This article analyzes the occurrence of coordinative conjunctions in five spoken corpora and five written corpora by utilizing the “Comprehensive Searching System: KOTONOHA,” a function of the corpus searching web-application “Chuunagon.” Principal component analysis was conducted on the results of the corpus survey in order to visualize the relationship among 10 genres as well as to describe each conjunction. Consequently, 10 genres were divided into 4 clusters. “Workplace Conversation” and “Simulated Public Speaking” are in one cluster, while the two conversation corpus known as NUCC and CEJC belong to another cluster. “Academic Presentation Speech” is in the same cluster as “Newspaper” or “Minutes of the Diet.” Furthermore, four groups of conjunctions which are specific to each cluster are found to correspond to the four groups of conjunctions in previous studies that analyzed conjunctions from the perspective of meaning and usage.

View full abstract

In this paper, we first presented an overview of the "Data for Word Stylistic Value" which shows continuous numerical values for the degree of stylistic differences (written hard style/spoken soft style) of many words. Subsequently, as an example of the use of this data, we attempted to estimate the "style of text" based on the "style of word" using a numerical value called the "text stylistic value." The texts used for the estimation were the core data of the "The Balanced Corpus of Contemporary Written Japanese" and were able to reproduce the stylistic characteristics of each register. Through this, we have shown that the "Data for Word Stylistic Value" and the "text stylistic value" can be used effectively for stylistic research, and that the interdependence between "style of word" and "style of text" can be empirically analyzed using quantitative methods.

View full abstract

The purpose of the present study is to explore the chronological variation of demonstrative pronouns using the Corpus of Historical Japanese. As the forms of demonstrative pronouns change with time, this study utilized the classification based on the word-initial (ko-, so-, ka-, and a-) and word-final forms (zero, -no, -re, -ko, -nata, and -chi). The locally weighted (LOESS) smoothing regression was performed to investigate the frequency change patterns of these word forms from the Nara to the Taisho period in Japan. The results show that (1) the total number of demonstrative pronouns increased, (2) the so- form increased, the ko- and a- forms were frequent in the 1700s and 1800s, and the ka- form increased after 1800, and (3) the zero form decreased, the -re form increased, the -no form increased after 1700, and the -ko, -nata, and -chi forms decreased in the early modern period after an increase in the Edo period.

View full abstract

The purpose of this paper is to clarify the characteristics of the style of weblog and Twitter. The results of the examination of the ratio of parts of speech and the ratio of parts of lexical stratification of weblog confirm that the style of weblog is extremely close to that of spoken language. In the case of Twitter, its style is, based on its ratio of parts of speech, considered to be similar to that of the written language. But it is reasonable to interpret this as an effect of the limited number of letters and characters used there. As concerns the ratio of parts of lexical stratification of Twitter, the ratio of native Japanese words ranks higher: Colloquial or slang words, words that emphasize the writer's feelings, or words that evaluate things are used frequently. The effect of web mark in the style was reexamined by logit transformation of data, and its result showed that in both weblog and Twitter, sentences with web mark were highly rated for adjectival nouns and interjections, while sentences without web mark were highly rated for prenoun adjectivals and conjunctions.

View full abstract

In this study, we analyzed the stylistic changes in sentence endings in novels by Soseki Natsume. Random forest, principal component analysis, and linear regression analysis were used to analyze the incidence of words appearing at the end of the sentences. Results showed that a change in the quantitative features of sentence endings occurred around 1908. In particular, seven works after "Sorekara," published in 1909, showed a similar trend. In addition, a consideration of eigenvectors in the principal component analysis revealed that the stylistic change in sentence endings recognized in 1908 was a marked increase in the use of the auxiliary verb "ta."

View full abstract

The Modifier-Verb Ratio (MVR) is a widely used index for assessing stylistic features of Japanese texts in terms of part-of-speech composition. It is unclear how such elements are related to impressions resulting from reading a text. A questionnaire survey examined the relationship between MVR/part-of-speech rates and subjective impressions. Multilevel analysis indicated that noun and verb rates negatively affected imaginability, which was not the case for the modifier rate. We found an interaction between different types of part-of-speech by replacing the part-of-speech rates with the mean part-of-speech rates of sentences, which suggested that the more adjectives authors use, the easier it is for readers to imagine the scene depicted by texts, but only in that specific part-of-speech compositions. On the other hand, we did not find any promising explanatory variables for the impression of the speed of an unfolding story. We have discussed the relationship between part-of-speech composition and stylistic impressions, the significance of focusing on the part-of-speech composition of sentences, and reexamined the working definition based on this discussion.

View full abstract

In this study, I conducted a stylistic typology of literary works from the Kamakura period, including wabun ‘native-style literature’ and gunki mono ‘war chronicles’, which are considered to be a type of wakan konkōbun ‘Japanese-Chinese hybrid style literature’, based on the properties of sentence predicates and a comparison with the results of Okawa (2020), using correspondence analysis and cluster analysis. As a result, I found that Kamakura period wabun could be divided into two types: 1. standard kikōbun ‘travelogues’ and wakashū ‘waka anthologies’ and 2. wabun and kikōbun and wakashū with strong stylistic features. Gunki mono were grouped together with kanbun kundoku ‘vernacular reading of Literary Sinitic’-inspired literature and diaries, revealing that such chronicles have stylistic characteristics similar to works such as "Konjaku Monogatari-shu" in terms of opposition of wa versus kan, and in terms of genre-related style.

View full abstract

In order to analyze the stylistic characteristics of Ryūnosuke Akutagawa's children's literature, this study employs eight children's literature of Akutagawa as the research object. This research makes a quantitative analysis of lexical richness, lexical difficulty, sentence length, and syntactic complexity. Three research findings reveal that: Firstly, in terms of lexicon, the use of the words is simple in Akutagawa's children's literature, and the vocabulary richness is relatively low. Secondly, in terms of sentences, the syntactic complexity and average sentence length of Akutagawa's children's literature are lower than those of adult literature. Thirdly, it is clear that the stylistic characteristics of Akutagawa's children's literature reflect some extent the stylistic characteristics of children's literature in general. Therefore, the reading difficulty of Akutagawa's children's literature is lower than that of adult literature. This may be due to the fact that the author assumed his own children, his relatives' children, and the doctor's child as their direct readers.

View full abstract

When analyzing qualitative data, it is important to present them in the form of crosstabulation tables so that the procedures used are evident. This paper introduces the terminology involved in crosstabulation tables and then presents the methods for testing crosstabulation tables with different conditions and several categories. Crosstabulation tables of two conditions (with and without correspondence) and i × j tables (without correspondence) are used. Additionally, the adverse effects of category mergers are discussed, including (1) increasing the quantity of data so that the number of data in each cell exceeds a certain number and (2) employing Fisher’s exact probability test as a method for dealing with cases, in which it is inappropriate to apply the chi-square test.

View full abstract