Facilitating apps recommendation in Google Play

DOIhttps://doi.org/10.1108/EL-05-2017-0119
Pages856-874
Published date01 October 2018
Date01 October 2018
AuthorFan Wu,Yung-Ting Chuang,Hung-Wei Lai
Subject MatterInformation & knowledge management,Information & communications technology,Internet
Facilitating apps recommendation
in Google Play
Fan Wu,Yung-Ting Chuang and Hung-Wei Lai
Department of Information Management, National Chung Cheng University,
Ming-Hsiung, Chia-Yi, Taiwan
Abstract
Purpose The purpose of this paper is to present a system that analyzes trustworthiness and ranks
applicationsto improve the search experience.
Design/methodology/approach The system adopts pointwise mutual information to calculate
comment semantics. It examines subjective (signed opinions, anonymous opinions and star ratings) and
objective factors (download numbers, reputation ratings) before ltering, ranking and displaying). The
authors invited three experts to check three categories and compared the results usingSpearman and two
statistics.
Findings A high correlation between the proposedsystem and the expert ranking system suggests that
the system can act as decisionsupport.
Research limitations/implications First, the authors have only tested the correlation between the
proposed system and an expert ranking system; user satisfaction was not evaluated. The authors plan to
conduct a latersurvey to gather user feedback. Second, the ranking systemevaluates applications using xed
weights and disregardstime. Therefore, in the future, the authors plan to enable theirsystem to weight recent
recordsover older ones.
Practical implications User discussion forums,although helpful, have drawbacks. Not all reviews are
trustworthy, and forums provide no ltering mechanisms to combat information overload. The solution to
this is the authorssystem that crawls a forum, lters information, analyzes the trustworthiness of each
commentand ranks the application for the user.
Originality/value This paper develops a formula to analyze thetrustworthiness of opinions, enabling
the system to act as decisionsupport when no professional advice is available.
Keywords Semantics, Content analysis, Mobile applications, Social networks, Sentiment analysis,
Automated operations
Paper type Research paper
Introduction
The development of applications for mobile devices (apps) has exploded in recent years.
Practical and convenient, many apps have been installed on smartphones, allowing these
devices to compete with desktop computers and attract more users. Beyond the high-tech
hardware, smartphones derivemost of their value from their versatile apps. The vendors of
mobile phone operating systems have constructed their own app, such as Google Play and
the App Store, to provideaccess to millions of apps to their users.
As there are numerous apps which do similar things, users may have trouble choosing
the best app for their needs. Some users may learn about the features of these apps from
advertisements or word of mouth. Others may follow the advice of professionals who post
comments online before they makea decision to buy an application. To facilitate the sharing
This research is supported by MOST 104-2410-H-194-090-MY2 and MOST 106-2410-H-194-027-MY2
of the Ministry of Science and Technology, Taiwan.
EL
36,5
856
Received29 May 2017
Revised30 September 2017
21October 2017
Accepted9 December 2017
TheElectronic Library
Vol.36 No. 5, 2018
pp. 856-874
© Emerald Publishing Limited
0264-0473
DOI 10.1108/EL-05-2017-0119
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/0264-0473.htm
of opinions, operating system vendors provide platforms from which users can share their
experiences with eachother and use this information to recommend their apps to users. Such
consultation with reviews and experiences is popular online, especially for apps that cost
money. Freedman (2011) and Dierkes et al. (2011) indicated that the reviews generated by
previous or experienced users have a signicantimpact on shopperspurchasing intentions
and decisions.
These online forums for user discussion and recommendations have some drawbacks.
Not all reviews are trustworthy. The forum does not lter out excess information, as users
still need to browse all reviews, searching for the most trustworthy and relevant
recommendations.
To resolve the above problems, the authors present a system which crawls a forum,
analyzes the trustworthiness of each comment and ranks them according to that quality.
Through this re-ranking, users can save time and nd the appropriateapp more easily. The
system adopts pointwisemutual information (PMI) (Turney, 2002) to calculate the semantics
of comments to distinguishthe positive or negative score of their semantic orientation. After
calculating the semantics of all comments,the system disregards comments with one of the
following two conditions: serious inconsistency between the star rating and the semantic
orientation of the comment or disregarding a comment which has outlier status compared
with other comments for the same app. It is assumed that the majority of users are rational
in their analyses, so extreme outliers may be out of step with the mainstream. The
comments are summed, ltered and re-ranked for the users convenience. Invited experts
checked the three categories, and the researchersevaluated the results using Spearman and
two other statisticsto compare the system rankings with those of the experts.
Literature review
Social networks
Social networks are a powerful tool with which to analyze human interactions. Wasserman and
Faust (1994) used social theory and statistics to analyze and understand human activities in
society. Jansen et al. (2009) studied the inuence of word of mouth through analyses of microblog
comments, sentiments, and opinions in a social network. Some researchers (Al-Ouet al.,2012;
Pan, 2010) studied trust in social networks by analyzing user behavior and/or opinions to
determine whether their counterparts can be trusted. Chen et al. (2009) and Yardi et al. (2009)
proposed social-networking theory to identify spammers and detects Web spam.
Trust and trustworthiness
Rousseau et al. (1998) classiedtrust into three types: relational trust, institutionaltrust, and
calculative trust. Similarly,the concept of trust has also been studied in many elds, such as
risk management (Ward and Smith, 2004), social psychology (Bartlett and Bartlett, 1995)
and behavioral sciences(Smelser and Baltes, 2001).
Although the concepts of trust and trustworthinessare abstract and not easy to describe,
researchers have used a variety of mechanisms to evaluate them. Castelfranchi and Tan
(2001) used achievement degree to evaluate the extent of trust among people in virtual
societies. Kim et al. (2008) developeda trust-based decision-making process a consumer can
use when making a purchase from a given site. Noorian et al. (2016) presented an adaptive
decentralized trust model in an electronic commerce website. Adler et al. (2008) proposed a
system that computes quantitative values of trust for the text in Wikipedia articles.
Similarly, some scholars evaluated the trust in eBays reputation system (Resnick et al.,
2006), distributed networkattacks (Sun et al.,2006),cloud service providers (Fan et al.,2014;
Google Play
recommendations
857

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT