Sentiment Analysis of Movie Review Using Integrated CNN-LSTM Mode (CNN-LSTM 조합모델을 이용한 영화리뷰 감성분석)
-
- Journal of Intelligence and Information Systems
- /
- v.25 no.4
- /
- pp.141-154
- /
- 2019
Rapid growth of internet technology and social media is progressing. Data mining technology has evolved to enable unstructured document representations in a variety of applications. Sentiment analysis is an important technology that can distinguish poor or high-quality content through text data of products, and it has proliferated during text mining. Sentiment analysis mainly analyzes people's opinions in text data by assigning predefined data categories as positive and negative. This has been studied in various directions in terms of accuracy from simple rule-based to dictionary-based approaches using predefined labels. In fact, sentiment analysis is one of the most active researches in natural language processing and is widely studied in text mining. When real online reviews aren't available for others, it's not only easy to openly collect information, but it also affects your business. In marketing, real-world information from customers is gathered on websites, not surveys. Depending on whether the website's posts are positive or negative, the customer response is reflected in the sales and tries to identify the information. However, many reviews on a website are not always good, and difficult to identify. The earlier studies in this research area used the reviews data of the Amazon.com shopping mal, but the research data used in the recent studies uses the data for stock market trends, blogs, news articles, weather forecasts, IMDB, and facebook etc. However, the lack of accuracy is recognized because sentiment calculations are changed according to the subject, paragraph, sentiment lexicon direction, and sentence strength. This study aims to classify the polarity analysis of sentiment analysis into positive and negative categories and increase the prediction accuracy of the polarity analysis using the pretrained IMDB review data set. First, the text classification algorithm related to sentiment analysis adopts the popular machine learning algorithms such as NB (naive bayes), SVM (support vector machines), XGboost, RF (random forests), and Gradient Boost as comparative models. Second, deep learning has demonstrated discriminative features that can extract complex features of data. Representative algorithms are CNN (convolution neural networks), RNN (recurrent neural networks), LSTM (long-short term memory). CNN can be used similarly to BoW when processing a sentence in vector format, but does not consider sequential data attributes. RNN can handle well in order because it takes into account the time information of the data, but there is a long-term dependency on memory. To solve the problem of long-term dependence, LSTM is used. For the comparison, CNN and LSTM were chosen as simple deep learning models. In addition to classical machine learning algorithms, CNN, LSTM, and the integrated models were analyzed. Although there are many parameters for the algorithms, we examined the relationship between numerical value and precision to find the optimal combination. And, we tried to figure out how the models work well for sentiment analysis and how these models work. This study proposes integrated CNN and LSTM algorithms to extract the positive and negative features of text analysis. The reasons for mixing these two algorithms are as follows. CNN can extract features for the classification automatically by applying convolution layer and massively parallel processing. LSTM is not capable of highly parallel processing. Like faucets, the LSTM has input, output, and forget gates that can be moved and controlled at a desired time. These gates have the advantage of placing memory blocks on hidden nodes. The memory block of the LSTM may not store all the data, but it can solve the CNN's long-term dependency problem. Furthermore, when LSTM is used in CNN's pooling layer, it has an end-to-end structure, so that spatial and temporal features can be designed simultaneously. In combination with CNN-LSTM, 90.33% accuracy was measured. This is slower than CNN, but faster than LSTM. The presented model was more accurate than other models. In addition, each word embedding layer can be improved when training the kernel step by step. CNN-LSTM can improve the weakness of each model, and there is an advantage of improving the learning by layer using the end-to-end structure of LSTM. Based on these reasons, this study tries to enhance the classification accuracy of movie reviews using the integrated CNN-LSTM model.
1. Introduction Today Internet is recognized as an important way for the transaction of products and services. According to the data surveyed by the National Statistical Office, the on-line transaction in 2007 for a year, 15.7656 trillion, shows a 17.1%(2.3060 trillion won) increase over last year, of these, the amount of B2C has been increased 12.0%(10.2258 trillion won). Like this, because the entry barrier of on-line market of Korea is low, many retailers could easily enter into the market. So the bigger its scale is, but on the other hand, the tougher its competition is. Particularly due to the Internet and innovation of IT, the existing market has been changed into the perfect competitive market(Srinivasan, Rolph & Kishore, 2002). In the early years of on-line business, they think that the main reason for success is a moderate price, they are awakened to its importance of on-line service quality with tough competition. If it's not sure whether customers can be provided with what they want, they can use the Web sites, perhaps they can trust their products that had been already bought or not, they have a doubt its viability(Parasuraman, Zeithaml & Malhotra, 2005). Customers can directly reserve and issue their air tickets irrespective of place and time at the Web sites of travel agencies or airlines, but its empirical studies about these Web sites for reserving and issuing air tickets are insufficient. Therefore this study goes on for following specific objects. First object is to measure service quality and service recovery of Web sites for reserving and issuing air tickets. Second is to look into whether above on-line service quality and on-line service recovery have an impact on overall service quality. Third is to seek for the relation with overall service quality and customer satisfaction, then this customer satisfaction and loyalty intention. 