I. Yintong: Database of Chinese historical phonology
David Prager Branner

  1. Description

    Yintong is a working prototype of a full-scale database of characters in the Guǎngyùn 廣韻, a major dictionary dating from 1008 C.E. All data was entered into the database by me and I am responsible for any necessary corrections in future. Yīntōng has been endorsed by the T‘ang Studies Society and a mirror site is now being set up at the American Oriental Society, where future development may take place.

    At present, the database has the following basic functions:

    1. If a character is entered, return information about the fǎnqiè associated with the character, the phonological values represented by those fǎnqiè, the page number of the Guǎngyùn where that reading appears.
    2. If a medieval Chinese reading is entered (in various forms), return a list of the other characters in the same xiǎoyùn.
    3. If two medieval Chinese readings are entered, return a list of any characters appearing in both xiǎoyùn.
    4. If a piece of Chinese text is entered, return a transcription of each character based on the Guǎngyùn’s readings.

  2. Content

    The database forming the “back end” of Yintong is the result of my own years of collation and correction of medieval Chinese character readings, beginning in 1998. There is no paper copy corresponding to this material. I have consulted three important collations of the Guǎngyùn and its predecessor, the Qièyùn (of 601 C.E.):

    Lóng Yǔchún 龍宇純. 1968. Táng xiě quánběn Wáng Rénxù Kānmiù bǔquē Qièyùn jiàojiān 唐寫全本王仁昫刊謬補缺切 韻校箋 [Collation and notes on the Táng manuscript of the Wáng Rénxù Corrected and supplemented Qièyùn]. Hong Kong: Chinese University of Hong Kong Press.

    Yue Nae-Wing [Yú Nǎiyǒng] 余迺永. 1993. Xīnjiào hùzhù jiàozhèng Sòngběn Guǎngyùn 新校互註宋本廣韻 [The Sòng dynasty edition of the Guǎngyùn, newly collated and cross-referenced]. Hong Kong: Chinese University of Hong Kong 中文大學出版社. Revised ed., Shanghai: Shànghǎi Císhū Chūbǎnshè 上海辭書出版社, 2000.

    Zhōu Zǔmó 周祖謨 (1914-1995; zì Yànsūn 燕孫). 1953. Guǎngyùn jiàoběn 廣韻校本 [A corrected edition of the Guǎngyùn]. Zhōu Zǔmó 周祖謨, ed. Shāngwù Yìnshūguǎn 商務印書館. Many reprints. Note the following: 2nd ed., (Beijing: 中華書局, 1988); 3rd ed. (韵學叢書), (Beijing: 中華書局, 2004).

    but the result is my own. The system for transcribing medieval Chinese phonological values (not actually pronunciations but phonological categories similar to modern “phonemes”) is also my own, and is designed to avoid reconstructive implications (see discussion of this issue in The Chinese Rime Tables). It was introduced in a 1999 article that is available on the Yintong website.

II. Basic Instructions for the Yintong Database
May, 2007

    The URL http://yintong.umd.edu brings up a splash page with fields for login (用戶) and password (密碼) on the right side. After you log in, the splash page has 5 links in the lower right corner:

    退出“logout”
    簡介“information about this site”
    單字搜尋“single-graph search”
    詩歌譯音“transcription of poetry”
    兩小韻找共同字“searching for graphs that are found in two different xiǎoyùn (homophone groups)”

    These are discussed below, one by one. Logout, of course, is straightforward; it returns you to the original splash page. The information page has not yet been posted.


  1. 單字搜尋 single-graph search

    When this page opens, there are a number of search fields available. The main body of the page consists of a line of management options at the top; then there are three colored sections labeled

    形 “form” (i.e., how a morpheme is written — its graph) 音 “sound” (i.e., phonologicalvalue) 義 “meaning”

    Below these there is a results area, with the left-hand labeled 搜索結果 “search results” and the right-hand labeled 運作信息 “report of action taken”. When the page is first opened, or after being cleared, the search results section may show 無結果 “no results”.

