Yelp Data Research Articles

A New Ontology for Restaurant Review Sentiment Analysis

ABSTRACTUnderstanding the aspects of a restaurant that contribute to people's sentiment is crucial for restaurant owners and marketers to improve customer experiences. A sentiment ontology for restaurant review sentiment analysis was proposed in this study. The proposed ontology identified six aspects of a restaurant evaluation: food, service, ambience, cleanliness, location, and price, based on 1,000 annotated Yelp data. Applying the proposed ontology in the annotated Yelp data showed that it helped improve the sentiment analysis accuracy by 11.2% on average. We believe that this study provides insights into the key factors that contribute to restaurant review sentiments.

The Evaluation of DyHATR Performance for Dynamic Heterogeneous Graphs

Dynamic heterogeneous graphs can represent real-world networks. Predicting links in these graphs is more complicated than in static graphs. Until now, research interest of link prediction has focused on static heterogeneous graphs or dynamically homogeneous graphs. A link prediction technique combining temporal RNN and hierarchical attention has recently emerged, called DyHATR. This method is claimed to be able to work on dynamic heterogeneous graphs by testing them on four publicly available data sets (Twitter, Math-Overflow, Ecomm, and Alibaba). However, after further analysis, it turned out that the four data sets did not meet the criteria of dynamic heterogeneous graphs. In the present work, we evaluated the performance of DyHATR on dynamic heterogeneous graphs. We conducted experiments with DyHATR based on the Yelp data set represented as a dynamic heterogeneous graph consisting of homogeneous subgraphs. The results show that DyHATR can be applied to identify link prediction on dynamic heterogeneous graphs by simultaneously capturing heterogeneous information and evolutionary patterns, and then considering them to carry out link predicition. Compared to the baseline method, the accuracy achieved by DyHATR is competitive, although the results can still be improved.

The Spatial Distribution of Urban Consumer Service Firms: Evidence from Yelp Reviews

A growing literature employs distance-based measures of localization to assess the spatial distribution of firms with a focus on manufacturing across a country. We analyse the spatial concentration of a variety of consumer services firms in the Phoenix, AZ area using geo-referenced Yelp data from over 29,000 establishments. Results from a K-density approach indicate substantial localization and service differentiation among localised firms. Firm concentration varies across service cost and quality; higher quality/cost establishments tend to cluster. Our results further understanding of the modern urban landscape as cities are increasingly centres of consumption.

Deep Learning-Based Imputation Method to Enhance Crowdsourced Data on Online Business Directory Platforms for Improved Services

ABSTRACT Popular online business directory (OBD) platforms, such as Yelp and TripAdvisor, depend on voluntarily user-submitted data about various businesses to assist consumers in finding appropriate options for transactions. Yet the crowdsourced nature of such data restricts the availability of attribute values for many businesses on the platform. Crowdsourced data often suffer serious completeness and timeliness constraints, with negative implications for key stakeholders such as users, businesses, and the platform. We thus develop a novel, deep learning–based imputation method, premised in institutional theory, to estimate missing attribute values of individual businesses on an OBD platform. The proposed method leverages a deep model architecture and considers both inter-business and inter-attribute relationships for imputations. An application to a Yelp data set reveals our method’s greater imputation effectiveness relative to prevalent methods. To illustrate the method’s practical utilities and values, we further examine the efficacy of business recommendations empowered by its imputed business attribute values, in comparison with those enabled by data imputed by benchmark methods. The results affirm that the proposed method substantially outperforms benchmarks for imputing missing attribute values and empowers more effective business recommendations. This study addresses crucial, prominent completeness and timeliness constraints in crowdsourced data on OBD platforms and offers insights for downstream applications that can improve user experiences, firm performance, and platform services.

The influence of relative popularity on negative fake reviews: A case study on restaurant reviews

Fake reviews are rising as online reviews become more prevalent. However, little is known about the mechanism underlying firms’ negative fake reviews. This study draws on signaling theory to analyze the effect of a firm’s relative popularity (vs. its competitors) on receiving negative fake reviews. Based on Yelp data for restaurants in New York City, we show that a restaurant’s relatively higher popularity over its competitors increases the number of negative fake reviews it receives. The positive impact of relative popularity on negative fake reviews is weakened by the restaurant’s reputation and competitors’ popularity. The effect is weaker for chain restaurants than for independent restaurants. This study examines the generation mechanism of negative fake reviews from a competition perspective and emphasizes relative popularity as a motivator behind such reviews. Implications for review system design and restaurant management are provided.

