Quantitative Analyses of Chinese Poetry of Tang and Song Dynasties: Using Changing Colors and Innovative Terms as Examples

1. Introduction¹

Tang (618-907 AD) and Song (960-1279) dynasties are two very important periods in the development of Chinese literary. The majority forms of the poetry in Tang and Song were Shi (詩) and Ci (詞), respectively. Tang Shi and Song Ci established crucial foundations of the Chinese literature, and their influences in both literary works and daily lives of the Chinese communities last until today.

Recognizing the importance of Tang Shi, a Chinese emperor of the Qing dynasty (1644-1912), Kangxi, ordered to compile a collection of Tang poems, Quan-Tang-Shi ( QTS, 全唐詩). QTS contains nearly 50 thousand works of about 2200 poets. A similar effort for compiling a collection of Song Ci from the private sector began in the Ming dynasty (1368-1644), and achieved a collection called Quan-Song-Ci ( QSC, 全宋詞) in the early Republican period of China (ca. 1937). QSC contains around 20 thousand works of about 1330 poets. The exact statistics about QTS and QSC may vary slightly depending on the sources.

In the past more than a thousand years, literary and linguistic researchers have had done a myriad of research about the poetry of the Tang and Song dynasties. Hence, it is beyond our capacity and not our objective to review the literature in this abstract. Traditional researchers studied and compared poetic works that were produced by different authors and in different time periods to produce insightful and invaluable analyses and commentaries (Tao, 1999). Most of the time, the researchers focused on the poems of selected poets. Even when computing supports become available, studying poems of specific poets (Jiang, 2003) is still an important and popular type of research in poetry.

Software tools facilitate the analysis of poetry from a panoramic perspective, and may lead to applications that would be very challenging in the past. For instance, Zhou and his colleagues (2009) analyzed the contents of collected couplets and Tang poems for creating couplets. Yan and his colleagues (2013) considered topic modeling in automatic composition of Chinese poetry. Lee (2012) concentrated on the linguistic analysis and teaching of QTS.

In this presentation, we will discuss some interesting findings in some quantitative analyses of QTS and QSC.

2. Colors and Imageries

Colors are an important ingredient in everyday lives, and actually carried important meanings in religion and social statuses in pre-modern Chinese societies. Wong (2011), a specialist in colors, discussed hidden meanings of colors in China in his book on “The colors of China”.

The beauty and imageries conveyed in the poetry originate from the collocations of the written words in the poems. Lo (2008) and Huang (2004) attempted to classify terms in the poetry by their semantic categories. The results can then serve as a foundation for analyzing the imageries hidden in the poems. Liu and his colleagues (2015) emphasized that colors play a crucial role in painting the imaginaries of poems: “colors in poems are like audios in movies”, and they analyzed the words related to colors in QTS.

Using methods for text analysis, one may analyze occurrences and collocations of colors in QTS and QCS. The main contribution of our work will be illustrating meaningful applications of text analysis for linguistics and literary. The collocations of color words that appeared frequently in QTS are certainly interesting (Liu, 2015). Yet, the analysis can be extended in at least two directions. First, did a poet have specific preferences on some collocations? Second, how were the collocations used by different authors?

Bai Juyi ² has the largest number of works in QTS. He used “白髮” ³ with “青衫” ³, “青雲” ³, “丹砂” ³, and “青山” ³, and “白首” ³ with “青山” and “紅塵” ³ relatively often. Liu Changqin ² , another important poet in the mid Tang period, used “白髮” with “滄洲” ³, and “白首” with “青山”, “滄洲”, and “青春” ³ relatively often. These different words and collocations convey the imagery of “aging”, and the variations in the word choices shed light on the subtle differences between the poets about how they expressed emotions about aging.

“白雲” ⁴ is a very frequent word in QTS. Collocating with different words would create different imageries in the poems, e.g., “黃葉” ⁵, “滄海” ⁶, “清露” ⁷, and “流水” ⁸. Each of these collocations may brew a different scene in readers’ minds, and some of these collocations are more popular than others. It should be interesting for researchers to extract the source poems ⁹ from the QTS to thoroughly study them.

Liu et al. (2015) reported that white (“白”) is the most frequent color in QTS. We may check and find that red (“紅”) is the most frequent color in QCS. It is interesting to investigate the changes (and their causes) of the popular colors from QTS to QCS. Again, using text analysis methods, we can find a good approximation of the trend, though obtaining the precise frequencies of the colors requires the techniques of word sense disambiguation (WSD). For instance, “金” ¹⁰ could represent a material or a color, so WSD is necessary to achieve precise statistics.

In QSC, the most frequent six colors are in the order of “紅”, “青”, “黃”, “綠”, “白”, and “碧” ¹¹, while, in QTS, the most frequent six colors are “白”, “青”, “紅”, “黃”, “碧”, and “綠”. “紅塵”, “殘紅”, “紅妝”, “紅葉”, “紅袖”, “紅日”, and “紅樓” ¹² are some of the most frequent red words in QSC. The changes in the dominant colors from QTS to QSC may be a result of the selection process and may be a result of the cultural shift, and is an academically interesting issue to purse further.

3. Word Inventions and Influences

Liu and Wang (2012) proposed a method to measure and compare the influences of the poems of a poet. Their methods considered whether or not the poems were selected to be included in famous collections.

While our goal is not to challenge Liu and Wang’s viewpoint, we would propose to consider also whether poets created new words that were used by later Chinese generations. Isn’t it practically meaningful and academically significant to create new words that future generations continue to use? At the time of Tang and Song, poets were at an excellent stage of the Chinese history to achieve such a cultural impact.

We conduct an analysis of frequent bigrams in QTS and QSC, and compare the differences. Words that appeared only in QSC are candidates of new words which were invented in Song dynasty. Words that appeared only in QTS are candidates of words that failed to survive in Chinese language. Although this process does not really guarantee a water-proof theoretical foundation for word invention, the findings should still serve as a persuasive factor in linguistic, literary, and historical research.

Here are some of such findings. “紅塵” ¹³ appeared in both QTS and QSC. “惺忪” ¹³ is a word that appeared in QSC ¹⁴ but not in QTS, and is still being used in modern Chinese. “空門” ¹³ is a word that appeared in QTS but not in QSC, and is being used in Chinese. “武皇” ¹³ is a word that appeared in QTS but not in QSC, and is not normally used in Chinese. “酴醾” ¹⁵ represents another type of instance. This word appeared much often in QSC than in QTS.

4. Discussions

A major challenge in analyzing the words in Chinese poetry is word segmentation. Traditional experience indicates that most words in poetry consist of one or two characters (2005). Relying on this heuristic, we can algorithmically analyze the corpora containing about 4.9 million characters at least approximately. To really understand and appreciate the poetry, one should not read them verbatim. Metaphor recognition can be essential for revealing the real intentions of the poets.

Observations resulting from quantitative analysis of QTS and QSC open windows to promising research opportunities about QTS, QSC, and their transition. In addition to colors, terms about astronomical objects, floral entities, meteorological phenomena, and geographical sights, are important participants in poetry. Innovative collocations of them paint impressive imageries in readers’ minds. Good computational tools can help researchers explore poets’ worlds more efficiently.

5. Responses to the Reviews

We are thankful for the precious critiques and comments of the DH2016 reviewers. Our responses, which could not exceed 300 words, for the comments cannot fit into this final version because of length constraints, so we place our complete responses in a separate file online at < http://www.cs.nccu.edu.tw/~chaolin/papers/dh2016liu.responses.pdf>.