Modeling online reviews with multi-grain topic models

By Ivan Titov, Ryan T. McDonald. The Web Conference, 2008

838 citations recorded by the source corpus.

In this paper we present a novel framework for extracting the ratable aspects of objects from online user reviews. Extracting such aspects is an important challenge in automatically mining product opinions from the web and in generating opinion-based summaries of user reviews [18, 19, 7, 12, 27, 36, 21]. Our models are based on extensions to standard topic modeling methods such as LDA and PLSA to induce multi-grain topics. We argue that multi-grain models are more appropriate for our task since standard models tend to produce topics that correspond to global properties of objects (e.g., the brand of a product type) rather than the aspects of an object that tend to be rated by a user. The models we present not only extract ratable aspects, but also cluster them into coherent topics, e.g., 'waitress' and 'bartender' are part of the same topic 'staff' for restaurants. This differentiates it from much of the previous work which extracts aspects through term frequency analysis with minimal clustering. We evaluate the multi-grain models both qualitatively and quantitatively to show that they improve significantly upon standard topic models.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/dd43f1bc6fdc2ddb2a568efff5a73afc4e382862

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Building a Sentiment Summarizer for Local Service Reviews (2008, background): “One possibility is to induce some kind of aspect clusters [7, 19], but this will rely on co-occurrence counts being sufficiently accurate.”
SAR: A sentiment-aspect-region model for user preference analysis in geo-tagged reviews (2015, background): “On the other hand, several proposals [13], [18], [19], [7] on opinion mining aim at identifying latent topical-aspects and predicting ratings/sentiments on identified aspects in product reviews.”
Jointly Modeling Aspects and Opinions with a MaxEnt-LDA Hybrid (2010, methodology): “In contrast, unsupervised, knowledge-lean topic modeling approach has been shown to be effective in automatically identifying aspects and their representative words (Titov and McDonald, 2008; Brody and Elhadad, 2010).”
A Joint Model of Text and Aspect Ratings for Sentiment Summarization (2008, methodology): “Following Titov and McDonald (2008) we use a collapsed Gibbs sampling algorithm that was derived for the MG-LDA model based on the Gibbs sampling method proposed for LDA in (Grifﬁths and Steyvers, 2004).”
Topic Modeling using Topics from Many Domains, Lifelong Learning and Big Data (2014, methodology): “Last, since we use product reviews as our data, our work is related to opinion mining using topic models, e.g., (Mei et al., 2007; Titov & McDonald, 2008; Zhao et al., 2010).”
Mining topics in documents: standing on the shoulders of big data (2014, methodology): “Topic models have been used for the task by many researchers [5, 10, 16, 21, 23, 25, 26, 32, 33, 37].”
FLAME: A Probabilistic Model Combining Aspect Based Opinion Mining and Collaborative Filtering (2015, methodology): “Topics learned under this assumption are local topics that preserve sentence-level word concurrences [20], while models like LDA [1] produce global topics that preserve document-level word concurrences.”
Sentiment Analysis in MOOC Discussion Forums: What does it tell us? (2014, background): “If we have enough posts with annotated sentiment and topic, many machine learning approaches could capture the mixture of document topics and sentiments simultaneously and substantially improve the accuracy of opinion tracking [17, 28].”
Survey on Aspect-Level Sentiment Analysis (2016, methodology): “790 hotel [40] reviews Titov & McDonald (2008) [39] product, hotel, and binary sentiment classification Ranking Loss: 0.”
Transformation Networks for Target-Oriented Sentiment Classification (2018, methodology): “The early methods mostly adopted supervised learning approach with extensive hand-coded features (Blair-Goldensohn et al., 2008; Titov and McDonald, 2008; Yu et al., 2011; Jiang et al., 2011; Kiritchenko et al., 2014; Wagner et al., 2014; Vo and Zhang, 2015), and they fail to model the semantic rela”
Fine-grained Opinion Mining with Recurrent Neural Networks and Word Embeddings (2015, methodology): “, 2009) and CRF (Shariaty and Moghaddam, 2011; Yang and Cardie, 2012; Yang and Cardie, 2013), as well as topic modeling techniques using Latent Dirichlet Allocation (LDA) model and its variants (Titov and McDonald, 2008; Lin and He, 2009; Moghaddam and Ester, 2012).”
An Unsupervised Neural Attention Model for Aspect Extraction (2017, background): “In recent years, Latent Dirichlet Allocation (LDA) (Blei et al., 2003) and its variants (Titov and McDonald, 2008; Brody and Elhadad, 2010; Zhao et al., 2010; Mukherjee and Liu, 2012) have become the dominant unsupervised approach for aspect extraction.”
Simultaneously Discovering and Quantifying Risk Types from Textual Risk Disclosures (2014, background): “Another variant (Titov and McDonald 2008, Chang and Chien 2009, Wang et al. 2009, Du et al. 2010, Lin et al. 2011) models the sentence-wide topic proportion in addition to the document-wide topic proportion in original LDA model.”
Dual Graph Convolutional Networks for Aspect-based Sentiment Analysis (2021, methodology): “Earlier methods (Titov and McDonald, 2008; Jiang et al., 2011; Kiritchenko et al., 2014; Vo and Zhang, 2015) are usually based on handcrafted features and fail to”
Latent Dirichlet allocation (LDA) and topic modeling: models, applications, a survey (2017, background): “According to the Table 1, used LDA model for variety subjects, such as: Scientiﬁc topic discovery [35], Source code analysis [27], Opinion Mining [30], Event detection [40], Image Classiﬁcation [41].”
An Unsupervised Aspect-Sentiment Model for Online Reviews (2010, methodology): “This approach overcomes the problems of frequent-term methods, as well as the issues raised by Titov and McDonald (2008b).”