Social networks and consumer behavior: Evidence from Yelp

Estimates of the effect of social networks on consumer behavior may be biased, because social network formation is endogenous. Thus, it is hard to distinguish the impact of a social network from the impact of the factors leading to its formation. I exploit a machine learning technique to systematically search for groups of individuals who are similar to the extent that friendship variation within these groups is unlikely to be correlated with unobservables that affect the outcome. I apply the method using Yelp data to estimate the effect of social network friendship on restaurant choices. I find that compared to non-friends, social network friends are 64 percent more likely to visit the same restaurant. Additional exploration of the heterogeneous effects suggests that the social network effect takes place mostly through the information channel.

Complex causal structures of neighbourhood change: Evidence from a functionalist model and yelp data

In this study, we articulate a functional model of neighbourhood change and continuity, adapted from a classical model proposed by Stinchcombe in 1968. We argue this model provides a relatively simple way to capture key aspects of the complex causal structure of neighbourhood change that are implicit in much neighbourhood change research but rarely formulated explicitly. To evaluate the model, we formulate six testable propositions, which we empirically test with large-scale data from Yelp.com. We illustrate our approach with the case of Toronto, but find broad support for all propositions in an analysis of six cities. A conclusion reflects on the value of incorporating functionalist models into neighbourhood research and policy.

Towards a Model of Urban Evolution Part IV: Evolutionary (Formetic) Distance—An Interpretation of Yelp Review Data

This paper is part IV of “towards a model of urban evolution”. It demonstrates how the Toronto Urban Evolution Model (TUEM) can be used to encode city data, illuminate key features, demonstrate how formetic distance can be used to discover how spatial areas change over time, and identify similar spatial areas within and between cities. The data used in this study are reviews from Yelp. Each review can be interpreted as a formeme where the category of the business is a form, the reviewer is a group and the review is an activity. Yelp data from neighbourhoods in both Toronto and Montreal are encoded. A method for aggregating reviewers into groups with multiple members is introduced. Longitudinal analysis is performed for all Toronto neighbourhoods. Transversal analysis is performed between neighbourhoods within Toronto and between Toronto and Montreal. Similar neighbourhoods are identified validating formetic distance.

Characterizing California licensure status and tobacco user experience with adverse events using Yelp data

Various tobacco vendors, including alternative tobacco product sellers, are listed on the popular crowdsourced business listing platform Yelp. Yelp is used to rate and choose tobacco, electronic nicotine delivery systems (ENDS) goods/services and includes self-reporting of user experiences with shops and products. We cross-referenced California Department of Tax and Fee Administration (CDTFA) licensed tobacco, vape, and head shop retail stores with publicly available Yelp business listings to identify licensed and unlicensed stores in California. We extracted metadata associated with store accounts and analyzed user comments and ratings for discussion of tobacco-related complaints and adverse events. We detected a total of 3,717 shops that were categorized as tobacco/vape/head shops on Yelp and by cross-referencing with CDTFA data, licensed businesses accounted for 49.5% (n = 1,841), licensed individual retailers 31.6% (n = 1,174), and suspected unlicensed storefronts 18.9% (n = 702). Businesses and individuals with a state tobacco retail license received a higher average rating from Yelp users (3.86 out of 5) compared to unlicensed shops (3.57) (p < 0.001). Additionally, 4,682 unique comments about licensed businesses, 1,535 unique comments about individual retailers, and 560 unique comments about unlicensed vendors were reviewed, with themes including discussion about defective and counterfeit products and adverse events including coughing, difficulty breathing and reports of hospitalization detected. In contrast, comments about licensed stores predominantly discussed customer service issues. Close to one-fifth of tobacco, vape and/or head shops reviewed on Yelp were not in CDTFA’s licensure database. Overall self-reported tobacco user experiences appeared to differ in content and severity based on whether an establishment was licensed. These results have the potential to identify unauthorized stores and adverse events associated with their tobacco and vaping products or services.

