This paper has been developed from a paper presented at the 2019 TEI Conference. The authors wish to express their gratitude to Nobutake Kamiya, Naoki Kokaze, So Miyagawa, and Wang Yifan for their input in completion of the paper and the related proposal to the TEI Technical Council. Kazuhiro Okada wishes to express his gratitude for support from the Introduction of TEI to Japanese Pre-modern Texts project at the National Institute of Japanese Literature; Satoru Nakamura was supported by JSPS KAKENHI Grant Number JP19K20626; Kiyonori Nagasaki was supported by JSPS KAKENHI Grant Numbers JP19H00516 and JP19H00526. The authors are sincerely grateful for the concepts established by W3C (2001; 2013; 2014) and the clear guide developed by Ishida (2016).
- 1 This paper was originally submitted to jTEI before submission of the proposal to the TEI-C Technica (...)
1Glossing is a custom found in most medieval manuscript cultures. In those cultures, phrases or words written in classical languages are explained by marginalia, interlinear glosses, and the like, normally in a vernacular language. Whereas most of these traditions have been abandoned and have not been incorporated into modern writing systems, glosses in the Japanese writing system have continuously evolved over the centuries, and are still used today. The Japanese gloss is now called rubi ルビ, a Japanese rendition of the movable body type size “ruby.” This paper describes the necessity of the ruby as an element set in the TEI Guidelines by explaining its characteristics. 1
2Other names for the ruby gloss include furigana 振り仮名 and bōkun 傍訓. Rubi was a slang term used by printers in the early twentieth century, derived from the name of a size of movable type imported from England, and it is now a standard term. In Japanese linguistics literature in Japanese, the term furigana is preferred, as in Ōshima (1989), Konno (2009), and Yada (2018). In this article, we consistently use the term “ruby,” following the convention adopted in markup languages.
3Whereas interlinear glosses are not found exclusively in the Japanese writing system, and similar phenomena can be observed in other East Asian material, the Japanese ruby is surely among the most complex glosses. A ruby basically functions as a phonetic gloss to an attached Chinese character, but it is not just a pronunciation aid; it can show multilayered reading or even encode the main voice of the text (see section 3.2 for a detailed discussion). Encoding multilayered ruby glosses within the existing TEI framework had been a challenge.
4A ruby gloss is visually aligned to some portion of text, but the relationship between a ruby and the text which it glosses and which it is embedded in can be flexible. This flexibility allows certain types of ruby to introduce nuanced or rather polyphonic expressions. First, the structure of polyphonic texts is, we believe, different from that of the interlinear glosses familiar from Western texts, normally tagged with <add> or <note>, whose purpose is usually to clarify the meaning of the text or to provide no more than a simple note, a common strategy used for encoding interlinear glosses (see Estill, 2016). Second, ruby glosses can be structurally multilayered: as demonstrated below, rubies can be doubled and even tripled, thus piling up multiple voices. This complexity, which has hindered existing ruby markup, may be seen as odd. But from another perspective, it is a unique characteristic of the ruby gloss. This article explores ways to conceptualize these phenomena within a manageable structure.
5Modeling a culturally specific phenomenon with TEI in order to introduce elements into a public standard, rather than creating a customized model, is also supported by the history of attempts to mark up ruby glosses. In fact, the Japanese community has faced difficulty with ruby markup in various projects: see Kawase, Ichimura, and Ogiso (2013), on their Japanese literature corpus building, and Cong and Takaku (2018), on Tang poems in secondary education textbooks in Japan. While the former encodes the glossed text with <w> and the ruby text with @ana, the latter directly introduces a tagset from HTML. It indicates the difficulty of sharing methods of encoding interlinear glosses such as <note>, which are not explicitly included in the Guidelines. In this regard, encoding the ruby model directly within the TEI Guidelines not only provides a markup system natural to the text but accommodates other East Asian materials. This advance will also benefit similar glossing practices around the world.
6This paper is organized as follows: in section 3 we describe ruby examples from a historical perspective. The first half of section 4 demonstrates encoded examples, and the last half deals with an efficient way of modeling the ruby examples discussed. Section 5 discusses the proposal in light of other East Asian materials.
7This section briefly explores the ruby examples to be considered. Before going deeper, we provide a glossary to assist the discussion:
-
glossed text (base text; <rb>)
-
A part of text that is glossed. Usually in kanji, or Chinese characters, but not limited to them.
