2018/04/23 by Christopher Mitcheltree, Mitcheltree, Christopher, Wharton, Skyler +2
Computer Science · #Computation and Language (cs.CL) #FOS: Computer and information sciences #Sentiment Analysis and Opinion Mining #Topic Modeling #Web Data Mining and Analysis
paper · pdf · doi:10.48550/arxiv.1804.08666
openalex publication_date 2018/04/23 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Reviews of products or services on Internet marketplace websites contain a\nrich amount of information. Users often wish to survey reviews or review\nsnippets from the perspective of a certain aspect, which has resulted in a\nlarge body of work on aspect identification and extraction from such corpora.\nIn this work, we evaluate a newly-proposed neural model for aspect extraction\non two practical tasks. The first is to extract canonical sentences of various\naspects from reviews, and is judged by human evaluators against alternatives. A\nk-means baseline does remarkably well in this setting. The second experiment\nfocuses on the suitability of the recovered aspect distributions to represent\nusers by the reviews they have written. Through a set of review reranking\nexperiments, we find that aspect-based profiles can largely capture notions of\nuser preferences, by showing that divergent users generate markedly different\nreview rankings.\n