Ghost characters

Address sign for Akenbara (𡚴原, Oaza-Kawachi 大字河内, Taga Town 多賀町.) The ghost kanji "妛" may be a misspelling of "𡚴".

In 1978, the Ministry of Trade and Industry established the standard JIS C 6226 (later JIS X 0208.) This standard defined 6349 characters as JIS Level 1 and 2 Kanji characters. This set of Kanji characters is called "JIS Basic Kanji". At this time, the following four lists of Kanji characters were used as sources.[1]^: 269f

1. Kanji Table for Standard Codes (Draft): IPSJ Kanji Code Committee (1971)
2. National Land Administrative Districts Directory: Geographical Society of Japan (1972)
3. Nippon Seimei's family name table: Nippon Life (1973, no longer extant)
4. Basic Kanji for Administrative Information Processing: Administrative Management Agency (1975)

At the time of the establishment of the standard, the authority for each character was not clearly stated, and it was pointed out that some characters had unknown meanings and usage examples. The term "ghost character" was coined from "ghost word," meaning a word that is included in dictionaries but has no practical use.[2] The most common examples are "妛" and "彁". These characters were never mentioned in the Kangxi Dictionary or the Dai Kan-Wa Jiten, a comprehensive collection of ancient Chinese character books.

In 1997, the drafting committee for the revised standard, led by its chairman, Koji Shibano, and Hiroyuki Sasahara of the National Institute for Japanese Language and Linguistics, investigated the literature referred to in the drafting of the 1978 standard. It was revealed that many of the characters that had been considered ghost characters were actually kanji used in place names.

According to the survey, prior to the drafting of the 1978 standard, the Administrative Management Agency had compiled eight lists of Kanji characters, including the above 1–3, in 1974, entitled "Frequency of Use and Correspondence Analysis Results of Kanji Characters for Selection of Standard Kanji Characters for Administrative Information Processing." This is accompanied by a list of kanji characters and their original sources. The results of this correspondence analysis, rather than the original sources, were referred to when selecting the JIS basic kanji at that time. Of these, many ghost characters were found to be included in those based on the Comprehensive list of administrative divisions of national land and List of Kanji characters for personal names by Nippon Life Insurance Company. In particular, the List of Kanji characters for personal names had no original source at the time of drafting the first standard, and its contents have been pointed out to be inadequate.[3]

In response to these results, the Standard Revision Committee restored the 1972 edition of the Comprehensive list of administrative divisions of national land from its proofreading history, and checked all the kanji appearing in the book against all the pages to confirm the examples.

In addition, as a replacement for the List of Kanji characters for personal names, which no longer exists, they conducted an exhaustive literature search, including a comparative study of the NTT and Nippon Telegraph and Telephone Public Corporation telephone directory databases and a survey of more than 30 ancient and modern character books.

12 kanji characters remain unidentified. 3 appear to be typos. Perhaps by coincidence, there are eight characters that were listed in the Japanese ancient dictionary or the Chinese ancient dictionary. As for "彁", no concrete source has been found.[1]^: 269f

Ghost characters have already been adopted into international standards such as Unicode, and changes to these standards are likely to cause compatibility problems, making it virtually impossible to modify or remove ghost characters.

The results of the aforementioned survey by Hiroyuki Sasahara et al. are summarized in Annex 7, "Detailed Description of Ward Locations," of JIS X 0208:1997. This section excerpts some of them.

JIS X 0208:1997 compiles the details of the sources of 72 characters whose sources have been identified, mainly those not listed in both Morohashi's Dai Kan-Wa Jiten and Kadokawa's Shin Jigen. However, this also includes characters that have been found to be misspelled by the original sources.

The list of delimiters appended as "source authority" in Annex 7 of JIS X 0208:1997 lists 72 characters, but the detailed text does not list "鰛(82-60)", which is only 71 characters.

The table below shows examples.

Since the establishment of the standard, the policy for compiling dictionaries has been to publish character books that are based on the assumption that all JIS basic Kanji characters are listed. For ghost characters, it is not possible to refer to past sources. Therefore, their treatment differed depending on the dictionaries and individual characters as follows.

Hiroyuki Sasahara et al. conducted a survey of ghost characters as part of the fourth revision of JIS X 0208 (1997). The so-called "83JIS revision" of the second standard (1983) had a history of causing great confusion due to incompatible modifications, mainly the replacement of punctuation marks. They couldn't let that happen again.

It was also necessary to maintain compatibility with Unicode. At the time of the survey, Unicode 1.1, which was enacted in 1993, already contained all JIS basic kanji. Therefore, the change at this juncture was not merely a domestic issue. In addition, Unicode adopts the source separation rule. According to this rule, for example, those recorded separately in JIS X 0208 are to maintain their separation. On the other hand, under the CJK Unified Ideographs policy, the same character code is assigned to different characters that are similar in China, Japan, Korea, etc. (see Han unification.) Because of this complicated process of Unicode creation, it was virtually impossible to exclude only Japanese ghost characters.

As a result, Japanese ghost characters remain in JIS Basic Kanji and Unicode.

Since the publishing of the standard, examples of ghost characters have appeared along with their widespread use. In particular, ghost characters are often mistakenly used for similar characters.

The "祢宜", the title of the deputy manager of a Japanese shrine, is sometimes misused as "袮宜." (ネ or 衤) In some cases, the Japanese surname "栩谷" is mistaken for "挧谷." (木 or 扌)[1]^{: 292, 304} Japanese folklorist Motoji Niwa introduced the surname "妛芸凡" (Akiōshi) in his book.[2][9]

The Asahi Shimbun database contained the name "埼玉自彊会" printed in the February 23, 1923 edition of the Asahi Shimbun, but when it was digitized, it was incorrectly labeled "埼玉自彁会." It has now been corrected.[10]

Unicode's CJK Unified Ideographs also have characters whose inclusion history is unknown and are sometimes called "ghost characters" as well.

For example, it has been pointed out that the character "螀" (U+8780), which was also registered in Unicode because it was included in the CCITT Chinese Primary Set, may be a typographical error that was adopted without sufficiently checking the source.[13]