-
ruby text (gloss; <rt>)
-
A text that glosses the glossed text. In a plain text transcription, parenthesized text shows ruby text. Usually written in hiragana (or katakana for loanwords), but not exclusively (kanji, romaji, and also in Taiwan, bopomofo will appear).
-
ruby gloss (ruby glossed text; <ruby>)
-
A glossed text and its corresponding ruby text.
-
right- (top-) or left- (bottom-)sided ruby
-
A ruby gloss whose position relative to the glossed text is semantically meaningful. In Japanese, the default position is on the right side in a vertical text; in a horizontal text, the default position will be the top side. A ruby gloss placed on the other side implies a special case. A pair of ruby glosses that gloss a text from both right and left sides or top and bottom sides is called a double-sided ruby. In other writing systems the default position may differ. In the transcriptions in this paper, a ruby text in the default position precedes a text on the other side.
-
kanji (Chinese character)
-
A script borrowed from the Chinese writing system for writing Japanese. Roughly speaking, it denotes either a word of Chinese origin, or the morphemic root of a Japanese native content word.
-
kana
-
Kana was developed from a phonological usage of kanji to write native words. It is also a cover term for hiragana and katakana (see below).
-
hiragana
-
Hiragana is a cursive development of kana. Among its uses in the contemporary Japanese writing system, it denotes the agglutinative part of the language, which cannot be fully specified with kanji. To illustrate the idea with English, given the word stronger, the root part, strong, would be written in kanji and the rest in hiragana. Hiragana is also preferred when use of kanji seems not to be suitable.
-
katakana
-
Katakana is developed from components or fragments of a phonological usage of kanji. It was a major component of the glossing system of Classical Chinese writing, and today katakana signals that the expression is special. It is usually used for modern words borrowed from Western languages and for words with colloquial nuances.
8A description of ruby gloss patterns and backgrounds now follows. An adequate history of the ruby gloss would require a book-length inquiry, but in this treatment we are going to concentrate on the history of its structure and related matters. For simplicity, we will first review modern Japanese ruby glosses. Then we will proceed to historical ones, and lastly to Taiwanese bopomofo examples.
9In modern Japanese texts, ruby glosses are used to gloss Chinese characters with kana syllabaries. The relationship between glossed and ruby texts have two general patterns that can be distinguished: either (A) the gloss is subject to the glossed text, or (B) the opposite. Ōshima (1989) summarizes the possible ruby gloss patterns as follows:
-
A-a.
Specifying phonation, pattern 1: to prevent confusion with another word with the same orthography (such as Ōgoto 大事[おおごと] with Daiji, Ani-san 兄[あに]さん with Nii-san, or Hinichi 日日[ひにち] with Hibi).
-
A-b.
Specifying phonation, pattern 2: to supply an accented reading (such as Senso 戦争[せんそ] with Standard Japanese Sensō or Chau 違[ちゃ]う with Chigau).
-
A-c.
Glossing a reading, pattern 1: to assist in the learning of new words. Orthography and pronunciation of unfamiliar words are presented.
-
A-d.
Glossing a reading, pattern 2: to explain unfamiliar orthography. The reader would already know the word, but under a different orthography.
-
B-e.
Specifying the meaning of the kana orthography. When the orthography is irregular for the gloss, as in rabirinsu 迷宮(ラビリンス) for meikyū, hito 女(ひと) for on’na, and hairu 侵(はい)る for okasu.2
-
B-f.
Specifying the items referred to by the gloss. Achira 大陸(あちら) for Tairiku and Ani-san 伊兵衛(あに)さん for Ibē san.
Figure 1. Original passage of Ōshima (1989).
- 3 Translation ours with slight modifications in numbering and illustrations.
(Ōshima 1989, 72–733)
10Ōshima (1989) adds other cases, for specifying the meaning of the Chinese character orthography (tabū 禁忌[タブー] [“taboo”] for kinki [“taboo” or “contraindication”]), and the equal (or rather independent) relationship between the gloss and the glossed text (Wocchingu 再発見[ウォッチング] [“watching”] for saihakken [“rediscovery”]). We could add more patterns of glossing other language expressions. Here, what is important is that the uniform structure of the ruby gloss allows all of the above patterns. In addition, these patterns are not exclusive.
