Faculty of Computing and Information Technology
Department of Rob otics and Digital Technology
Technical Rep ort
A Japanese Electronic Dictionary Pro ject Part
The Dictionary Files
J W Breen
November
Enquiries
Technical Rep ort Co ordinator
Rob otics and Digital Technology
Monash University
Clayton VIC
Australia
tr coord rdt monash edu au
Contents
Abstract and Keywords
Introduction
Pro ject Overview : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
Japanese Orthography and Text Pro cessing : : : : : : : : : : : : : : : : :
Japanese English Dictionaries
The Main Dictionary File
The EDICT File : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
File Structure : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
Lexicographic Details : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
File Compilation : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
The Character Dictionary File
Character File Structure : : : : : : : : : : : : : : : : : : : : : : : : : : : :
Character File Compilation : : : : : : : : : : : : : : : : : : : : : : : : : :
Copyright Issues
Conclusion
Acknowledgements
A The Japanese Writing System
B Japanese Text Pro cessing
C The Unico de Co ding System
Abstract
Electronic multi lingual dictionaries have seen considerable development in the last decade
The standardization of co ding systems for the orthography of many Asian languages in
the same p erio d combined with the increased availability of low cost micro electronic
storage and display systems has op ened up considerable demand and p otential for dictio
nary systems in these languages This rep ort describ es an on going pro ject to develop and
maintain a comprehensive electronic Japanese English dictionary capable of use within
a variety of search and display electronic text reading supp ort and machine translation
environments The pro ject les are b eing developed in the public domain The dictionary
les have at the time of writing attained the status of b eing the ma jor freely available
electronic rep ository of Japanese English dictionary material in the world
Keywords
dictionary lexicography Japan JIS WNN SKK kanji hiragana katakana Unico de
Chapter
Introduction
Pro ject Overview
This pap er describ es an on going pro ject to develop and maintain a comprehensive elec
tronic Japanese English dictionary capable of use within a variety of search and display
electronic text reading supp ort and machine translation environments The pro ject les
are b eing developed in the public domain The pro ject which was initiated and co ordi
nated by the author has b een assisted by the volunteer lab our of many p eople who have
selected translated formatted and edited material for the dictionary
The dictionary les developed in the pro ject consist of
EDICT the main dictionary le of Japanese words with their pronunciations and
English translations
KANJIDIC a le of explanatory information and indices for the kanji char
acters in the JIS Level and character sets describ ed in the JIS X
standard
The EDICT and KANJIDIC les have at the time of writing attained the status
of b eing the ma jor freely available electronic rep ository of Japanese English dictionary
material in the world and are b eing used by thousands of researchers teachers and
students via a large number of software packages which have b een developed to use or
incorp orate them The EDICT le now has approximately entries of which some
are Japanese prop er names
An adjunct to the development of the dictionary les is the development of software
which accesses and displays the material in a variety of contexts An additional pap er
is in preparation which describ es a number of software systems developed by the author
and others to carry out this task
Japanese Orthography and Text Pro cessing
The Japanese language has a unique system of orthography which p oses particular chal
lenges to b oth computer text pro cessing and the lexicographer This pap er assumes on
the part of the reader an elementary knowledge of Japanese orthography an overview of
which is included in App endix A For a fuller description of the Japanese writing systems
the reader is directed to an introductory text such as Neustupn y or O Neill An
overview is also in Unger and a more historical treatment in Seeley
The ability to store and pro cess the many thousands of characters used in Japanese
orthography is central to an electronic dictionary system A basic knowledge of the
mechanisms used to store and pro cess Japanese characters is assumed in this pap er An
overview of the co ding systems is included in App endix B and the reader is referred to
Lunde for a thorough treatment of the sub ject
Chapter
Japanese English Dictionarie s
Japanese dictionaries take two general forms The rst is a word oriented dictionary of the
form found in most languages Such dictionaries are a lexically ordered set of entries each
of which b egins with a head word and contains synonyms explanations etc In Japanese
such dictionaries usually have their headwords in kana and are ordered according to the
1
standard gojuon order Both katakana and hiragana headwords are used and are treated
as equal for the purp ose of ordering There are literally hundreds of such dictionaries in
print and use in Japan The ma jor and most authoritative is the kojien w which
is the Oxford English Dictionary of Japanese dictionaries
The second form of dictionary the character dictionary is unique to languages such as
Chinese and Japanese which use non phonetic symbols Japanese character dictionaries
known as kanwajiten wM Chinese Japanese dictionary typically have the characters
group ed by radical that is an identi able element of the character usually o ccurring at
the left or top of the character Within each radical group characters are ordered by
