neal young / Le19Unsupervised
The paper considers the problem of summarising a collection of reviews for an item; the application considered is that of reviews for doctors. We consider the case where there are aspects about the item that are captured in the review, and are organized in an ontology in the form of a DAG. There is also sentiment about the aspects, so the reviews can be thought of as aspect/sentiment pairs. The goal is to find a subset of them to cover all the pairs, where coverage takes into account the ontology structure. We propose and analyse algorithms, and perform quantitative and qualitative experiments on real data.
© Copyrights are reserved by the publishers.
Download for personal and limited academic use only.