Text summarization and sentiment classification both aim to capture the mainideas of the text but at different levels. Text summarization is to describethe text within a few sentences, while sentiment classification can be regardedas a special type of summarization which "summarizes" the text into a even moreabstract fashion, i.e., a sentiment class. Based on this idea, we propose ahierarchical end-to-end model for joint learning of text summarization andsentiment classification, where the sentiment classification label is treatedas the further "summarization" of the text summarization output. Hence, thesentiment classification layer is put upon the text summarization layer, and ahierarchical structure is derived. Experimental results on Amazon onlinereviews datasets show that our model achieves better performance than thestrong baseline systems on both abstractive summarization and sentimentclassification.
URL
Affiliations
Abstract
Translation (by gpt-3.5-turbo)
Summary (by gpt-3.5-turbo)