• No results found

PREDICTING APPLICATION REVIEW RATING WITH TEXT MINING

N/A
N/A
Protected

Academic year: 2020

Share "PREDICTING APPLICATION REVIEW RATING WITH TEXT MINING"

Copied!
6
0
0

Loading.... (view fulltext now)

Full text

(1)

Available Online at www.ijpret.com 186

INTERNATIONAL JOURNAL OF PURE AND

APPLIED RESEARCH IN ENGINEERING AND

TECHNOLOGY

A PATH FOR HORIZING YOUR INNOVATIVE WORK

PREDICTING APPLICATION REVIEW RATING WITH TEXT MINING

MS. TEJASHRI R. GAIKWAD1, PROF. A. B. RAUT2

1. Student ME CSE, HVPM COET, Amravati, Maharashtra, India. 2. Associate Professor CSE, HVPM COET, Amravati Maharashtra, India.

Accepted Date: 05/03/2015; Published Date: 01/05/2015

\

Abstract: With the rapid evolution of smart phones, mobile application have become essential part of our lives. Today most of the android applications development and modification are based on customer reviews and feedbacks. Everyday new applications are launched in the market and these applications are available on android market and apple store. It is very challenging for a common users to find out the relevant application satisfying his or her need because user cannot trust on application just by seeing the rating 1 to 5 because these ratings does not include other vital problems such as graphic user interface, battery power, speed, user friendliness of application. In this paper we are discussing a system which provide the review application depending on the user comments, the application get review considering all aspects such as graphic user interface, battery power, speed, user friendliness of application and all other which are really important and easy to understand.

Keywords: Online reviews, Application Store, Text Mining, Android market.

Corresponding Author: MS. TEJASHRI R. GAIKWAD

Access Online On:

www.ijpret.com

How to Cite This Article:

(2)

Available Online at www.ijpret.com 187 INTRODUCTION

Sentiment analysis is a popular technique for summarizing and analyzing consumers’ textual reviews about products and services. Generally, to understand customer demands and feedback over the application, developers and end-users can communicate and understand each other through user forums or the website of application stores. For example, on the website of application stores, user can rate and write opinion about different applications. Based on the user rating, application stores websites provide the ranking of applications in different categories. The rating information, however, does not provide information for developers to improve the applications. But sometimes it’s necessary for developers to understand the textual content of user reviews. Many users provide the reasons why they do not like the application and what they like in the application, for example, “Game is good! However, needs more levels.” But the textual content of user reviews has at times been ignored. The developers can improve the application by increasing levels and the necessary things which user wants. If our focus is only on the star rating, no one will know how to improve this application or why this application is so popular and good.

1. Literature review and related work

There are several studies which examine and study the detailed textual information contained within the online reviews. Content analysis was utilized to identify the factors affecting the evaluative content of book reviews in sociology. As this technique is extraordinarily time-consuming while dealing with a large amount of data, such as online book reviews, text mining is one of the fields of data mining which gaining attention in information system research [1]. The first relevant stream of literature assesses the effect of online product reviews on sales. Research in this direction has generally assumed that the primary reason that reviews influence sales is because they provide information about the product or the vendor to potential consumers [2].

(3)

Available Online at www.ijpret.com 188 Several studies have extracted users’ opinions from online reviews to predict products’ sales [1] and examined and study the effect of semantic characteristics on the helpfulness votes [2]. However, few studies have realized that the usefulness and importance of identifying the topics contained in user reviews in the application stores by automatically extracting the semantic information of reviews, thereby possibly helping developers to better understand user’s responses and opinion. This research explores the potential ability to utilize a text-mining technique to quantify and measure the semantic information contained in user reviews for applications.

Although reading online reviews may help developers to improve applications and update where it is necessary, the user reviews can be large in number in many cases the large quantity of textual reviews available for an application can be overwhelming, thereby hindering the developer’s ability to track the key information. In order to help the developers to more easily find helpful review information to improve the application, a text mining based method of summarizing online reviews with the help of text miner was investigated. Our objective is to assist developers to better understand customer demands and thus promote the sales of application by employing the text mining based approach.

2. Analysis of Problem

With the rapid evolution of smart phones, mobile applications have become essential part of our lives. Every smart phone now contained the various applications. Recently with the popularity of smart phones and tablet personal computers, thousands of application developers, working as additional service provides, have emerged and a so called producer-consumer network has expand. The sale of application is important for both developers and applications providers such as the Android market and Apple store. Both developers and application providers are deeply concerned with ways to improve online application stores so that sales and customer satisfaction remain high.

