The Study of Comparing Korean Consumers' Attitudes Toward Spotify and MelOn: Using Semantic Network Analysis

Abstract

This study examines Korean users' attitudes and emotions toward Melon and Spotify, which lead the music streaming market. We used Text Mining, Semantic Network Analysis, TF-IDF, Centrality, CONCOR, and Word2Vec analysis. As a result of the study, MelOn was used in a user's daily life. Based on Melon's advantages of providing various contents, the advantage is judged to have considerable competitiveness beyond the limits of the streaming app. However, the MelOn users had negative emotions such as anger, repulsion, and pressure. On the contrary, in the case of Spotify, users were highly interested in the music content. In particular, interest in foreign music was high, and users were also interested in stock investment. In addition, positive emotions such as interest and pleasure were higher than MelOn users, which could be interpreted as providing attractive services to Korean users. While previous studies have mainly focused on technical or personal factors, this study focuses on consumer reactions (online reviews) according to corporate strategies, and this point is the differentiation from others.

1. Introduction

Advances in Information Technology (IT) have brought to rapid changes in various industries, and the changes are still underway. The music industry is also going through a turbulent period with the incorporation of IT technology [Avdeeff, 2012]. Technological advances have transformed the concept of music from listening with physical constraints such as place and time to relishing with a new concept that anyone can enjoy anytime, anywhere [Krause et al., 2015; Heye and Lamont, 2010]. On account of research on the digitalization of the music industry has been discussed for a long time [Snadler, 2007], researchers nowadays enable research and analysis of various musical elements. However, there has been research and controversy in the past about digital music (called online music) consumption. Sharing P2P files provided by Napster, it initially approached mainly from an illegal and unscrupulous perspective [Fisher, 2004; Levin et al., 2007]. However, over time, music consumption has expanded from analog to digital and has been chosen by consumers, and it has changed from an illegal and unscrupulous perspective to a legitimate and moral perspective [Sinclair and Green, 2016].

Some studies have shown that online music consumption accounts for 50% of the global music market [IFPI, 2016]. Therefore, it is evidence that online music is leading the growth of the music market and being possible a new source of revenue. In particular, many users use online services using smart devices, and the rate of using smartphones (85.8%) is overwhelmingly higher than other devices such as computers (38.5%) and TVs (31.5%) [The Korea Creative Content Agency, 2020]. In other words, this could be interpreted that consumers who consume online music prefer streaming services on smartphones based on accessibility and convenience [Heye and Lamont, 2010]. In a report on the music industry published by the Korea Creative Content Agency [2020], the Korean music market ranked sixth globally, indicating that 63.6% of users who use music streaming and download services have experienced digitalization. In the study, 52.4% of total users only used streaming services. It means that online music service is expanding in the Korean music market. The development of IT technology has brought about a significant change in the music industry. The technology changed the concept of music consumption and brought about the music market’s growth. As a result, the environment of consuming music is also changed from offline to online.

This study aims to understand the attitudes of Korean consumers toward online music services by comparing and analyzing Spotify, a global leader company, and MelOn, a Korean leader company. Furthermore, we would extract important keywords from users’ reviews of each company and analyze differences in the service evaluation of users. Previous studies focused on changes due to demographic differences such as income or sex and effects due to technical factors, but this study has a differentiation in that it focuses on consumer reactions (using reviews) by corporate strategies. The purpose of this study is as follows. First, by analyzing the reviews of streaming music users, we would strive to derive user emotion and interest keywords. The user’s attitude and preference for streaming music services would be identified based on the result. Second, by comparing and analyzing users’ reviews of Spotify and MelOn, we would identify how Korean users experience the two companies. Finally, we would provide guidelines that reflected user emotions to online music service companies.

2. Theoretical Background

2.1 Previous Studies on Online Music

The online music market began with the first evaluation of illegality and pirates [Cesareo and Pasteore, 2014[. At the time, studies were conducted to prevent such illegal activities [Levin et al., 2007], and psychological and demographic characteristics of illegal activities were conducted [Gray, 2012]. In other words, the online music market was recognized as having a negative effect on the existing music market. However, as a new way of making profits was generated in the music market, the perception of online music changed rapidly. Moreover, the online music market has grown steadily. For example, in IFPI’s report [2019], which was the comparative analysis of album sales and online (including streaming) services from 2001 to 2018, album sales, which generated more than $23billion in revenue in 2001, were reduced to $4.7billion in 2018, but streaming services achieved rapid growth over $8.9billion in 2018 after achieving $1billion in 2012 [IFPI, 2019; Korea Creative Content Agency, 2019].