2. Theoretical Background 2.1 On-line Service Quality Barnes & Vidgen(2000; 2001a; 2001b; 2002) had invented the tool to measure Web sites' quality four times(called WebQual). The WebQual 1.0, Step one invented a measuring item for information quality based on QFD, and this had been verified by students of UK business school. The Web Qual 2.0, Step two invented for interaction quality, and had been judged by customers of on-line bookshop. The WebQual 3.0, Step three invented by consolidating the WebQual 1.0 for information quality and the WebQual2.0 for interactionquality. It includes 3-quality-dimension, information quality, interaction quality, site design, and had been assessed and confirmed by auction sites(e-bay, Amazon, QXL). Furtheron, through the former empirical studies, the authors changed sites quality into usability by judging that usability is a concept how customers interact with or perceive Web sites and It is used widely for accessing Web sites. By this process, WebQual 4.0 was invented, and is consist of 3-quality-dimension; information quality, interaction quality, usability, 22 items. However, because WebQual 4.0 is focusing on technical part, it's usable at the Website's design part, on the other hand, it's not usable at the Web site's pleasant experience part. Parasuraman, Zeithaml & Malhorta(2002; 2005) had invented the measure for measuring on-line service quality in 2002 and 2005. The study in 2002 divided on-line service quality into 5 dimensions. But these were not well-organized, so there needed to be studied again totally. So Parasuraman, Zeithaml & Malhorta(2005) re-worked out the study about on-line service quality measure base on 2002's study and invented E-S-QUAL. After they invented preliminary measure for on-line service quality, they made up a question for customers who had purchased at amazon.com and walmart.com and reassessed this measure. And they perfected an invention of E-S-QUAL consists of 4 dimensions, 22 items of efficiency, system availability, fulfillment, privacy. Efficiency measures assess to sites and usability and others, system availability measures accurate technical function of sites and others, fulfillment measures promptness of delivering products and sufficient goods and others and privacy measures the degree of protection of data about their customers and so on. 2.2 Service Recovery Service industries tend to minimize the losses by coping with service failure promptly. This responses of service providers to service failure mean service recovery(Kelly & Davis, 1994). Bitner(1990) went on his study from customers' view about service providers' behavior for customers to recognize their satisfaction/dissatisfaction at service point. According to them, to manage service failure successfully, exact recognition of service problem, an apology, sufficient description about service failure and some tangible compensation are important. Parasuraman, Zeithaml & Malhorta(2005) approached the service recovery from how to measure, rather than how to manage, and moved to on-line market not to off-line, then invented E-RecS-QUAL which is a measuring tool about on-line service recovery. 2.3 Customer Satisfaction The definition of customer satisfaction can be divided into two points of view. First, they approached customer satisfaction from outcome of comsumer. Howard & Sheth(1969) defined satisfaction as 'a cognitive condition feeling being rewarded properly or improperly for their sacrifice.' and Westbrook & Reilly(1983) also defined customer satisfaction/dissatisfaction as 'a psychological reaction to the behavior pattern of shopping and purchasing, the display condition of retail store, outcome of purchased goods and service as well as whole market.' Second, they approached customer satisfaction from process. Engel & Blackwell(1982) defined satisfaction as 'an assessment of a consistency in chosen alternative proposal and their belief they had with them.' Tse & Wilton(1988) defined customer satisfaction as 'a customers' reaction to discordance between advance expectation and ex post facto outcome.' That is, this point of view that customer satisfaction is process is the important factor that comparing and assessing process what they expect and outcome of consumer. Unlike outcome-oriented approach, process-oriented approach has many advantages. As process-oriented approach deals with customers' whole expenditure experience, it checks up main process by measuring one by one each factor which is essential role at each step. And this approach enables us to check perceptual/psychological process formed customer satisfaction. Because of these advantages, now many studies are adopting this process-oriented approach(Yi, 1995). 2.4 Loyalty Intention Loyalty has been studied by dividing into behavioral approaches, attitudinal approaches and complex approaches(Dekimpe et al., 1997). In the early years of study, they defined loyalty focusing on behavioral concept, behavioral approaches regard customer loyalty as "a tendency to purchase periodically within a certain period of time at specific retail store." But the loyalty of behavioral approaches focuses on only outcome of customer behavior, so there are someone to point the limits that customers' decision-making situation or process were neglected(Enis & Paul, 1970; Raj, 1982; Lee, 2002). So the attitudinal approaches were suggested. The attitudinal approaches consider loyalty contains all the cognitive, emotional, voluntary factors(Oliver, 1997), define the customer loyalty as "friendly behaviors for specific retail stores." However these attitudinal approaches can explain that how the customer loyalty form and change, but cannot say positively whether it is moved to real purchasing in the future or not. This is a kind of shortcoming(Oh, 1995). 3. Research Design 3.1 Research Model Based on the objects of this study, the research model derived is shows, Step 1 and Step 2 are significant, and mediation variable has a significant effect on dependent variables and so does independent variables at Step 3, too. And there needs to prove the partial mediation effect, independent variable's estimate ability at Step 3(Standardized coefficient
shows, Step 1 and Step 2 are significant, and mediation variable has a significant effect on dependent variables and so does independent variables at Step 3, too. And there needs to prove the partial mediation effect, independent variable's estimate ability at Step 3(Standardized coefficient
이메일무단수집거부
이용약관
제 1 장 총칙
제 2 장 이용계약의 체결
제 3 장 계약 당사자의 의무
제 4 장 서비스의 이용
제 5 장 계약 해지 및 이용 제한
제 6 장 손해배상 및 기타사항
Detail Search
Image Search
(β)