11Contemporary Japanese texts do not have ruby glosses unless they are needed. However, as seen in the above patterns, especially patterns A-b, B-e, and B-f, ruby glosses can be a device for rich expressions. Moreover, occasionally the text is glossed on both the right and left, or the top and bottom sides. The examples below illustrate the usual single- and double-sided ruby glosses.
- 4 Generally speaking, in the Japanese writing system, one kanji denotes a single morpheme or syllable (...)
12Note that single-sided ruby glosses have typographical complexities that are structurally straightforward. (As we will see later, HTML schemas have been concerned with typographical matters.) The examples below, figures 2, 3, and 4 (written top to bottom and right to left) show some typographical conventions when glossing words with complex orthography: 4
-
“真面目(ま/じ/め)” (from figure 2)
-
“浴衣(ゆかた)” (from figure 3)
-
“あんたの/鼾(いびき)” (from figure 4)
Figure 2. Kawakami Hiromi. 1998. Yashi, Yashi 椰子・椰子, p. 21, detail. Tokyo: Shinchosha.
Figure 3. Kawakami Hiromi. 1998. Yashi, Yashi, p. 21, detail. Tokyo: Shinchosha.
Figure 4. Kawakami Hiromi. 1998. Yashi, Yashi, p. 29. Tokyo: Shinchosha.
13Unlike its appearance in the second (figure 3), the ruby text (“いびき”) is not glossed to the word “の.” This is because ruby glosses can be placed alongside a neighboring word if it would not be misleading, especially if the glossed text is in kanji and the neighboring words are in kana. This is not a structural but a typographical matter; see W3C (2012, §3.3.5, figs. 121, 122) for further illustrations.
14Yet another example of a single-sided ruby gloss is of the pattern II-e (figure 5, written top to bottom and right to left). Here the translator rendered the name of an Ainu divinity, kamuy’otopus, in Japanese characters that were equivalent to the Chinese characters “神髪彦(カムイオトプシ)の.” We note that again the ruby text is allowed to continue alongside a character that is not being glossed.
Figure 5. Kindaichi Kyosuke, editor and translator. 1936. Ainu jojishi yūkara アイヌ叙事詩ユーカラ [Yukar: Ainu Epics], p. 219. Tokyo: Iwanami Publishers.
15The next example is double-sided. In this example (figure 6), “大王(おおきみ)(だいおう),” the preferred reading ōkimi is presented as the top ruby text and another reading, daiō, is presented parenthesized as a bottom-sided ruby. It should be noted that, as this is a textbook of Japanese history for Japanese junior high schools, it has numerous interlinear glosses, birth and death years for example. However, these glosses are not tied with a particular text, but just function as pointers, explanations, or something else. We consider them not to be ruby glosses, but general annotations that fall under the semantics of <note>.
Figure 6. Ishii Susumu, Gomi Fumihiko, Sasayama Haruo, and Takano Toshihiko, editors. 2007. Shōsetsu nihonshi 詳説日本史 [Advanced Japanese History], p. 21. Tokyo: Yamakawa Shoten.
16Lastly, although not yet confirmed by a real example, a double-sided ruby could overlap and not nest. In a blog entry discussing complex ruby structures and their encodings, Uchida cites a constructed example presented by a now-deleted tweet:
まず、@koueiheiさんによるルビの例。「労働者階級意識」といふ親文字列のうち、「労働者階級」に「プロレタリアート」といふルビ。同時に「階級意識」に「かいきゅういしき」といふルビ。これは入れ子方式ではマーク付けできない。
もっとも、日本語でも英語でも「階級」と「意識」は2つの単語に分解できるんぢゃないか、したがって無理にcomplex rubyで扱ふ必要な無いんぢゃないか、といふ指摘もあらう。
(Uchida 2010)
Twitter user @koueihei (who had deleted their account at the time of writing) offers a glossed text that reads “労働者階級意識” —a phrase that is difficult to translate. In this example, @koueihei has added two ruby glosses: プロレタリアート (Proletariat) to the (upper?/right?) side of “労働者階級,” and “かいきゅういしき” to the (bottom?/left?) side of “階級意識” (translation of Klassenbewußtsein or class consciousness). The double-sided ruby could be visualized as in either figure 7 or figure 8.