In general, consumers can take pleasure with music by paying a certain amount of money to listen to online music, downloading and using streaming services, or watching advertisements provided in the middle of the music and listening to music for free. In addition, if consumers pay more than a certain fee, consumers could use both services. The new way of providing music also changed the supply chain of music companies [Graham et al., 2004], and in particular, online music services resulted in a direct connection between consumers and artists, reducing the influence of the companies [Hughes et al., 2003]. In addition, however, consumers could listen to their preferred music more easily and quickly.

Research on online music services has been conducted actively. The research has been conducted on various perspectives, including the negative role of sales as a substitute in the music market [Hiller, 2016; Aguiar and Waldfogel, 2018], the role of supplementary materials [Aguiar, 2017; Kretschmer and Peukert, 2015], and the economic analysis and legal issues of streaming music service platforms [Aguiar and Waldfogel, 2018]. Besides, Tepper and Hargittai [2009] and Smith [2012] conducted research based on demographic criteria. According to the studies, the younger prefer to use the higher use of YouTube and online music, and the older prefer using analog music such as CDs. In addition, research on consumer technology adaptation and music piracy in the online music market was conducted, and specific patterns according to income and gender were also found [North and Hargreaves, 2008]. It could be explained that this research has a difference from previous studies in that users’ emotions after experiencing the service depending on the company’s strategy would be directly analyzed and make a conclusion from the results.

2.2 Spotify and MelOn

Spotify was a small company that started in Sweden in 2006. Spotify introduced a streaming method based on advertisements [Sletten, 2021]. Two types of free and paid services were provided, and free users had to listen to advertisements every 30 minutes to continue listening to the following music. Such a system was revolutionary because the existing way to listen to music had to pay a specific cost to buy CDs or LPs or illegally reproduce music on the Internet. However, Spotify’s new system reduced the user’s debt for illegal copying or downloading music and even provided satisfaction that they paid legitimate costs through watching advertisements [Carver, 2016]. Although Spotify suffered management difficulties due to copyrights, the company had 345 million subscribers as of March 2021, and nearly half are paid users. Spotify launched its service, providing various content and music in South Korea in 2021.

MelOn, which SK Telecom established in 2004, went through Loen Entertainment in 2008 and was merged by Kakao in 2016. It has been divided from Kakao as a subsidiary, and the total sales increased by more than half a billion dollars in 2020. MelOn accounted for about 40% of the market share in Korea, and more than 50% of users listened to music by streaming or downloading services [Korea Creative Content Agency, 2020]. In particular, in the survey, the primary age group using MelOn was 25 to 29 years old, and the number of users subscribed to the service for more than three years. Spotify and MelOn are leaders in the online music market worldwide and Korea. However, with launching Spotify’s service in Korea, which could be a changer for the market, direct competition with MelOn has become inevitable. This circumstance could be explained as the background and reason of this study focusing on Spotify and MelOn.

2.3 The Differences between Spotify and MelOn

The streaming service model strategy provided by Spotify and MelOn to users is different. MelOn uses a strategy to listen to music after purchasing a subscription fee. If the license is not subscribed, the user could only listen to up to 60 seconds per song, and more than it could not be played. On the other hand, Spotify provides a seven-day free experience without a procedure for log-in or registration. In addition, from June 2021, if a user registers credit card information, the user could use the service for free for three months and use various other functions. Furthermore, MelOn provides a variety of music-related content, including music videos, radio, comment writing, and most of the lyrics for the songs provided. On the other hand, Spotify provides songs and radio services focusing on listening to music and does not provide functions such as writing comments or videos. In addition, the lyrics of songs are provided fewer than MelOn, and only paid users could check the lyrics. The characteristic differences between Spotify and MelOn could be summarized (<Table 1>).

<Table 1> The Difference between Spotify and MelOn Service Models

The difference between Spotify and Melon analyzed, Spotify is making efforts to meet customer needs based on an algorithmic customized playlist of individuals. It analyzed the reviews of various users and explained that Spotify also showed advantages in sound quality. However, the cost for service and the number of Korean users who could listen, MelOn is far ahead of Spotify. Spotify and MelOn’s service models and strategies are found distinct differences. However, research for comparative analysis from the user’s point of view according to the difference in service strategies is not enough. In this study, we intend to proceed with the analysis by judging that such differences in strategies would affect the attitude and emotion of users.