Personalized multi-faceted trust modeling to determine trust links in social media and its potential for misinformation management

In this paper, we present an approach for predicting trust links between peers in social media, one that is grounded in the artificial intelligence area of multiagent trust modeling. In particular, we propose a data-driven multi-faceted trust modeling which incorporates many distinct features for a comprehensive analysis. We focus on demonstrating how clustering of similar users enables a critical new functionality: supporting more personalized, and thus more accurate predictions for users. Illustrated in a trust-aware item recommendation task, we evaluate the proposed framework in the context of a large Yelp data set. We then discuss how improving the detection of trusted relationships in social media can assist in supporting online users in their battle against the spread of misinformation and rumors, within a social networking environment which has recently exploded in popularity. We conclude with a reflection on a particularly vulnerable user base, older adults, in order to illustrate the value of reasoning about groups of users, looking to some future directions for integrating known preferences with insights gained through data analysis.

Continuous marketing in the social business platform based on identifying consumer boring products

In the social business platform, continuous marketing to consumers can fully explore the consumer purchasing potential. However, since consumers can be influenced by their social friends, their tastes often change, which resulting in the cold start problem of familiar users (CSPFU), and the traditional product recommendation methods are difficult to achieve satisfactory results because they focus on identifying the preferable new products instead of boring familiar products. Therefore, a consumer multi-stage compensation product evaluation model (CMCPEM) based on the multidimensional correlation of products and customers to identify the products that consumers may feel tired is proposed. Specifically, the multidimensional correlation indexes are firstly proposed to depict the preferences of the consumer for the target product to be identified, other consumers who have social contagion and structural equivalence relationships with the consumer and other consumers of homogeneous products. After the direct linear, non-linear and indirect fusion of these multidimensional correlation indexes, the compensation indexes (CIs) are proposed to comprehensively describe the first stage of product evaluation process of consumers. Then, J test in the non-nested model is used to screen out the non-nested CIs that consumers focus on. Finally, in the third stage, the final decision result is given by comprehensively considering CIs that consumers focus on and the indexes that represent consumers’ favorite. Experiment results on YELP data confirm the effectiveness of CMCPEM in successfully launching the continuous marketing campaign.

Abstract 12361: Yelp Data on Fast Food and Exercise Facilities Predict Cardiovascular Events

Introduction: Yelp includes information on nearly all businesses and socioeconomic data in any region of the United States. This database maybe used to obtain socio-economic data that may affect the incidence of cardiovascular disease. This project focuses on access to poor quality food such as fast food and lack of exercise that may have a negative impact in cardiovascular disease. A dataset including the number of fast-food restaurants, non-fast-food restaurants, grocery stores, gyms, other fitness facilities, nursing homes, and pharmacies were aggregated by zip code. In addition, demographic variables such as gender, age, and comorbidities were included. Cardiovascular outcomes were obtained from the Myocardial Infarction Data Acquisition System (MIDAS) database of all hospitalizations in New Jersey. Ten percent of the zip codes that did not have sufficient data were excluded. Hypothesis: Yelp variables will predict the occurrence of stroke, myocardial infarction and heart failure requiring hospitalization. Methods: The yelp data were used to predict three outcomes as identified by the MIDAS database: stroke, myocardial infarction and heart failure requiring hospitalization. Linear models followed by stepwise selection of the predictors were used to define the final models used in these analyses. Results: The results of the regressions in the table show that the number of fast-food restaurants and nursing homes were associated with increased frequency of all 3 outcomes controlling for population size. Access to grocery stores, gyms, and other fitness facilities are associated with lower rate of stroke, myocardial infarction and heart failure requiring hospitalization. Conclusions: Socioeconomic variables obtained from the yelp database predicted the occurrence of stroke, myocardial infarction and heart failure requiring hospitalization with high degree of accuracy.

