[雙語(yǔ)翻譯]軟件工程外文翻譯--移動(dòng)應(yīng)用軟件工程研究的未來(lái)趨勢(shì)（英文）

1、Future Trends in Software Engineering Research for Mobile AppsMeiyappan Nagappan Department of Software Engineering Rochester Institute of Technology Rochester, NY, USA mei@se.rit.eduEmad Shihab Dept. of Computer Science

2、 and Software Engineering Concordia University Montreal, QC, Canada eshihab@cse.concordia.caAbstract—There has been tremendous growth in the use of mobile devices over the last few years. This growth has fueled the devel

3、opment of millions of software applications for these mobile devices often called as ‘a(chǎn)pps’. Current estimates indicate that there are hundreds of thousands of mobile app developers. As a result, in recent years, there h

4、as been an increasing amount of software engineering research conducted on mobile apps to help such mobile app developers. In this paper, we discuss current and future research trends within the framework of the various

5、stages in the software development life-cycle: requirements (including non-functional), design and development, testing, and maintenance. While there are several non-functional requirements, we focus on the topics of ene

6、rgy and security in our paper, since mobile apps are not necessarily built by large companies that can afford to get experts for solving these two topics. For the same reason we also discuss the monetizing aspects of a m

7、obile app at the end of the paper. For each topic of interest, we first present the recent advances done in these stages and then we present the challenges present in current work, followed by the future opportunities an

8、d the risks present in pursuing such research. Index Terms—Mobile apps, Mining app markets.I. INTRODUCTIONIn the context of this paper, a mobile app is defi ned as the application developed for the current generation of

9、mobile devices popularly known as smart phones. These apps are often distributed through a platform specifi c, and centralized app market. In this paper, we sometimes refer to mobile apps simply as apps. In the past few

10、years we are observing an explosion in the popularity of mobile devices and mobile apps [17]. In fact, recent market studies show that the the centralized app market for Apple’s platform (iOS) and Google’s platform (Andr

11、oid), each have more than 1.5 million apps [8]. These mobile app markets are extremely popular among developers due to the flexibility and revenue potential. At the same time, mobile apps bring a whole slew of new challe

12、nges to software practitioners - such as challenges due to the highly-connected nature of these devices, the unique distribution channels available for mobile apps (i.e., app markets like Apple’s App Store and Google’s G

13、oogle Play), and novel revenue models (e.g., freemium and subscription apps). To date the majority of the software engineering research has focused on traditional “shrink wrapped” software, such as Mozilla Firefox, Eclip

14、se or Apache HTTP [79]. However, re-cently researchers have begun to focus on software engineering issues for mobile apps. For example, the 2011 Mining Software Repositories Challenge focused on studying the Android mobi

15、le platform [90]. Other work focused on issues related to code reuse in mobile apps [84], on mining mobile app data from the app stores [34], testing mobile apps [70] and teaching programming on mobile devices [95]. Ther

16、efore, we feel it is a perfect time to reflect on the accomplishments in the area of Software Engineering research for mobile apps and to draw a vision for its future. Note that we restrict to just the software engineeri

17、ng topics for mobile apps in this paper, and even that not exhaustively due to space restrictions (we skip topics like usability or performance engineering since an entire paper can be written on each of these topics). W

18、e do not discuss the advancements in other areas of research for mobile apps such as cloud based solutions, or networking in mobile apps. The purpose of this vision paper is to serve as a reference point for mobile app w

19、ork. We start by providing some background information on mobile apps. Then, we discuss the current state-of-the-art in the fi eld, relating it to each of the software development phases, i.e., requirements, development,

20、 testing, and maintenance as shown in Figure 1. We also talk about two non-functional requirements: energy use and security of mobile apps. Finally, even though it is not one of the software development phases, we talk a

21、bout the software engineering challenges and recommendations for monetizing mobile apps. Along with a discussion of the state-of-the-art, we also present the challenges currently faced by the researchers/developers of mo

22、bile apps. Then we discuss our vision for the future of software engineering research for mobile apps and the risks involved, based on our experiences. Our hope is that our vision paper will help newcomers to quickly gai

23、n a background in the area of mobile apps. Moreover, we hope that our discussion of the vision for the area will inspire and guide future work and build a community of researchers with common goals regarding software eng

24、ineering challenges for mobile apps. A word of caution though - the discussion of the current state-of-the-art is not meant to be a systematic literature survey (for a more comprehensive study please refer to Sarro et al

25、. [88]), and the future directions of research are based on our opinions that have been influenced by our knowledge of the research in this community.Fig. 2. Overview of the various stakeholders with respect to modern da

26、y mobile apps.Fig. 3. Snapshot of an app in the app marketcategories. Therefore, now as researchers we have access to apps that have been self reported to be in the same domain. This gives researchers tremendous potentia

27、l to conduct research that can be controlled for the domain of the app. Often we see that a software engineering research study is done on an IDE, like Eclipse and another OSS project like the browser Firefox [20]. Howev

28、er, we do not know what domains of applications that these results transfer to. In the world of mobile apps, if we conduct our research on only game apps, then we can be more certain that our fi ndings would apply to oth

29、er game apps.Additionally, all these various data points are available for hundreds of thousands of apps in a public facing website making it a rich dataset for researchers to crawl.A. Common ChallengesIn the next sectio

30、n, we discuss the accomplishments, chal- lenges and risks for each of the development phases. However, one challenge seems to be a common challenge that impacts all of the development phases, public access to data. Such

31、access challenges manifests in three ways. Firstly, app stores restrict public access to their data and typically only allow for access to a subset of all the user reviews. For example, in the case of the Google Play sto

32、re, one can only access 500 reviews for an app. Secondly, app stores do not provide the source code of the apps, or any other associated artifacts like test code, or design and requirement documents. Only the app binary

33、and release notes are made available.Finally, with respect to the release notes, and the app binary, one can only get them for the latest release. There is no historical information that can be collected from the app sto

34、re (except user reviews). The only way to gather historical information on the various releases of the mobile apps is to continuously mine the app stores at regular intervals (like daily or weekly basis).III. REQUIREMENT

35、SA number of studies have focused on requirement extraction for mobile applications. Contrary to traditional work on software requirements, which mainly focused on analysis of the requirements and specifi cations documen

36、t, the majority of mobile app-related studies leveraged app reviews posted by users to extract requirements. For example, Iacob et al. [40] used linguistic rules to detect feature requests from user reviews. Then, they s

37、ummarize the feature requests to generate more abstract requirements. Galvis-Carreno and Winbladh [27] extract topics from user reviews in order to revise requirements. They show that their automatically extracted requir

38、ements match with manually extracted requirements. Guzman and Maalej [30], [60] use natural language processing (NLP) techniques to identify app features in the reviews and use sentiment analysis to determine how users f