Clustering for Simultaneous Extraction of Aspects and Features from Reviews

Abstract

This paper presents a clustering approach that simultaneously identifies product features and groups them into aspect categories from online reviews. Unlike prior approaches that first extract features and then group them into categories, the proposed approach combines feature and aspect discovery instead of chaining them. In addition, prior work on feature extraction tends to require seed terms and focus on identifying explicit features, while the proposed approach extracts both explicit and implicit features, and does not require seed terms. We evaluate this approach on reviews from three domains. The results show that it outperforms several state-of-the-art methods on both tasks across all three domains.

Is hybrid AI suited for hybrid threats? Insights from social media analysis

“…Similar approaches are developed by adopting supervised sequence labelling. Thus, Hidden Markov models and conditional random fields are used by Chen and colleagues to extract aspect and polarity from social data [45]. Although approaches above show promising results, opinion mining techniques making use of machine learning become problematic for social data exploration, which involves several different domains, multi languages and distinct text types, because models have to be trained for each one, and large sets of training data are required to achieve good results.…”

Use cases for social data analysis with URREF criteria

“…Similar approaches are developed by adopting supervised sequence labelling. Thus, Hidden Markov models and conditional random fields are used by Chen and colleagues to extract aspect and polarity from social data [16]. To alleviate the need to use large amounts of labelled data sets for training purposes, unsupervised methods based on topic modeling [17] or ontologies [18] were also developed.…”

Trend analysis in online data with unsupervised classification and appraisal categories

“…Similar approaches are developed by adopting supervised sequence labelling. Thus, Hidden Markov models and conditional random fields are used by Chen and colleagues to extract aspect and polarity from social data [29]. To alleviate the need to large amount of labelled date for training purposes, unsupervised methods based on topics models [30] or ontologies [26] were also developed.…”