3. Research Methodology

3.1 The Subject and Method of Analysis

For this study, review data was collected. Social Network Service (SNS) was mainly used, and the primary analysis procedure is as follows. After setting the search terms: “Melon App” and “Spotify.” The websites for the review data were Naver Blog, Daum Blog, Instagram, Twitter, and Facebook. The programs used for data collection were Textom and Python.

3.2 Text Mining

The text form is generally divided into unstructured and structured data. For example, data classified based on certain conditions and criteria could be defined as standardized data. On the other hand, data composed of various documents or pages on the Internet could be unstructured data. Text mining is a process of finding hidden meanings of text by analyzing various patterns, models, and flows of data from documents [Feldman, 1995; Nahm, 2004]. Text mining has various advantages, such as extracting features from structured or unstructured data, collecting research-related topics or terms from documents, and categorizing in-document data (Choudhary et al., 2009). Because of these advantages, we set up the plan to collect data through text mining.

3.3 Term Frequency - Inverse Document Frequency (TF-IDF)

Term Frequency is a result of calculating the frequency of all words in a single document, and the result could be interpreted that the higher the frequency, the higher the importance of the document [Li and Liu, 2012]. Although, to interpret the importance according to the frequency, it is a possibility that the word is also high in documents on other issues. Therefore, the Inverse Document Frequency value should be obtained, and words with high frequency in other documents should be excluded [Robertson, 2004]. In this study, TF-IDF was used to extract accurate keywords. The method is suitable for extracting words with high frequency from collected reviews and low frequency from other reviews or Internet pages.

3.4 Semantic Network Analysis

Semantic Network Analysis (SNA) is a methodology that could interpret the meaning of the result in detail through using the strength of nodes and links [Feldman and Dagan, 1995]. SNA is different from focusing on a two-way mode for identifying the shape and pattern of a relationship in the existing network analysis. SNA is a differentiated methodology in that the content of messages flowing in the relationship of results could be interpreted by including the two-way mode analysis.

The advantages and characteristics of SNA are as follows [An, 2017; Jang and Barnett, 1994]. First, collecting data using Social Network Service (SNS) is mainly done, and it enhances efficiency to collect tens of thousands of reviews in a few hours. Second, compared to the survey, SNA could be described that the reviews of participating users are vast and more objective in individual expressions of opinion. Furthermore, users express their opinions in various ways, making it possible to obtain unexpected results. Finally, SNA has the characteristic that research could be conducted by quantitatively analyzing a large amount of qualitative data. For this study, we utilize SNA with these advantages and characteristics.

3.5 Convergence of iteration Corealtion (CONCOR) Analysis

In general, when analyzing keyword-oriented big data, research topics are analyzed through frequency analysis of the main keyword. CONCOR analysis, however, is able to analyze one step more in detail and in-depth than a frequency analysis study. Additionally, CONCOR analysis classifies structural equivalence based on correlations existing between formalized keywords [Wasserman and Faust, 1994]. In addition, CONCOR has the advantage of forming a meaningful cluster and explaining the characteristics centering on the attributes related to between keywords. CONCOR is also the most commonly used analysis method to understand structural equivalence. Therefore, the CONCOR analysis will be used in this study, and the program used is Ucinet 6.0.

3.6 Analyzing Emotional Vocabulary

As the last step of this study, the topics and keywords were summarized again, and the emotional intensity scores of keywords were calculated and compared using Textom’s emotional vocabulary dictionary. The model used in this study was Word2Vec, and five words closest to the keyword were selected. Then, the words were specifically compared and analyzed in the consumers’ evaluation in detail. The Word2Vec model learns and identifies words belonging to sentences in data as a specific criterion for analyzing word matching. The model expresses meaningfully similar words [Mikolov et al., 2013].

<Figure 1> Research Process and Employed Tools

4. Results of Analysis

4.1 Result of Collecting Data

Review data was collected on SNS with keywords ‘MelOn App’ and ‘spotify.’ In the case of Spotify, since it started its service on February 2, 2021, in Korea. 5,172 cases of MelOn data and 4,269 cases of Spotify data were collected. Through data preprocessing, 4,630 cases of MelOn and 3,858 cases of Spotify were used for analysis.