Figure 7. A possible rendering of an example in Uchida (2010), where the vertical bar shows morpheme boundaries. In this variant, the bottom-side ruby corresponds with word divisions.
Figure 8. Another rendering of figure 6. In this variant, the bottom-side ruby corresponds with morpheme divisions.
17According to Uchida, word-to-morpheme alignment, as shown in figure 7, cannot be represented using nested structure. It is true, as Uchida also points out, that the glossed text consists of three morphemes, and that it does not need to encode both rubies as a single ruby gloss structure, as we can see from the glossing boundaries in figure 8. It is enough then to encode both rubies at the boundary of “階級” and “意識,” meaning that it is not necessarily an encoding problem. However, we can imagine a case where the left/bottom-side ruby was Klassenbewußtsein, and hence remains a problem for nested encoding. In any case, this example is surely one of the rarest structures among ruby glosses. But it is theoretically possible, and the possibility that examples will emerge warrants consideration of encoding strategies that can handle them.
- 5 Although the Japanese writing system adopts Chinese characters, the way it uses them cannot be summ (...)
18Historical ruby glosses are not too distant from modern ruby glosses. Ruby is originally derived from interlinear glosses to Classical Chinese texts, or Kanbun Kundoku (漢文訓読).5 A summary of the application of Japanese glosses to Classical Chinese texts can be found in Whitman (2011) and Kosukegawa (2014). See Whitman (2011) for a comparison with glosses in Latin manuscripts. Kosukegawa (2014) summarizes the Japanese glossing practice as follows:
-
The source texts of kundoku glossing were in classical Chinese.
-
…
-
Kundoku involves the act of directly adding glosses to a classical Chinese text.
-
- 6 Any language that has a tradition of reading Chinese characters is called Sinoxenic. Sinoxenic lang (...)
The method of adding glosses has both similarities and differences in each Sinoxenic language.6
(Kosukegawa 2014, 1)
The third point is particularly relevant to the ruby phenomenon. Here Kosukegawa (2014) translates the summarization of the Japanese glossing practice by Ishizuka (2001, 2) as “adding of annotations or symbols to aid in the comprehension of a classical Chinese [text] while reading it.” The purpose of glossing, not translating, is to preserve the original glossed text as it is. The purpose of Japanese ruby is practically the same: to leave the intended graphic expression in the glossed text while conveying the meaning.
19What distinguishes ruby glosses from glosses to the Classical Chinese text is that ruby glosses are applied to text in the same language. Yada (2018) identifies the earliest ruby gloss from late twentieth-century materials. Konno (2009, 93) demonstrates that B-e–type ruby glosses can be attested as early as the sixteenth century. Such glosses are as a rule not later additions but glosses by the authors themselves (see also Yada 2018, 803).
- 7 We do not say much about the practice of glossing Classical Chinese texts. However, it should be no (...)
20Double-sided ruby glosses flourished between the eighteenth and nineteenth centuries, when the Tokugawa Shogunate ruled the nation. Murakami (2017) claims that those double-sided ruby glosses emerged from the Sinological practices of Ogyū Sorai (1666–1728), who valued understanding texts through the living vernacular language rather than the vernacular-glossed Classical Japanese language. According to Murakami (2017), while the right side of the ruby gloss generally shows the pronunciation, the left side of the ruby gloss is a vernacular explanation of the glossed text.7 Murakami’s (2017) argument is important to account for the question as to why the complex glosses to Classical Chinese texts turn into the more modern double-sided rubies: the old glosses, sometimes successively accumulating, as a rule show independent interpretations and have nothing to do with position; in contrast, Sorai’s double-sided ruby glosses are systematically positioned. Sorai’s double-sided ruby glosses are intended to explain Classical Chinese texts in the vernacular Japanese language, and the same motivation lies behind double-sided rubies in Japanese texts.
- 8 This practice was a fruit of the adaptation of Chinese vernacular-influenced literature, called bái (...)
21Explanatory vernacular glosses also developed during the days of the Tokugawa Shogunate.8 In a limited timeframe from the eighteenth to the nineteenth century, publications for the general public in which most Chinese characters were glossed appeared widely. Yanaike (2009) claims that such glosses, rather than the glossed text itself, were what people actually read. Such blanket use of glossing was demanded because of the Japanese population’s low literacy in Chinese characters, and was possible because of printing: almost all Japanese publications were produced using woodblock printing. Only in the last decade of the nineteenth century did movable type printing supersede woodblock printing. The ruby gloss became uneconomical in movable type printing, and the introduction of modern compulsory education reduced the need for the glosses.
