Building a Sentiment Summarizer for Local Service Reviews

Auteur-es

Blair-Goldensohn, Sasha; Hannan, Kerry; McDonald, Ryan; Neylon, Tyler; Reis, George; Reynar, Jeff

Nombre Auteurs

5

Titre

Building a Sentiment Summarizer for Local Service Reviews

Année de publication

2008

Référence (APA)

Blair-Goldensohn, S., Hannan, K., McDonald, R., Neylon, T., Reis, G., & Reynar, J. (2008). Building a Sentiment Summarizer for Local Service Reviews. WWW Workshop on NLP Challenges in the Information Explosion Era (NLPIX). https://storage.googleapis.com/pub-tools-public-publication-data/pdf/34368.pdf

Mots-clés

ND

URL

https://storage.googleapis.com/pub-tools-public-publication-data/pdf/34368.pdf

Accessibilité de l'article

Open access

Champ

Natural Language Processing

Type contenu (théorique Applicative méthodologique)

Applicative

Méthode

In this paper, we present a system that summarizes the sentiment of reviews for a local service such as a restaurant or hotel.
In particular we focus on aspect-based summarization models [8], where a summary is built by extracting relevant aspects of a service, such as service or value, aggregating the sentiment per aspect, and selecting aspect-relevant text.
The model we employ for sentiment classification is a hybrid that uses both lexicon-based and machine learning algorithms.
The model we employ for aspect extraction is a hybrid but this time we combine a dynamic aspect extractor, where aspects are determined from the text of the review alone, and a static extractor, where aspects are pre-defined and extraction classifiers trained on a set of labeled data.

Cas d'usage

Online customer reviews

Objectifs de l'article

Our goal is to create a general system that can handle all services with sufficient accuracy to be of utility to users.
In this paper, we present a system that summarizes the sentiment of reviews for a local service such as a restaurant or hotel. In particular we focus on aspect-based summarization models [8], where a summary is built by extracting relevant aspects of a service, such as service or value, aggregating the sentiment per aspect, and selecting aspect-relevant text.

Question(s) de recherche/Hypothèses/conclusion

Research question(s) : The ability to analyze a set of online reviews and produce an easy to digest summary is a major challenge for online merchants, review aggregators and local search services. In this study, we look at the problem of aspect-based sentiment summarization.

Hypothesis(es) : Central to our system is the ability to exploit different sources of information when available. In particular, we show how user provided document level sentiment can aid in the prediction of sentiment on the phrase/sentence level through a variety of models. Furthermore, we argue that the service domain has specific characteristics that can be exploited in order to improve both quality and coverage of generated summaries. This includes the observation that nearly all services share basic aspects with one another and that a large number of queries for online reviews pertain only to a small number of service types.

Conclusion(s) : The resulting system is highly precise for frequently queried services, yet also sufficiently general to produce quality summaries for all service types. The main technical contributions include new sentiment models that leverage context and user-provided labels to improve sentence level classification as well as a hybrid aspect extractor and summarizer that combines supervised and unsupervised methods to improve accuracy.

Cadre théorique/Auteur.es

Aspect-based sentiment summarization (Carenini, Ng, et Pauls, 2006 ; Gamon et al, 2005 ; Hu et Liu, 2004 ; Popescu et Etzioni, 2005 ; Zhuang, Jing, et Zhu, 2006)

Lexicon-based sentiment analysis (Hu et Liu, 2004 ; Turney, 2002 ; Wiebe, 2000)

Machine-learning for sentiment analysis (Choi et al., 2005 ; Dredze, Blitzer, et Pereira, 2007 ; Mao et Lebanon, 2006 ; McDonald et al., 2007 ; B. Pang, Lee, et Vaithyanathan, 2002 ; Snyder et Barzilay, 2007)

Summarizing sentiment (Beineke et al., 2003)

Summarizing sentiment by extracting and aggregating sentiment over ratable aspects (Gamon et al., 2005 ; Hu et Liu, 2004 ; Hu et Liu, 2004 ; Carenini, Ng, et Paul, 2006 ; Carenini, Ng, et Zwart, 2005 ; Popescu et Etzioni, 2005)

Concepts clés

Sentiment analysis

Aspect Ratings

Données collectées (type source)

Dataset of local business reviews collected on Google Maps (maps.google.com)

Définition des émotions

No definition

Negative, neutral, positive labeling

Ampleur expérimentation (volume de comptes)

The amount of review input data varies, but each example has a minimum of 16 input reviews (56 total sentences) of input text.

Department Store (43 Reviews)
Greek Restaurant (85 Reviews)
Children’s Barber Shop (16 Reviews)
Hotel/Casino (46 Reviews)

Technologies associées

Aspect-based summarization models

Mention de l'éthique

ND

Finalité communicationnelle

In the future we plan to adapt the system to products, which is a domain that has been well studied in the past. Just as in services, we believe that hybrid models can improve system performance since there again exists a pattern that a few products account for most review queries (e.g., electronics). Additionally, there is a set of aspects that is common across most products, such as customer service, warranty, and value, which can be utilized to improve the performance for less queried products.

Résumé

Online user reviews are increasingly becoming the de-facto standard for measuring the quality of electronics, restau- rants, merchants, etc. The sheer volume of online reviews makes it dicult for a human to process and extract all meaningful information in order to make an educated pur- chase. As a result, there has been a trend toward systems that can automatically summarize opinions from a set of re- views and display them in an easy to process manner (1, 9). In this paper, we present a system that summarizes the sen- timent of reviews for a local service such as a restaurant or hotel. In particular we focus on aspect-based summarization models (8), where a summary is built by extracting relevant aspects of a service, such as service or value, aggregating the sentiment per aspect, and selecting aspect-relevant text. We describe the details of both the aspect extraction and sentiment detection modules of our system. A novel aspect of these models is that they exploit user provided labels and domain specic characteristics of service reviews to increase quality.