We investigate the effect of using sentiment expression boundaries in predicting sentiment polarity in aspect-level sentiment analysis. We manually annotate a freely available English sentiment polarity dataset with these boundaries and carry out a series of experiments which demonstrate that high quality sentiment expressions can boost the performance of polarity classification. Our experiments with neural architectures also show that CNN networks outperform LSTMs on this task and dataset.

We investigate the application of kernel methods to representing both structural and lexical knowledge for predicting polarity of opinions in consumer product review. We introduce any-gram kernels which model lexical information in a significantly faster way than the traditional n-gram features, while capturing all possible orders of n-grams n in a sequence without the need to explicitly present a pre-specified set of such orders. We also present an effective format to represent constituency and dependency structure together with aspect terms and sentiment polarity scores. Furthermore, we modify the traditional tree kernel function to compute the similarity based on word embedding vectors instead of exact string match and present experiments using the new models.