22The example below is a page from an early modern text, Hanabusa sōshi (figure 9). It is written top to bottom and right to left. Here, the line on the far left is to be transcribed:
にむかひ梁(りやう)武帝(ぶてい)の仏(ぶつ)に淫(いん)して民膏(みんかう)(たみのあぶら)を費(ついや)し国の衰(おとろ)へとなり
(Tsuga Teishō. (1749). Hanabusa sōshi 英草紙. Vol. 1, f. 3r. doi:10.20730/200016727.)
Figure 9. Tsuga Teishō. (1749). Hanabusa sōshi. Vol. 1, f. 3r. doi:10.20730/200016727.
National Institute of Japanese Literature, Japan. CC BY-SA 4.0.
In this example, a sequence of two characters 民膏, or the wealth of the people, is doubly glossed, being Sino-Japanese reading in an old-fashioned spelling “minkō みんかう” (cf. Mandarin mín gāo): (1) on the right side, with the native word explanation “tami no abura たみのあぶら” and (2) on the left side. The left-side gloss, which literally means “the fat of the people,” is a rather literal rendition of the glossed characters, meaning that it explains the characters rather than the word.
23The next example is taken from a Japanese song, Igirisu dodoitsu (figure 10, written top to bottom and right to left). The first line has triple ruby glosses:
まことリットル(りつとる)(少(すくな)き)
(Bōan Koikoi Sanjin. Late 18c. Gengo chūkai igirisu dodoitsu言語注解英語土渡逸. Vol. 1, f. 1v. doi:10.20730/200018433.)
Figure 10. Bōan Koikoi Sanjin. Late 18c. Gengo chūkai igirisu dodoitsu. Vol. 1, f. 1v. doi:10.20730/200018433.
National Institute of Japanese Literature, Japan. CC BY-SA 4.0.
In this example, the left-side gloss is itself glossed. The glossed text (1), in katakana script, is a rendition of “little.” An equivalent transcription in hiragana is given in the right-side ruby gloss (2), and a translation is given in the leftmost left-side ruby gloss (3). To the best of our knowledge, such complicated ruby glosses appear initially in the late eighteenth century and again briefly, around 1870. Although not unlikely in other eras, technically speaking a deeper gloss is possible. As TEI is intended for documenting rather than creating, we believe that a ruby schema should be capable of encoding historical examples as long as it does not complicate the schema itself.
24There is another writing system, besides the Japanese system, that has ruby-like glosses: bopomofo, used in Taiwan. Bopomofo, or the Zhùyīn Fúhào 注音符號 system, is a transcription system for Mandarin Chinese and other Chinese languages. Bopomofo was learned in order to gloss phonetically from Japanese kana, so we consider it one of the ruby gloss systems.
25Bopomofo has unique characteristics in ruby text positioning. It allows both top and right sides of the glossed text in horizontal writing, meaning that its default value for the ruby text position is not definable (Ministry of Education 2000; note that the right-hand position in horizontal writing is preferably called intercharacter position, a counterpart to the term interlinear position). The function of bopomofo, reflecting the monosyllabic nature of the Chinese language, is mostly to present the phonetic value of a glossed character. Metaphorical usage may technically arise, but it does not change the structure of bopomofo.
Figure 11. Horizontal right-sided (intercharacter) bopomofo example on the street.
nesnad. 2016. “File:Bopomofo on Taiwanese street - with English - Nov 2016 2.jpg.” In Wikimedia Commons. https://commons.wikimedia.org/wiki/File:Bopomofo_on_Taiwanese_street_-_with_English_-_Nov_2016_2.jpg
26Similar examples can be obtained from schoolbooks used in mainland China, but not as frequently as in Taiwan. In Korea and Vietnam, similar vernacular reading aids were applied to words of Chinese origin written in Chinese characters before modern writing systems emerged. The TEI ruby gloss encoding schema will help to encode Korean and Vietnamese texts with such glosses.
