This paper describes our deep learning-based approach to multilingualaspect-based sentiment analysis as part of SemEval 2016 Task 5. We use aconvolutional neural network (CNN) for both aspect extraction and aspect-basedsentiment analysis. We cast aspect extraction as a multi-label classificationproblem, outputting probabilities over aspects parameterized by a threshold. Todetermine the sentiment towards an aspect, we concatenate an aspect vector withevery word embedding and apply a convolution over it. Our constrained system(unconstrained for English) achieves competitive results across all languagesand domains, placing first or second in 5 and 7 out of 11 language-domain pairsfor aspect category detection (slot 1) and sentiment polarity (slot 3)respectively, thereby demonstrating the viability of a deep learning-basedapproach for multilingual aspect-based sentiment analysis.
展开▼
