Konkureerivate mobiilirakenduste erijoontel ja tunnustel põhinev võrdlev analüüs
Files
Date
2016
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Tarkvara kvaliteedi mõõtmine on väljakutse paljudele ettevõtetele. Kasutaja kogemuse kogumine ja struktureerimine on väga keeruline. Kasutajate tagasiside on tavaliselt liiga üldine ning tarkvara probleemide leidmine ja seostamine tarkvara võimalustega (nt. kas tegemist on turvaveaga või kasutajaliidese probleemiga) on muutumas keeruliseks. Samal ajal näevad kliendid vaeva rakenduste võimalusepõhise võrdlemisega. Klientidel ei ole ühtegi viisi tuvastamaks, milline võimekus töötab hästi eri rakendustes ja milline mitte.\n\rEttevõtted üritavad sundida kliente andma „struktureeritud“ tagasisidet, kuid tagasiside vorme täidetakse tihti pealiskaudselt või eiratakse täielikult. Kuna võrgumüük on tänapäeval tüüpiline tarkvararakenduste evituskanal, hoitakse enamikku kasutajate tagasisidest võrgus ja avalikult kättesaadavalt veebis (nt. Google Play, Apple Store mobiilitarkvara puhul). Sellegipoolest on automaatne väärtusliku info eraldamine, positiivsete ning negatiivsete arvamuste eristamine ning rakenduste klassifitseerimine võimaluste rühmade järgi keeruline. Konkureerivate rakenduste võimalustepõhine võrdlemine on jätkuvalt raske ülesanne. Üks probleemidest on kommentaaride suur arv, mis teeb arvustajate hoiakute jälgimise keeruliseks. Analoogiliselt on keeruline leida koondarvamust tarkvara iga aspekti (või ka „võimaluse“) kohta. Sel põhjusel on hoiakuteanalüüs muutumas aina populaarsemaks. Selles valdkonnas on teostatud palju uurimusi ning loodud ja rakendatud on mitmeid meetodeid ja tööriistu. Võttes aluseks rakenduse ülevaadetest eraldatud informatsiooni, püstitatakse antud töös kolm eesmärki:\n\r1. Tuvastada etteantud rakenduste võimalused.\n\r2. Tuvastada need rakendused, mida võib funktsionaalsuse (nt. võimaluste) poolest pidada konkureerivateks.\n\r3. Võrrelda neid rakendusi kasutades hoiakute analüüsi.
Measurement of software quality is a challenge for many companies. It is very complicated to extract and structure experience that users had. The feedback is usually too general, and it is becoming tough to figure out which problems a piece of software has and in which specific features (e.g. security problem, UI). At the same time, customers are struggling with the problem of comparing applications based on their features. There is no way for customers to know exactly which functionality works well, and which does not, in different applications.\n\rCompanies are trying to make customers provide "structured" feedback. However, feedback forms are often filled in superficially and partly or completely ignored. Since selling online is nowadays the typical delivery channel of software applications, most of the customer reviews are stored online and thus publicly available on the web (e.g. Google Play, Apple Store – for mobile software). However, automatically extracting valuable information and separating positive from negative opinions, as well as classifying software apps by feature groups is difficult. Comparing competing applications based on features they have is still a hard problem. One of the problems is the large amount of comments, which makes it difficult to keep track of the reviewers’ variety of sentiments. Likewise, it is hard to figure out a summarized opinion about each aspect (also, widely used the word “feature”) of the software. That is why approaches to sentiment analysis are becoming more and more popular. Much research has been done in this field, and various methods and tools have been developed and applied. Based on information extracted from app reviews, in this thesis, I tackle three goals:\n\r1. For a given app, identify features that this software application has.\n\r2. Identify those applications that can be considered competing apps with regards to functionality (i.e., the set of features provided).\n\r3. Compare these applications using sentiment analysis.
Measurement of software quality is a challenge for many companies. It is very complicated to extract and structure experience that users had. The feedback is usually too general, and it is becoming tough to figure out which problems a piece of software has and in which specific features (e.g. security problem, UI). At the same time, customers are struggling with the problem of comparing applications based on their features. There is no way for customers to know exactly which functionality works well, and which does not, in different applications.\n\rCompanies are trying to make customers provide "structured" feedback. However, feedback forms are often filled in superficially and partly or completely ignored. Since selling online is nowadays the typical delivery channel of software applications, most of the customer reviews are stored online and thus publicly available on the web (e.g. Google Play, Apple Store – for mobile software). However, automatically extracting valuable information and separating positive from negative opinions, as well as classifying software apps by feature groups is difficult. Comparing competing applications based on features they have is still a hard problem. One of the problems is the large amount of comments, which makes it difficult to keep track of the reviewers’ variety of sentiments. Likewise, it is hard to figure out a summarized opinion about each aspect (also, widely used the word “feature”) of the software. That is why approaches to sentiment analysis are becoming more and more popular. Much research has been done in this field, and various methods and tools have been developed and applied. Based on information extracted from app reviews, in this thesis, I tackle three goals:\n\r1. For a given app, identify features that this software application has.\n\r2. Identify those applications that can be considered competing apps with regards to functionality (i.e., the set of features provided).\n\r3. Compare these applications using sentiment analysis.