27The ruby gloss has long been a problem for Japanese text encoding. Within the XML framework, Hara and Yasunaga (2002) were among the first to address this problem. Their scope was, however, limited to the simple single-sided ruby. It is practically the same with Yamaguchi et al. (2011), whose goal was to build an XML vocabulary for Japanese linguistic corpora. They treated ruby text as an attribute for the ruby gloss element, which is useful only for the phonation gloss (types A-a, A-b).
28Such a simple structure works for most ruby glosses, but faces numerous problems. Taking an example from corpus building, if ruby text is going to be analyzed, how can it be described? It is impossible to tag an attribute value for a ruby text in XML. In fact, Murayama, Ogiso, and Nakamura (2017), building on the work of Yamaguchi et al. (2011), describe their difficulties in handling these values in their environment. In the same vein, a double-sided ruby can be described only as long as the ruby gloss element is allowed to nest, and the ruby text refers to the same glossed text. Kawase, Ichimura, and Ogiso (2013) attempted to introduce a TEI framework based on Yamaguchi et al. (2011). They were aware of this problem, and introduced HTML’s ruby vocabulary while calling for a more sophisticated structural approach in TEI. Later, Murayama, Ogiso, and Nakamura (2017) addressed this issue by introducing a new nesting left-side ruby element to their own schema. Cong and Takaku (2018) also introduce HTML’s ruby vocabulary to the TEI framework. However, HTML is not intended for documentation of the text, and this is a crucial limitation. Cong and Takaku (2018) also marked up ruby texts character by character, rather than at the morphemic level, which would break up the logical relationship of ruby and glossed text.
- 9 As acknowledged in W3C (2001), their work was inspired by Dürst (1997), in which ruby glosses were (...)
29The history of ruby encoding in HTML is marked by pro- and anti-SGML (XML) positions on structure. A ruby encoding system was first proposed in the earliest draft of W3C (2001) (which eventually became a module of XHTML 1.1),9 in late 1998, and implemented in 1999 by Microsoft informally in their Internet Explorer 5. In its earliest form, W3C (2001) accommodates only single-sided ruby; however, the final form also covers double-sided ruby. The published form defines the envelope <ruby> element, the container elements <rb>, <rt>, and respective <rbc> and <rtc> elements as the containers of <rb>s and <rt>s, with the compatibility-easing <rp> element, whose role was to give parentheses to unsupported implementations. One thing to note is that the first draft, which was produced during the transition from HTML to XHTML, allowed the omission of the end tags of <rb> and <rt>, but reversed that in the final form. However, the Internet Explorer implementation was based on an earlier draft of this document, and they did not follow the later version. This gap between W3C and implementers, although not limited to this small corner, has continued to cause practical issues until now, although W3C abandoned their development of XHTML 2.0. Another thing worth mentioning is that, in this schema, <rb> was allowed once per <ruby> element, and not allowed under the <rbc> element (see W3C 2013 for a comparison between approaches in XHTML and HTML5); therefore, if an encoder wished to divide the ruby text character by character (as shown in the examples of figures 2, 3, and 4), they needed to tag every unit of ruby-text correspondence, and ignore the semantic unit.
- 10 As for the encoding model after HTML5, note well that before W3C (2013) was published, HTML5 droppe (...)
30As an extension to HTML5, W3C (2014) set out to create ruby encoding that met the requirements of W3C (2012) with the styling component covered by CSS3 modules.10 This extension addressed the gap noted in W3C (2013), which investigated the ways in which W3C (2012) was insufficient to meet ruby-related criteria. Based on JISC (2004), W3C (2012) had tried to show the traditional typography standard that should be realized in modern browsers (Yabe 2010). In contrast with fixed layout settings, where traditional typography has worked, it was not an easy task to accomplish this in content-flow layout; content-flow allows adjusting the layout to fit everything from smart-phone browsers to large monitors for desktop PCs. Ruby glosses present a number of challenges for content-flow layout; to name but two examples: How and at what distance from the glossed text should ruby text be rendered, given a limited allotted boundary? What if a line break occurs in the middle of the ruby gloss? Kawabata (2014) reviews difficulties in rendering ruby glosses properly in different versions of HTML.
