SLT (2022)
Keyphrases
- sentiment analysis
- pre trained
- opinion mining
- multi modal
- sentiment classification
- text classification
- multi aspect
- medical images
- social media content
- sentence level
- product reviews
- text mining
- training data
- user generated
- knowledge discovery
- public opinion
- natural language processing
- domain dependent
- machine learning
- sentiment lexicon
- data sets