Preamble

A pre-occupation of the reports collected here to date has been the K-Wave outside of Korea. We have taken both comparative perspectives on various countries and engaged with the situation in individual countries, too, for example the UK and the USA. While the K-Wave is a phenomenon predicated upon engagement with Korean popular culture on the part of people residing overseas, it has not escaped notice in South Korea. This month, we draw upon a large corpus of South Korean press articles to see what insights into the trajectory of public discourse surrounding the K-Wave can be gained from an examination of simple frequency, then put that into comparative perspective by examining the frequencies of related terms in a large collection of English-language book publications. As a contrast to last month’s exploration of how simple visual interpretation of Sankey diagrams can supplement interpretation, this month we direct our attention to the possibility that a common visualisation technique can hinder the accurate interpretation of underlying data.

Data

The Korean-language data we use is the 물결 21 corpus (hereafter the Trends 21 Corpus), accessed through its online tool, which can be found here. This is a collection of texts built from roughly 3,300,000 news articles from four South Korean newspapers of record. That is, including their often idiosyncratic Romanisations/translations, 동아일보 (or Dong-A Ilbo), 조선일보 (or Chosun Ilbo), 중앙일보 (or JoongAng Ilbo), and 한겨레신문 (or Hankyoreh Sinmun). Due to the characteristics of Korean orthography, it is not entirely appropriate to count the number of words as we might for an English-language corpus. Taking the number of eojeol as an analogy, that is, orthographic units surrounded by whitespace that may consist of complex units such as combinations of nouns and particle, we arrive at a total or 840,000,000 eojeol. This quantity of data should be sufficient to draw robust conclusions about the press discourse around the K-Wave over the time period covered by this corpus: 2000 to 2019.

The English-language data comes from the Google NGram Viewer. This consists of over 500,000,000 books dated from between 1500 and 2022. These books cover a range of languages outside of English including, for example, Chinese and German. English, however, makes up the majority of this data. As of 2011, the English sub-corpus of earlier version of this data contained roughly 361 billion words. Here, however, we restrict ourselves to the same time period as that covered by the Trends 21 Corpus. We are not able to ascertain the precise amount of data this covers. While we consider it reasonable to compare these two sources to gain insights into the relative chronologies and degrees of public interest in the K-Wave in Korea and the wider Anglosphere, we acknowledge that interpretation is somewhat complicated by the Trends 21 Corpus’ exclusive focus on select publications of a single type.

Findings

The relative and absolute frequencies of the term ‘한류’ (hallyu) in the Trends 21 Corpus are visualised below.

The relative frequencies should be understood as the frequency of the term per million eojeol. The two visualisations presented here appear strikingly different, but this is in fact an artifact of the way we have fitted a line to the underlying data. The line chart for the absolute frequencies simply joins the points, whereas the line seen on the visualisation of relative frequencies uses local regression to fit a smooth curve through the points. We present it here to highlight the effect that such smoothing can have on the interpretation of visualisations. Focusing on the points reveals a relative frequency with a distribution very similar to that of absolute frequency over time.

Here, we restrict our discussion to what the distribution of the absolute frequency of appearance of the term ‘한류’ in this corpus could tell us about K-Wave discourse in the South Korean press. First, three distinct peaks are visible in the data in 2005, 2011/12, and 2016. These peaks can be seen as coinciding with some high-profile developments in the K-Wave. 2005 is temporally proximate to the repeated broadcasts of the drama 겨울연가 (English Winter Sonata) in Japan, first on the NHK satellite channel, then on its general terrestrial channel. The third broadcast of the series, the first on the general channel, finished airing there only in the second half of 2004 (details here). This can be seen as the inception of the K-Wave as a regional phenomenon in East Asia as a whole, rather than ‘merely’ an increase in engagement with South Korean pop culture in China. 2011 and 2012 were significant for the mainstreaming of the K-Wave outside of East Asia. Notable events included prominent acts such as Girls’ Generation touring in Europe in 2011 and the phenomenal success of 2012’s Gangnam Style on Youtube. Finally, 2016 saw the deployment of the USA’s THAAD missile defense system in South Korea which has been linked with the quasi-official prohibition on some aspects of engagement with Korean popular culture in China. These events may all be considered newsworthy in their own rights, but the increased frequency with which the terms ‘한류’ appears at these times is highly suggestive of these separate events being framed either as part of or in relation to the single phenomenon of the K-Wave.