31The encoding models that were adopted in HTML5 (and those inherited by WHATWG’s HTML Living Standards) have, in principle, a tabular structure and an interleaving structure. Each has its own strength and weakness, in terms of logical structure. The interleaving structure provides a simple parsing mechanism and conforms to desktop publishing implementations. Most actual HTML encoding tends to take this approach. However, since not a few Japanese words have two or more Chinese characters to be represented, it breaks up the logical word structure and hinders deeper encoding. In contrast, the tabular structure is more intuitive, but the repetition of the same <rb> or <rt> tags would cause ambiguity in encoding, especially when the numbers of <rb> and <rt> tags differ. As both <rb> and <rt> are content elements in their own right, they prioritize presentation functions at the expense of their original logical structure.
32The double-sided ruby, termed complex ruby in HTML schemas, is modeled using both a nested structure and a boxed structure. The nested structure gives a simpler and more flexible representation of the structure. The boxed structure creates a table layout within which details are filled in. One weakness of the nested model is that it cannot handle some overlapping cases, as discussed by Uchida (2010). However, from the viewpoint of stand-off versus inline markup, this problem is rooted in a limitation of inline markup in XML. HTML5 addresses this problem by allowing empty <rt> elements (Ishida 2016). In the context of the TEI Guidelines, blank elements can be replaced by linking elements.
33As we have seen, the HTML encoding schemas were designed principally to allow ruby to be rendered in browsers. But TEI is concerned principally with meaning, not rendering. Therefore, for the TEI schema, it was necessary to invent a new model for descriptive encoding. TEI has not only introduced a nested ruby gloss element, but also provided a stand-off mechanism for correspondence between spans of ruby and glossed texts.
- 11 In the original proposal, there was a limit of one <rb> and one <rt> element per <ruby> element so (...)
34Our conception was, basically, that the schema should allow maximum flexibility through a simple model. In particular, while we allow nesting of the ruby gloss element (similar to the parallel segmentation method used in a critical apparatus), we also allow, if necessary, the double-end-point-attached method in order to accommodate the most complex examples, which require the use of <anchor>.11 We do not find convincing reasons for maintaining multiple encoding strategies, and for the reasons discussed above, we avoid employing either the interleaving structure or the boxed structure.
35The following examples show how our proposal addresses the examples discussed above.
Example 1. “真面目(ま/じ/め)” .
<p>
<ruby>
<rb><anchor xml:id="ma"/>真<anchor xml:id="ji"/>面<anchor xml:id="me"/>目</rb>
<rt><anchor target="#ma"/>ま<anchor target="#ji"/>じ<anchor target="#me"/>め</rt>
</ruby>
</p>
Example 2. “浴衣(ゆかた)” .
<p>
<ruby>
<rb>浴衣</rb>
<rt>ゆかた</rt>
</ruby>
</p>
Example 3. “あんたの鼾(いびき)” .
<p>あんたの <ruby>
<rb>鼾</rb>
<rt>いびき</rt>
</ruby>
</p>
Example 4. “神髪彦(カムイオトプシ)の” .
<p>
<ruby>
<rb>神髪彦</rb>
<rt xml:lang="ain">カムイオトプシ</rt>
</ruby>の </p>
Example 5. “大王(おおきみ)([だいおう])” .
<p>
<ruby>
<rb>
<ruby>
<rb>大王</rb>
<rt place="top">おおきみ</rt>
</ruby>
</rb>
<rt place="bottom">(だいおう)</rt>
</ruby>
</p>
Example 6. “労働者階級意識” .
<p>
<ruby>
<rb xml:id="rb1">労働者<anchor xml:id="rb1-1"/>階級<anchor xml:id="rb1-2"/>意識<anchor xml:id="rb1-3"/></rb>
<rt from="#rb1" to="#rb1-2" place="left">プロレタリアート</rt>
<rt from="#rb1-1" to="#rb1-3" place="right">かいきゅういしき</rt>
</ruby>
</p>
Example 7. Example from Hanabusa soshi.
<p>にむかひ<ruby>
<rb>梁</rb>
<rt>りやう</rt>
</ruby><ruby>
<rb>武帝</rb>
<rt>ぶてい</rt>
</ruby>の<ruby>
<rb>仏</rb>
<rt>ぶつ</rt>
</ruby>に<ruby>
<rb>淫</rb>
<rt>いん</rt>
</ruby>して<ruby>
<rb><ruby>
<rb>民膏</rb>
<rt place="right" type="primary">みんかう</rt>
</ruby>
</rb>
<rt place="left" type="secondary">たみのあぶら</rt>
</ruby>を<ruby>
<rb>費</rb>
<rt>ついや</rt>
</ruby>し国の<ruby>
<rb>衰</rb>
<rt>おとろ</rt>
</ruby>へとなり </p>
Example 8. Example from Igirisu dodoitsu.