(4)

Available Online at www.ijpret.com 189 3. Workflow of Text mining

Figure 1: Workflow of data mining

DATA COLLECTION

We collect data from the Android Market and Apple Store, online application providers for smart phones and tablet personal computers running the Android system. The Android Market provide an ideal study environment for this paper because the competition in the market is furious and developers must understand users’ demands and wants from the application. The Android Market offer users a friendly feedback system to share their opinions and experiences about the applications which they used. The user review system on the website includes detailed comments and a five-star user rating system. One of the advantages of the Android Market online user review system is that many users can note the deficiency of the applications so that developers can improve them. To collect review data, we chose one of the most popular applications in the Android Market, as our target application.

TOPIC MODELING

In machine learning and natural language processing, a topic model is a type of statistical model for discovering the abstract topics that occur in our collection of documents. Various extended LDA models have been used for automatic topic extraction from large-scale principal sum. In the context of social tagging systems, where multiple users are commenting resources, the resulting topics reflect a shared view of the document; and the tags of the topics reflect a common vocabulary. As for community detection, the most representative approaches include centrality or between’s-based approaches and graph partitioning-based approaches. Girvan and Newman extended the between’s measure to edges and designed a clustering algorithm which gradually removes the edges with the highest between’s

Data collection

Topic Modeling

Preprocessing

(5)

Available Online at www.ijpret.com 190 value. Many studies provide various models and algorithms for topic mining and community detection; yet, few of them have integrated those models and algorithms, performed topic mining for detected communities. Although much work has been done in the topic modeling, there remains a need to apply these topic models empirically. The abundance of information created by the online review system provides a significant obstacle to identifying the information concerned. This paper take the initiative to investigate how to identify the latent topics contained in the application review data, potentially helping developers to more easily identify useful user demand information.

Preprocessing

Preprocessing was performed on the data which we collect from android market related to our specified application for reviewing data before the subsequent text analysis. Punctuation,

numbers and other non-alphabetic characters were removed first. .

Latent Features

The latent and potential topics identified from our dataset are quality this feature talks about the game’s graphics and sound quality. The key words related to this feature include: graphics, screen, sound. Performance this feature refers to the game’s speed and update situation. The key words related to this feature include: update, speed, space. User Interface this feature is about how the game interacts with users, for example, if users complain with the advertisements popping up during the game. The key words related to this feature include: advertisement.

TEXT MINING WITH TEXT MINER

Manual Coding

Coder will code scores of each feature independently based on reviews. The final result of the scoring will be the average of two independent scoring. Some scores are related to the positive feature of the application and some scores are related to the negative feature of the application. And there is one of the score which is related to the review which does not mention the feature.

4. Objective

(6)

Available Online at www.ijpret.com 191

 We import text data into Text Miner from different sources and which is in different formats.

 Create statistical, rule-based, and hybrid models through which we can understand and predict customer sentiments.

5. CONCLUSION

In this paper, we examined a previously ignored yet important research question concerning the online user reviews: How can we enhance the communication between application developers and users? We addressed this question by investigating a text mining based method for mining and summarizing the users’ opinions about the applications, which are contained in the online reviews. We first identified the features users mostly mentioned in the online reviews by employing LDA algorithm. Then the feature-related sentiment words are captured by Miner and we further established a model to predict the feature rating and overall rating based on reviews with Enterprise Miner.

REFERENCES

1. Tianxi Dong, Johnhua kim, lin920130 Predicting Application Review rating with SAS Text Miner Paper 269-2013.

2. Ghose, A., & Ipeirotis, P. G. (2011). Estimating the Helpfulness and Economic Impact of Product Reviews: Mining Text and Reviewer Characteristics. Knowledge and Data Engineering, IEEE Transactions on, 23(10), 1498-1512.

3. Cerrito, P. (2011). Sentiment Mining Using SAS® Text Miner®. Paper presented at the SAS Global Forum 2011, Las Vegas.

4. Blei, D.M., Ng, A.Y., & Jordan, M., I (2003) Latent dirichlet allocation. Journal of Machine Learning Research, 3, 993–1022.

5. Blei, D. M. & Lafferty, J. D.(2006). Dynamic topic models. The Proceedings of the 23rd international conference on machine learning (ICML2006), page 113-120, Pittsburgh, Pennsylvania, USA.

References

Related documents

*Appearance, *Actions, *Words, *Thoughts and Feelings, and *Character Relationships.. Characterization:

fundamentals masks important sectoral, local, and regional shocks that buffeted banks with particular credit or market risks. Using a narrative approach similar to that of Friedman

[r]

A cross-sectional study was conducted during the months of June to September 2017 to assess knowledge among private medical professionals about HIV infection,

Table 1 above shows the usage of UML diagrams based on a survey done exclusively on the Sri Lankan software development organizations. For our Modular Transformation project we

Our  relationship  was  always

(Industry)/ (Information technology) Submodule name Introduction to Project Management Submodule number WI-1.132.2. Main module Personnel and Project Management

Clark: Media Will Never Influence Learning, Educational Technology Research and Development Vol. 21–29 • Media are “mere vehicles that deliver instruction but do