The location of urban healthcare services: Evidence from Phoenix Yelp reviews

AbstractThe U.S. healthcare sector is a large and important industry. However, relatively little is known about the localization and colocalization patterns of healthcare services within an urban area. By using Yelp data from over 3000 healthcare service establishments in the Phoenix area, this paper applies a recent spatial econometric technique to the analysis of healthcare services. The results reveal that spatial concentration varies across distances and types of healthcare services. Localization is more prevalent at closer distances, and general practitioners tend to be more dispersed than specialists. Colocalization with hospitals is most likely for specialists while family practices and dentistry services tend to locate away from hospitals. Different types of location patterns across different types of healthcare services are relevant for informing policies addressing healthcare access and competition in healthcare markets.

Socioeconomic data predict adverse cardiovascular outcomes

Abstract Background/Introduction Socioeconomic status (SES) factors predict the occurrence of adverse cardiovascular (CV) outcomes. Purpose SES has been studied in many contexts. We present data on SES in a new context using yelp, an American public company that publishes crowd-sourced reviews about businesses. A dataset containing aggregated yelp data for New Jersey zip codes pertaining to the number of fast-food restaurants, grocery stores, gyms and other fitness centers, nursing homes, and pharmacies was created. Cardiovascular outcomes were obtained from the Myocardial Infarction Data Acquisition System (MIDAS). Methods Linear regression models of the SES factors were used to predict the occurrence of hospitalized stroke, myocardial infarction (MI), and heart failure (HF). Results Overall, the number of fast-food restaurants, grocery stores, nursing homes, and pharmacies were associated with increased rate of adverse CV outcomes, while gyms and other fitness centers were associated with improved outcomes. Also, the SES factors of grocery, nursing home, pharmacy and fast-food restaurants were associated with increased rate of MI (p&amp;lt;0.0001). In addition, the SES factors of grocery, nursing home, pharmacy and fast food were associated with increased rate of HF (p&amp;lt;0.0001). On the contrary, gyms and other fitness centers were associated with lower rate of CV outcomes (p&amp;lt;0.0001). With respect to stroke, the SES factors of grocery, nursing home, pharmacy and fast-food restaurants were associated with increased rate of this condition (p&amp;lt;0.0001, figure). The y axis of the figure is proportional to the square root of the number of strokes in each zip code per 10,000 people. Conclusions In conclusion, persons living in zip codes of low socioeconomic status are associated with a high rate of occurrence of MI, HF and stroke. The number of gyms and other fitness centers is associated with a low rate of these outcomes. This information may be important in designing and implementing preventive strategies. Funding Acknowledgement Type of funding sources: None.

Collaborative Filtering With Awareness of Social Networks

In this article, we present the NetRec method to leverage the social network data of users in collaborative filtering. We formulate two new network-related terms and obtain convex optimization problems that incorporate assumptions regarding users’ social connections and preferences about products. Our theory demonstrates that this procedure leads to a sharper error bound than before, as long as the observed social network is well structured. We point out that the larger the noise magnitude in the observed user preferences, the larger the reduction in the magnitude of the error bound. Moreover, our theory shows that the combination of the network-related term and the previously used term of nuclear norm gives estimates better than those achieved by any of them alone. We provide an algorithm to solve the new optimization problem and prove that it is guaranteed to find a global optimum. Both simulations and real data experiments are carried out to validate our theoretical findings. The application of the NetRec method on the Yelp data demonstrate its superiority over a state-of-the-art social recommendation method.

An interpretable framework for investigating the neighborhood effect in POI recommendation.

Geographical characteristics have been proven to be effective in improving the quality of point-of-interest (POI) recommendation. However, existing works on POI recommendation focus on cost (time or money) of travel for a user. An important geographical aspect that has not been studied adequately is the neighborhood effect, which captures a user's POI visiting behavior based on the user's preference not only to a POI, but also to the POI's neighborhood. To provide an interpretable framework to fully study the neighborhood effect, first, we develop different sets of insightful features, representing different aspects of neighborhood effect. We employ a Yelp data set to evaluate how different aspects of the neighborhood effect affect a user's POI visiting behavior. Second, we propose a deep learning-based recommendation framework that exploits the neighborhood effect. Experimental results show that our approach is more effective than two state-of-the-art matrix factorization-based POI recommendation techniques.