<l>まこと<ruby>
<rb><ruby>
<rt place="right">りつとる</rt>
<rb>リットル</rb>
</ruby></rb>
<rt place="left"><ruby>
<rb>少</rb>
<rt place="right">すくな</rt>
</ruby>き</rt>
</ruby></l>
Example 9. Example of bopomofo.
<tagsDecl>
<rendition selector="rt" scheme="css"> ruby-position: inter-character; /*
making use of editor’s draft CSS 3 module, CSS Ruby Annotation Layout
Module Level 1 */ </rendition>
</tagsDecl> … <p>
<ruby>
<rb>牆</rb>
<rt>ㄑㄧㄤˊ</rt>
</ruby><ruby>
<rb>面</rb>
<rt>ㄇㄧㄢˋ</rt>
</ruby><ruby>
<rb>請</rb>
<rt>ㄑㄧㄥˇ</rt>
</ruby><ruby>
<rb>勿</rb>
<rt>ㄨˋ</rt>
</ruby><ruby>
<rb>攀</rb>
<rt>ㄆㄢ</rt>
</ruby><ruby>
<rb>爬</rb>
<rt>ㄆㄚˊ</rt>
</ruby>,</p>
- 12 In our original proposal, @type was allowed for the <ruby> element and required when nested. Howeve (...)
- 13 Descriptions in the TEI Guidelines, chapter 3.4.2 differ somewhat in details. When encoders divide (...)
36Some additional comments are worth making. First, in the case of double-sided ruby glosses encoding, if the order of priority is apparent, lower-priority glosses should be encoded in the shallower level and have a @type for both <rt>s; their respective values should be something like "primary" and "secondary" in the working language of the encoding project (see example 7; in this case, the right-sided ruby gloss is encoded in a shallower position than the left-sided one).12 If the order is not apparent, encoders should treat them in a consistent manner. Second, in a writing system where the default value of the ruby text position cannot be fixed, documenting it in <tagUsage> and <rendition> may be helpful (see example 9). Third, the correspondence between layouts of the ruby and glossed texts should be made clear with the aid of the <anchor> element.13 In an HTML context, reflecting a perception that ruby text is glossed per character, encoders may wish to encode example 1 as follows (example 11):
Example 11. Not recommended encoding example of “真面目(ま/じ/め)”.
<p>
<ruby>
<rb>真</rb><rt>ま</rt>
<rb>面</rb><rt>じ</rt>
<rb>目</rb><rt>め</rt>
</ruby>
</p>
Such a structure is not prohibited, but it prevents users from further analyzing and encoding the ruby and the glossed text.
37The new schema permits the encoding of complex ruby glosses. Ruby glosses can be nested, and adjusted for unusual spanned alignment, making it possible to encode triple ruby glosses and the nonnested glosses. It has also a downside in that these most complex ruby glosses cannot be directly translated into HTML, the most widely used ruby gloss encoding schema. In addition, the default ruby position cannot be fixed. Thus, every encoding project should use @place to specify a desired output.
38Our goal has been to describe the rationale for introducing to the TEI Guidelines a schema that fully accommodates the rich structure of ruby. This richness, as presented in Ariga (1989, 332), is seen in an 1874 essay by a Japanese author where the text is written in Classical Chinese but expected to be read with detailed accompanying glosses in Japanese. In the dialog a Chinese word 鄙吝 bǐlìn (in Sino-Japanese form, birin), or stingy, was transformed into Japanese slang kechinbō ケチンボー using an unusual ruby gloss to give a vivid tone to a formal narrative. (See also Takashi Wilkerson and Wilkerson 2000 on the playfulness of contemporary ruby.) Our proposal was to allow such a structure to be encoded in conformance with the Guidelines.
39Such inclusion of localized meaning not only takes a great step toward fully including a local textual tradition but also gives the TEI community a way to contribute to the global humanities by encouraging the inclusion of other underrepresented textual traditions. We hope this proposal encourages introducing other underrepresented local textual structures into the Guidelines.
SVN keywords: $Id: jtei-cc-ra-okada-200-source.xml 1184 2023-03-02 20:46:55Z ron $