Retrieval of Relevant Opinion Sentences for New Products

Park, Dae Hoon; Kim, Hyun Duk; Zhai, ChengXiang; Guo, Lifan

doi:10.1145/2766462.2767748

Cited by 17 publications

(14 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In order to quantitatively evaluate the effectiveness of our proposed method, we compare our model MSRM, against three well-established typical benchmark models: MEAD-SIM [11], ReviewSpecGen [11], and Translation [11]. MEAD-SIM considers both query-relevance and centrality.…”

Section: Experiments Resultsmentioning

confidence: 99%

“…Yu et al [10] used IMDb's structured data for documents categorization. Park et al [11] employed product specifications and reviews to retrieve relevant sentences for new products. Other characteristics such as helpfulness and coverage are ignored.…”

Section: Related Workmentioning

confidence: 99%

“…We leverage Query likelihood retrieval model [16] to analyze relevance, which is proved effective in [11]. We model the relevance of a product with conditional probability p(t, | ), which can be interpreted as the posterior probability that a new or unpopular product is relevant to a review sentence t of other product .…”

Section: Modeling Relevancementioning

confidence: 99%

See 2 more Smart Citations

MSRM: A Novel Model to Retrieve Meaningful Opinion Sentences for New Products

Du¹,

Xu²,

Huang³

2018

dtcse

View full text Add to dashboard Cite

Abstract. Online reviews are more and more important for potential consumers to make purchase decisions. The referred information of the new products or unpopular products that have no reviews or very few reviews is limited and this situation makes it difficult for consumers to obtain enough information to understand these products. Indeed, this is a new issue recently that needs to be addressed. In this paper, we study the problem of automatically retrieving meaningful opinion sentences for a new product or unpopular product from reviews of other similar products. The retrieved meaningful opinion sentences should possess three properties: helpfulness, relevance and coverage. We propose a meaningful sentences retrieval model (MSRM), which is centered on these three properties to extract meaningful opinion sentences for a new product or unpopular product. We employ product specifications to estimate similarity between products, and model helpfulness, relevance and coverage properties respectively, finally incorporate these three properties into MSRM to mine meaningful sentences from reviews of similar products. Through a series of experiments on real data sets, experiment results show MSRM achieves much better performance.

show abstract

Section: Experiments Resultsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

MSRM: A Novel Model to Retrieve Meaningful Opinion Sentences for New Products

Du¹,

Xu²,

Huang³

2018

dtcse

View full text Add to dashboard Cite

show abstract

“…For supervised learning, [7] developed a method using support vector machine (SVM) to automate the review helpfulness evaluation, [8] used entropy-based approach to explore the online review helpfulness and [5] used bilinear approach for classification of Amazon data.…”

Section: B Related Workmentioning

confidence: 99%

An Entropy based Product Ranking Algorithm using Reviews and Q and A Data

Anjum

Sabharwal

2016

International Conferences on Distributed Multimedia Systems

View full text Add to dashboard Cite

-Amazon.com, along with several other commercial websites for products and services, provides a platform for consumers to share their opinions by providing reviews and answering product related questions (QA data). These opinions can be quantitative, qualitative or a combination of both. Owing to the large corpus of such data available, there are several learning and classification approaches available to scrutinize them e.g., those based on Entropy measures, machine learning, stochastic, and natural language processing etc. In this paper, we review some of the prominent techniques and explore a hybrid approach, involving Entropy, Bilinear and statistical measures, to use heterogeneous consumer data and simultaneously analyze and rank products for customers. With experimental results, we show that our approach effectively ranks products using (1) text reviews (2) QA data and (3) star rating of products. We also make a case that the ranks calculated are more relevant to the customers and can enable better prediction on the products sale for the sellers.

show abstract

“…The use of categorical attributes (e.g., user, topic, aspects) in the sentiment analysis community (Kim and Hovy, 2004;Pang and Lee, 2007;Liu, 2012) is widespread. Prior to the deep learning era, these information were used as effective categorical features Tan et al, 2011;Gao et al, 2013;Park et al, 2015) for the machine learning model. Recent work has used them to improve the overall performance (Chen et al, 2016;Dong et al, 2017), interpretability (Amplayo et al, 2018a;Angelidis and Lapata, 2018), and personalization (Ficler and Goldberg, 2017) of neural network models in different tasks such as sentiment classification (Tang et al, 2015), review summarization (Yang et al, 2018a), and text generation (Dong et al, 2017).…”

Section: Introductionmentioning

confidence: 99%

Rethinking Attribute Representation and Injection for Sentiment Classification

Amplayo¹

2019

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conferen

View full text Add to dashboard Cite

Text attributes, such as user and product information in product reviews, have been used to improve the performance of sentiment classification models. The de facto standard method is to incorporate them as additional biases in the attention mechanism, and more performance gains are achieved by extending the model architecture. In this paper, we show that the above method is the least effective way to represent and inject attributes. To demonstrate this hypothesis, unlike previous models with complicated architectures, we limit our base model to a simple BiLSTM with attention classifier, and instead focus on how and where the attributes should be incorporated in the model. We propose to represent attributes as chunk-wise importance weight matrices and consider four locations in the model (i.e., embedding, encoding, attention, classifier) to inject attributes. Experiments show that our proposed method achieves significant improvements over the standard approach and that attention mechanism is the worst location to inject attributes, contradicting prior work. We also outperform the state-of-the-art despite our use of a simple base model. Finally, we show that these representations transfer well to other tasks 1 .

show abstract

Retrieval of Relevant Opinion Sentences for New Products

Cited by 17 publications

References 31 publications

MSRM: A Novel Model to Retrieve Meaningful Opinion Sentences for New Products

MSRM: A Novel Model to Retrieve Meaningful Opinion Sentences for New Products

An Entropy based Product Ranking Algorithm using Reviews and Q and A Data

Rethinking Attribute Representation and Injection for Sentiment Classification

Contact Info

Product

Resources

About