stroke count either the total number of strokes or the number of strokes excluding the
index radical The radicals used for this purp ose are usually based on the classical
radicals of Chinese orthography Use of such a dictionary requires some skill and practice
in b oth identifying the index radical and counting the strokes that comprise the character
Often a separate index is available which enables each character to b e found by either its
on or kun reading The information in each entry typically includes the full set of on and
kun readings the meaning s asso ciated with the character and a selection of the ma jor
comp ounds which b egin with the character Larger dictionaries also include the ma jor
variants of the character and some etymological information
Again there are many kanwajiten in use in Japan The ma jor reference dictionary is
the volume Morohashi daikanwajiten ewM
Dictionaries which combine Japanese with another language such as English obviously
need to have sections or parts which cover b oth Japanese English and English Japanese
They fall into two groups those prepared for Japanese sp eakers and those prepared for
English sp eakers There are many dictionaries in the rst group and relatively few in the
second The reason for the existence of the two groups of dictionaries lies with the nature
of written Japanese Without at least familiarity with the kana and in many cases a large
number of the joyo kanji dictionaries prepared for the Japanese domestic market are of
1
eY fty sounds The standard by table of kana which is usually used for syllable ordering
little use to non sp eakers of Japanese The following examples from two such dictionaries
illustrate this p oint
universe n gD cosmos H O H m mc Hc f" world
V JOZ H T F all mankind etc etc
and
md . / a trial HbcI Z!B the case is under on trial
Dictionaries prepared for those with limited Japanese reading skills usually rely on
the use of romaji particularly in the headwords As such dictionaries tend to b e aimed at
the market of less skilled users b oth the quantity and quality of the entries are limited
The following examples are from two such dictionaries
universe n uch u gD zen sekai Tf"
and
uch u g D N universe cosmos
As students progress in Japanese studies the dictionaries prepared for Japanese sp eak
ers b ecome more accessible This applies particularly to Japanese English dictionaries
English Japanese dictionaries remain a problem and it is often claimed that there is no
English Japanese dictionary available which adequately serves the requirements of mo d
erately advanced students One recent dictionary attempts to bridge this gap It is
essentially a version of a domestic English Japanese dictionary which has had the English
pronunciations removed and the readings of all Japanese words which are in kanji added
in the form of furigana i e small hiragana ab ove each character While this is a con
siderable advance it still lacks su cient contextual information in English to enable a
non sp eaker of Japanese to select b etween the various translations The following example
is from that dictionary
2
universal a g D H H J m H T f " m H q E B O R m
D H JmH H
For several decades there have b een quite adequate character dictionaries available to
English sp eakers The Nelson which sets the b enchmark in this type of dictionary
has details of over characters and comp ounds In recent years several other
dictionaries have b een published including the Spahn Hadamitzky which broke new
ground by also listing comp ounds which have the character in reference in other than the
rst p osition
2
The hiragana denoted app ears as furigana
Chapter
The Main Dictionary File
The EDICT File
The initial dictionary used in the pro ject was the EDICT le supplied as part of the
MOKE Mark s Own Kanji Editor PC text editor package developed by Mark Edwards
The original EDICT despite its acronym English DICTionary was a basic Japanese
English dictionary text le with the following format
1
kanji headword yomikata English English
or
kana headword English
2
This format is similar to that used in MOKE s henkan les and was based on the
3 4
le format used in the SKK FEP
As originally released with MOKE EDICT had under entries compiled by
Edwards Like MOKE s reference les it was enco ded using EUC MOKE provides a
facility the search the le sequentially for a text string o ccurring in the English elds and
also to match on a kanji headword encountered in a text le It also can app end extra
records to the EDICT le
As compatibility with MOKE which at the initial stages of the pro ject was the only
readily available MS DOS Japanese text editor app eared to b e desirable the EDICT le
format was retained Edwards gave his p ermission for the original EDICT entries to b e
included in the pro ject le Hindsight indicates that the EDICT format has b een less than
ideal but as there is now a considerable number of software packages developed around
that format it is unlikely to change
1
?Yk way of reading The term commonly used for a phonetic transliteration of a word
2
henkan or conversion les play an imp ortant part in Japanese text editors as they provide the
mappings needed to translate kana sequences into the appropriate kanji
3
Simple Kana to Kanji an FEP package developed by Masahiko Sato at Tohoku University for Nemacs
4
FrontEnd Program a term commonly used in Japan for software which takes keystrokes either
roma ji or kana and converts the text into the appropriate kanji
File Structure
A number of le design goals were developed for the pro ject
the le structure would b e accessible by users with appropriate text editors This
meant in e ect that the master les would b e held in a text form No indexing