The relative frequencies of the terms ‘Korean Wave’ and ‘Hallyu’ (both without regard for variable capitalisation) in the Google Ngrams Viewer data are visualised below. The original search terms and an alternative visualisation using the default settings of Google Ngram Viewer can be found here.

The relative frequency in this case is the frequency of the terms as a percentage of the total number of words in the corpus.

The trends discernible in the Trends 21 Corpus are far less pronounced for our English-language data. While we see the frequency of both the terms rise to between 2000 and 2005, we do not see a comparable dip in frequency over the following years. We do, however, see that 2011/12 is an inflection point for the terms and their frequency markedly increases from this period. It is conceivable that this represents the growth in interest and engagement with the K-Wave outside of East Asia, specifically in the Anglosphere, to which the contemporaneous spike in press discourse surrounding the K-Wave in South Korea could plausibly be attributed. This is followed by a striking departure from the Korean data. Rather than further dips and peaks, we see the steady increase in frequency of both terms, with only ‘Hallyu’ appearing to tail off towards the end of the surveyed period. Thus, we may tentatively conclude that the sustained high-level of engagement with South Korean popular culture outside of Korea has become the status quo there, rather than something which is sufficiently novel to merit a great deal of news coverage. Conversely, the phenomenon, and potentially the media, products, and concepts which underpin it, are continuing to accrue ever higher levels of engagement overseas, particularly in the Anglosphere.

This, at least, is a plausible interpretation interpretation of the visualisation of our ‘smoothed’ lines. It is, however, interesting note a detail which is obscured by the above visualisations and, consequently, encourages us to greater circumspection in drawing definitive conclusions. According to the underlying data the term ‘Hallyu’ appeared with greater frequency than the term ‘Korean Wave’ in the years 2007, 2014, 2015, and 2019. This detail is also not discernible when the visualisation presented using the default settings of the Google Ngram Viewer is examined. Below, we present a ‘bump chart’, which instead emphasises the changes in which of the terms is used most frequently.

Conclusion

Examining the changing frequency of terms in Korean and English-language big data corpora grants us some insight into the similarities and differences surrounding discourse on the K-Wave in Korea and the Anglosphere. The obvious commonality is that the frequency with which ‘한류’ or its English analogues appear in the corpora has increased greatly over the first twenty years of the twenty first century, arguably in parallel with the growth in international engagement with South Korean popular culture. Reporting on ‘한류’ in the newspapers of record in South Korea, though, has fluctuated to both a greater degree and over shorter timescales than engagement with the K-Wave in English-language publications. While the period from 2006 to around 2012 did not see a great increase in frequency in the appearance of the terms ‘Korean Wave’ or ‘Hallyu’, there was certainly no appearance of a decline in their frequency comparable with that observed in the Korean-language data. Whether this should be attributed to the nature of the texts included in the corpora (ie., one might expect news to be more responsive to public interest and current events than other types of publication) is a possibility that cannot be ruled out.

We further considered the limitations of arriving at definitive interpretations of underlying data on the basis solely of visualisations. In many circumstances, the ability of visualisations to be manipulated to highlight particular aspects of data or patterns found therein is a considerable advantage for the efficient communication of concepts that would be more challenging to convey in text. In so doing, however, it is important to be aware of what can be obscured, lost, or misrepresented by even very widely-used, commonly-accepted visualisation techniques.

The significance of the changing relative frequency of the terms ‘Hallyu’ and ‘Korean Wave’ in the English-language data is not yet established. To gain some insight into whether there is any relationship between engagment with the K-Wave and the relative frequency the Korean loanword and localised calques it will be instructive to draw a comparison between two Anglosphere nations which differ in terms of their engagement with the K-Wave (i.e., the UK and the USA) and supplement that with a look at such comparative terminology in other European languages. That is precisely what we aim to do next month.

Acknowledgement
This work was supported by the Core University Program for Korean Studies of the Ministry of Education of the Republic of Korea and Korean Studies Promotion Service at the Academy of Korean Studies (AKS-2021-OLU-2250004)