    1a. The management options consist of the following:

    a link labeled 運用設定“preferences”
    a drop-down menu labeled 輔助程序“supplemental programs”
    a button labeled 全面搜索“comprehensive search”
    a button labeled 全清除“clear all”
    a link labeled 主頁“main page”
    a link labeled 退出“log out”

    Clear all, main page, and log out should be self-evident.

    “Comprehensive search” performs a search based on all the fields filled in in the middle (three colored areas) section. Comprehensive search can also be done by hitting the return key when the cursor is in any of the search fields.

    1b. Most of the search fields that have been implemented to date are under “form”:

    字:“graph”
    康熙部首:“Kāngxī dictionary radical”
    四角號碼:“4-corner number”

    “Graph” is where the user enters a Chinese character to be looked up. I will return to this searchfield again under Section 2, when I walk through a sample search.

    1c. In the “sound” section, there are only two search fields:

    中古音切韻音系:“medieval reading, Qièyùn sound-system”
    反切:fǎnqiè

    The medieval reading is the phonological value of a morpheme in the medieval system (c. 500-1250 C.E.). The Chinese of that period had no knowledge of alphabets, and instead used something called “fǎnqiè” to represent sound. Fǎnqiè is a pair of characters, each of which has a pronuniciation; those two pronunciations, combined in a certain, indicate some third syllable's pronunciation. It's messy, but in order to study medieval Chinese, we need to pay close attention to fǎnqiè. A fǎnqiè is usually marked by either the character fǎn 反 or the character qiè 切 after the two “spelling” characters, so there are three characters in all; but only the first two are distinctive.

    Above the “medieval reading” search-field there is a little link that says 輔助 “assistance”. If you click it, a small window will open with a number of drop-down menus. These are intended as a substitute for regular expression (regexp) searches. They enable the user to search for medieval readings for which s/he knows only a part. The “assistance” window will generate complex search requests for a variety of different possibilities — for instance, words that have either initial {p} or initial {ph}, rather than only those with {p} or only those with {ph}.

  2. A sample single-character search.

    Put the character 輔 into the “graph” field and hit return or the “comprehensive search” button. Some of the other “form” fields will be populated, which we can ignore for now. And three new buttons have appeared to the right of the “graph” field:

    U+shows the Unicode number (useful for identifying graphs that aren't in your font)
    shows the font (i.e., converts back if you've just shown the U+ number)
    calls the Unihan website and gets a graphic of the character; useful again if your font is limited. I have received permission from the Unicode organizers to make calls on their database for this purpose.

    Now look at the “search results” section, where we see 以形爲主 “form as primary feature” (other possibilities will be discussed later). The sought-for character 輔 appears again, this time as a link — the link will simply perform the same search we've just done, but in cases where a number of different characters are returned, these links may be useful.

    On the next line, we see several pieces of information:

    [中古音]“medieval reading” (a label only)
    buoQ-3cAn actual medieval reading, in my own transcription (see article Branner 1999a in my vita).
    (遇_三上麌並)The same reading, but in one of the Chinese-only transcriptions rather than Roman-letter transcription. Notice the underscore — it takes the place of 開合, which is not contrastive in tōngshè; that is, tōngshè does not distinguish 開 from 合.
    扶雨切a historical fǎnqiè-value (ending in qiè) for this particular reading
    廣韻 262the source of this fǎnqiè, which happens to be page 262 of the book Guǎngyùn 廣韻

    All the items between {buoQ-3c} and the fǎnqiè are links. Most simply conduct a new single-character search on whichever graph you click on. However, {buoQ-3c} and the character qiè 切 in the fǎnqiè (you see it's boldfaced) perform other sorts of searches. I will come back to these below.

    2a. Searching by character, with multiple results

    Some characters have more than one reading. As an example, clear the fields with the “clear” button, and then doing a search on 三. 三 has not one but two readings. They are listed on separate lines.