4.2 Result of Analyzing MelOn Data

After preprocessing data, elimination of stop-words was conducted, and data were extracted in the form of nouns and adjectives. The processed data identified the frequency of each word, and the top 50 words with a high frequency are as follows (<Table 2>). For example, Melon showed a high frequency of use, good, song, possible, function, recommendation, YouTube, photo, video, and payment. Additionally, TF-IDF analysis was conducted, and the result is as follows (<Table 3>).

<Table 2> Top 50 Frequency Keywords in MelOn

<Table 3> Analyzing TF-IDF in MelOn

Connection centrality means that a specific node is directly connected to another node, and Proximity centrality refers to how a node is central in the entire network. Eigenvector centrality means that the degree of authority is based on the importance of connected nodes. Finally, mediation centrality means the degree to which a specific word is located between two other words. For finding the relationship among the words, we analyzed four centralities. The analysis included the top 25 words. However, the case of Proximity- and Mediation- centrality had no obvious difference between the data. Because of the result, we focused on Connection- and Eigenvector-centrality.

The aggregated TF-DIF top 100 keywords were made into a one-way mode matrix, and the correlation between keywords was calculated and divided by group using Ucinet’s CONCOR analysis, and the keyword frequency of each group was summed to derive six categories finally. Category 1 set the topic as ‘content and function’ and accounted for 11%. The topic contained the most keywords and was confirmed by text, and consumers often evaluated their software use experiences with the topic, so the topic was highly related. Category 2, “Car Audio Connection,” possessed 10% of the frequency, and keywords such as Android, Bluetooth, CarPlay, Speakers, Audio, Car Audio, Support, Smartphones, Navigation, Vehicles, Apple, and Cars were found to be common attributes. Category 3 accounted for 2% under the topic of ‘smart Product Connection.’ Many contents mainly were related to song playback. ‘Life’, set as Category 4, constituted 20%, and the summarized main keywords were a Song, Video, Album, Life, Impression, and Memory. Category 5, ‘singer’ accounted for 1% of the frequency, and Kang Daniel and Voting were derived as keywords. Category 6 was set as ‘Price’, and the ratio was 8% of the frequency. The main keywords were Benefits, Vouchers, Points, and Ratings. Keywords and main keywords for each category were shown in the following table.

<Table 4> Results of Centralities in MelOn

<Table 5> Keywords Classified by CONCOR Analysis in MelOn

4.3 Result of Analyzing Spotify Data

The results of frequency analysis of the Spotify data used were as follows. The result showed that Spotify had a high frequency of songs, uses, likes, company, and BTS (<Table 6>). <Table 7> shows the analysis result of TF-IDF.

<Table 6> Top 50 Frequency Keywords in Spotify

<Table 7> Analyzing TF-IDF in Spotify

<Table 8> Results of Centralities in Spotify

<Table 9> Keywords Classified by CONCOR Analysis in Spotify

The results of each centrality analysis of Spotify are as follows. As with Melon’s centrality analysis results, Spotify was found to have no significant difference in Proximity- and Mediation- centrality. The results of Spotify’s CONCOR analysis were as follows. Spotify was classified into five categories. Category 1 was ‘Content and Function’, Category 2 was ‘Music Conflict’, Category 3 was ‘Hobby Life’, Category 4 was ‘singer’, and Category 5 was ‘stock Investment’, and the details of each category were as follows.

4.4 Result of Comparing Emotional Vocabulary

Among the results of MelOn and Spotify, the adjective form was extracted, and the overall emotional score was calculated by using Textom’s emotional vocabulary analysis. MelOn was summed with 75.71% positive adjective frequency, 74.59 emotional intensity score, 77.15% positive adjective frequency for Spotify, and 76.63 emotional intensity score. Spotify had a slightly higher positive evaluation than MelOn. In terms of detailed emotions, MelOn’s “likes” and “rejection” were higher than Spotify while Spotify’s “interest” and “pleasure” were higher than MelOn’s.

<Table 10> Result of Analyzing Emotional Vocabulary

<Table 11> Result of Detailed Emotion Vocabulary

4.5 Keywords and Comparative Analysis of Emotional Vocabulary

Based on the results of the CONCOR analysis, the main keywords of MelOn and Spotify were summarized and classified again, and the following categories and keywords were extracted.

Textom’s emotional dictionary was used to calculate the adjective form and emotional intensity score, and the emotional vocabulary results of the keywords are as follows. MelOn was higher in the “device connection” category than Spotify. Spotify scored higher than MelOn in the categories of “content and function,” “musical taste,” and “price.” However, the overall average of MelOn was slightly higher than Spotify. The result was different from the emotional vocabulary analysis.