What Yelp Fake Review Filter Might Be Doing?

Online reviews have become a valuable resource for decision making. However, its usefulness brings forth a curse ‒ deceptive opinion spam. In recent years, fake review detection has attracted significant attention. However, most review sites still do not publicly filter fake reviews. Yelp is an exception which has been filtering reviews over the past few years. However, Yelp’s algorithm is trade secret. In this work, we attempt to find out what Yelp might be doing by analyzing its filtered reviews. The results will be useful to other review hosting sites in their filtering effort. There are two main approaches to filtering: supervised and unsupervised learning. In terms of features used, there are also roughly two types: linguistic features and behavioral features. In this work, we will take a supervised approach as we can make use of Yelp’s filtered reviews for training. Existing approaches based on supervised learning are all based on pseudo fake reviews rather than fake reviews filtered by a commercial Web site. Recently, supervised learning using linguistic n-gram features has been shown to perform extremely well (attaining around 90% accuracy) in detecting crowdsourced fake reviews generated using Amazon Mechanical Turk (AMT). We put these existing research methods to the test and evaluate performance on the real-life Yelp data. To our surprise, the behavioral features perform very well, but the linguistic features are not as effective. To investigate, a novel information theoretic analysis is proposed to uncover the precise psycholinguistic difference between AMT reviews and Yelp reviews (crowdsourced vs. commercial fake reviews). We find something quite interesting. This analysis and experimental results allow us to postulate that Yelp’s filtering is reasonable and its filtering algorithm seems to be correlated with abnormal spamming behaviors.

Investigating negative reviews and detecting negative influencers in Yelp through a multi-dimensional social network based model

In this paper, we propose an investigation of negative reviews and define the profile of negative influencers in Yelp. The methodology adopted to achieve this goal consists of two phases. The first one is theoretical and aims at defining a multi-dimensional social network based model of Yelp, three stereotypes of Yelp users, and a network based model to represent negative reviewers and their relationships. The second phase is experimental and consists in the definition of five hypotheses on negative reviews and reviewers in Yelp and their verification through an extensive data analysis campaign. This was performed on Yelp data represented by means of the models introduced during the first phase. Its most important result is the construction of the profile of negative influencers in Yelp. The main novelties of this paper are: (i) the definition of the two social network based models of Yelp and its users; (ii) the definition of three stereotypes of Yelp users and their characteristics; (iii) the construction of the profile of negative influencers in Yelp.

An Information Recommendation Technique Based on Influence and Activeness of Users in Social Networks

Over the past few years, the number of users of social network services has been exponentially increasing and it is now a natural source of data that can be used by recommendation systems to provide important services to humans by analyzing applicable data and providing personalized information to users. In this paper, we propose an information recommendation technique that enables smart recommendations based on two specific types of analysis on user behaviors, such as the user influence and user activity. The components to measure the user influence and user activity are identified. The accuracy of the information recommendation is verified using Yelp data and shows significantly promising results that could create smarter information recommendation systems.

Big Data Aspect-Based Opinion Mining Using the SLDA and HME-LDA Models

In order to make better use of massive network comment data for decision-making support of customers and merchants in the big data era, this paper proposes two unsupervised optimized LDA (Latent Dirichlet Allocation) models, namely, SLDA (SentiWordNet WordNet-Latent Dirichlet Allocation) and HME-LDA (Hierarchical Clustering MaxEnt-Latent Dirichlet Allocation), for aspect-based opinion mining. One scheme of each of two optimized models, which both use seed words as topic words and construct the inverted index, is designed to enhance the readability of experiment results. Meanwhile, based on the LDA topic model, we introduce new indicator variables to refine the classification of topics and try to classify the opinion target words and the sentiment opinion words by two different schemes. For better classification effect, the similarity between words and seed words is calculated in two ways to offset the fixed parameters in the standard LDA. In addition, based on the SemEval2016ABSA data set and the Yelp data set, we design comparative experiments with training sets of different sizes and different seed words, which prove that the SLDA and the HME-LDA have better performance on the accuracy, recall value, and harmonic value with unannotated training sets.