    Now try searching for 平. 平 has two readings, and the second of these (being-3a) occurs redundantly in two different places in the Guǎngyùn. So {being-3a} is given once, along with its corresponding non-Roman transcription 梗_三平庚並, but there are 2 fǎnqiè and two pages cited in the Guǎngyùn for that single reading.

    Now try searching for 連. 連 has only one reading, {lan-3b} (山開三平仙來), but there are two different fǎnqiè listed — one from the Guǎngyùn and one from the Jíyùn.

    Now try searching for 聯. It has only one reading and only one fǎnqiè, but there is an additional, interesting piece of information: 廣韻云聮說文作聯 “The Guǎngyùn says that the graph 聮 is written as 聯 in the Shuōwén.” For us, 聯 is the normal graph, but the Guǎngyùn happens to treat 聮 as the normal graph and notes 聯 as a Shuōwén variant. Information about graphic variants in the database is still somewhat incomplete, but eventually all such examples will be cross-referenced.

    2b. Searching by reading

    Suppose you want to find out what other characters there are that have the reading as 連 and 聯. You can do this either by entering that reading {lan-3b} (without the curly brackets) into the “medieval reading” field, or by clicking on that reading in the search results. Doing the latter will populate the “medieval reading” field, and then conduct a “comprehensive search”.

    Note: Since the “form” section is not automatically cleared by this process, the search results become a little cluttered at this point.

    First, we have 最接近的詞素 (形 中古音 反切) “nearest morpheme”. That terminology originates in the fact that we have been leaving room for “form” searches in which the graph itself is not necessarily supplied, but perhaps only some supplementary information (such as 諧聲 音符) that is shared among a number of different graphs.

    Then we have the 以形爲主 result — the same one we saw earlier when we searched for 聯 alone.

    Then we have 以中古音爲主 “medieval reading as primary feature”, and a long list of characters and their fǎnqiè. We see our reading {lan-3b} and the Chinese equivalent at the head of that list, numbered 1. (There is no 2, but there could be, if we had put in a partial reading and asked the software to find us all matches.) 15 matching characters are found, along with their fǎnqiè information.

    2c. Searching by fǎnqiè

    Entering a fǎnqiè into the fǎnqiè search field, or clicking on the graph qiè 切 in a search result, searches not for the reading in question but for that exact fǎnqiè.

  3. 詩歌譯音 Transcription of poetry

    If you go back to the main page and then click on “transcription of poetry”, you come up with a different search page. There is only one search field, a large one, to accommodate texts of poetry. The software will transcribe each character in the poem and present the whole thing in 5 or 7 columns (if the total number of characters is divisible by 5 o r 7, which suggests regular 5- or 7-syllable line length).

    There are a number of useful options available in the Preferences (運用設定)

    If the poem has a different but still regular number of characters per line, the software will not know how to break lines correctly unless you enter the number of characters per line under under “每行字數”.

    You may switch between my transcription and the traditional Chinese non-Roman notation (丁聲樹與李榮格式), marking it 是 “yes” rather than 否 “no”).

    You may also turn on the display traditional prosodic symbols (顯示平仄符號 — ○ for píng 平 and ● for 仄) in addition to the transcription. You may control the size of these symbols (平仄符號大小).

    Don’t forget to click “OK” in order to register the preferences.

    After a poem has been transcribed, it will be displayed using HTML tables. If you have trouble copying and pasting the content into your own document, click on “可剪貼文版本” to open a Unicode text-only version of the same material, with one line of the poem per paragraph and the readings of each syllable tab-delimited. This format can easily be converted to table notation in word processors and spreadsheets.

  4. The union of two different xiǎoyùn (homophone groups)

    On the splash page, the fourth of the five basic links is 兩小韻找共同字 “searching for graphs that are found in two different xiǎoyùn (homophone groups)”. This brings up a page with two parallel fields for searching by reading. If both fields are filled and the search button is clicked, the software will first do the two searches and then at the bottom of the results display a list of those characters that have both readings.



Branner, Yintong: Database of Chinese historical phonology, 2007.05