We analyzed a keyword comparison between them by using the Word2Vec model. We judged to exclude “content and function” and “device connection” because the keywords had high congestion and could not be extracted meaningful keywords. Finally, an analysis focusing on “musical taste” and “price” was conducted based on the result. The analysis results were as follows.

<Table 12> Category and Keyword after Analyzing CONCOR

<Table 13> Emotional Vocabulary Analysis Result for Each Keyword

<Table 14> Result of Word2Vec in Musical Taste Category

<Table 15> Result of Word2Vec in Price Category

5. Conclusion

5.1 Research Conclusion

This study analyzed users’ attitudes and emotions toward MelOn, a leading Korean company that provides online music services, and Spotify, the world-leading company recently launched services in Korea. The analysis results are as follows. First of all, the main interests of users using MelOn were content and function, car audio connection, and life. In particular, car audio connection and life were found to have 32% of lavish attention, which could be interpreted as meaning that melon was used in everyday life for Korean users. MelOn does not simply exist as an online music service app, but MelOn has competitiveness in the market by establishing itself as an essential app in everyday life. However, MelOn was found that users had a higher negative attitude than Spotify. Despite MelOn having a high market share and long-term service in the Korean market, rejection, fear, and anger of users’ negative attitudes were higher than Spotify. In the case of “anger,” MelOn was about twice as high as Spotify. In other words, it means that MelOn users had an extremely negative attitude toward the service, which was judged to be actively analyzed and reflected in Melon’s customer management or strategy in the future.

Second, as a result of Spotify, the main interests of users were content and function, music conflict, and stock investment. In the case of MelOn, connection with various IT devices or automobiles was one of the main concerns. However, Spotify users were highly interested in the content itself. Besides, the users recently expressed high interest in the conflict over music sources supplied by Korean music producers. While the music conflict happened, Korean users had considerable limitations to listening to domestic music. Therefore, the music conflict is an acute problem for Korean users. An unusual point for the result of Spotify was that the users showed high interest in stock investment, not related to music. The point is a possibility that the users experience the service with positive emotion. Based on the emotion, they could judge that the value of Spotify would be increased. The users were found to have more positive emotions about the service than MelOn. In particular, interest and joy were higher than MelOn, and rejection was also lower than MelOn. From the perspective of these results, it could be interpreted that Spotify provides sufficiently attractive services to Korean users.

Third, due to emotional intensity analysis through Word2vec analysis, MelOn users were mainly related to domestic music sources such as IU, Indie, and guysome. In the case of Spotify, topics related to overseas music sources such as Mariah Carey, Justin Bieber, and Billboard chart. In the price category, users thought that MelOn and Spotify were both expensive, and there were various experiences such as free experience period and price. Among the detailed categories, in the case of Melon, results directly related to service interruption such as termination, termination application, and mobile termination. The result indicates that MelOn users’ satisfaction is lower than they think of the quality of service. It could be interpreted that the customer churn rate would be increased in the future. Therefore, MelOn is needed a detailed analysis to solidify its position in the domestic market, and Spotify is required to secure domestic customers through a strategic approach to the service desired by Korean users.

5.2 Research Implication and Limitation

The implication of this study is as follows. First, previous studies were conducted by analyzing Spotify and MelOn, respectively. However, this study is meant to conduct a comparative analysis to derive the results of the difference between the company’s strategy and service provision. Second, this study is significant because it was approached from the user’s point of view, not from the technical point. According to the difference in the company’s services, the users’ emotions were directly analyzed through reviews left by the users. The results of this study are believed to be helpful in practice for companies that provide online music services.

The limitations of this study are as follows. First, when collecting data on MelOn, we set up the keyword as the MelOn App. Melon is a homonym with a melon meant a fruit. Therefore, when we searched the MelOn app, it is a possibility and limitation that information meaning fruit melon could be derived or duplicated. Through several processes, we attempted more detailed classification. However, the process needs to be classified more clearly and accurately in future studies. Second, the user’s evaluation of the app is multilateral. However, in this study, only review data was used for analysis. Therefore, it has a limitation to reflect the multilateral perspective of the user. It seems necessary to collect and analyze multilateral data of users in future studies. In addition, in the case of Spotify, it has not much data on reviews accumulated so far, so it is judged that research should be conducted based on more data in the future.