information would b e held in the le itself and it would b e the task of software using
the le to reformat the le into say an indexed database structure or to provide
an index system external to the le As part of this it was necessary that there
b e ready identi cation of tokens consisting of kanji and or kana strings or English
words for the purp oses of indexing and searching
the le would lend itself as far as p ossible to bidirectional use i e in supp ort of
b oth Japanese English and English Japanese searches The one to many nature of
dictionary entries presented a problem which could not b e totally resolved without
the introduction of a le structure which included a complex indexing arrangement
This would however violate the rst goal
After some exp erimentation with various indexing search and display techniques
it was concluded that this goal could b e met adequately with a le of the EDICT
structure provided the English translation elds for each Japanese entry contained
enough contextual information to enable a user following a search on an English
keyword which resulted in a display of all the entries which contained that word to
select the appropriate Japanese entry
Lexicographic Details
In the development of the contents of the le the following lexicographic principles have
b een followed
no use is made of the roma ji system except when it is appropriate to include it in
an English eld such as with prop er nouns and with ob jects such as sukiyaki and
sushi which are well known by their Japanese names This decision was made in
part on principle the Japanese language is rarely written seriously using roma ji
and partly to encourage the learning and use the kana at an early stage in language
acquisition
as is the normal practice with Japanese dictionaries all Japanese verbs and adjec
5
tives are listed in their dictionary form Users of such dictionaries are exp ected
to b e able to identify the dictionary form of a word With electronic dictionary
les this presents a problem only when an attempt is made to match entries against
text which contains in ected forms of such words English verbs have b een included
in their in nitive form thus removing the need to mark that they are verbs In a
few cases it has b een necessary to distinguish b etween transitive and intransitive
Japanese verbs with the vt and vi marker
5
Japanese has no in nitive form for verbs The usual practice is to record verbs in dictionaries in their
familiar or plain non past form
adjectival nouns or quasi adjectives N & X keiyodoshi also sometimes referred
to as na adjectives b ecause their nonpast prenominal form ends with a D na are
6
recorded without their D ending but will have the mo di er an included as a
grammatical indicator
words such as comp ounds and loanwords which are commonly combined with the
verb e suru are given the grammatical marker vs Such entries do not have
the e in either the headword or the yomikata and the English elds are de ned
in the form of nouns or participles
Example da d co oking vs
nouns which may take the genitive case particle H no are marked with the mo di er
a no
di ering Japanese sp ellings of a headword arising from the inconsistent use of okurig
7
ana are handled by including separate entries for each common sp elling This is an
inelegant but exp edient way of handling what the Japanese refer to as the okuri
gana problem as it enables text reading software to make appropriate matches
with entries As it adds to the size of the le and could b e seen to b e encouraging
p o or writing practices it is hop ed that at some time such software will incorp orate
metho ds for the deterministic identi cation of okurigana patterns thus enabling the
entries in the le to b e con ned to the standard sp ellings
it was decided to include prop er nouns in the le b oth the family and given names
commonly used in Japan and place names Such entries do not commonly o ccur in
dictionaries however it was considered worthwhile given that the dictionary le is
intended to assist inter alia with text reading
in keeping with the nature of the pro ject the sp elling of the English elds in the
entries has b een left in the style of the various contributors Thus the le contains
a mixture of British and North American sp ellings As far as p ossible translations
have b een edited to ensure that they are broadly comprehensible to sp eakers of all
varieties of English and that ma jor national variants are identi ed
a number of abbreviated indicators are used in parentheses in the English elds to
provide grammatical and other information ab out the entries Among these are
vt transitive verb
vi intransitive verb
id idiomatic expression
col collo quialism
vul vulgar expression or word
pn p erson name family or given
6
an adjectival noun
7
okurigana is the term for writing part of a verb form in kana Practices vary in okurigana usage for
example the word ikebana ower arrangement is often written b oth z t and z t The former is
preferred by purists as it indicates the reading of the z kanji but the latter is also often seen
pl place name
giv given name
fam familiar language
p ol p olite V teineigo language
hum humble kenjogo language
hon honori c or resp ectful sonkeigo language
pref pre x
suf su x
uk word usually written using kana alone
uK word usually written using kanji alone
Here are examples of entries in the le
bDc go o d bye
D to sob to weep
z architecture
O faucet tap
changing household registry
Bt withering atrophy contraction
C g i Koromogawa pn pl
F p osthumous song or p o em
G do ctor medical
K to raise child to b e brought up to grow up
File Compilation
As mentioned ab ove the original EDICT le had fewer than entries A further
entries were keyed by the author from a rst year student dictionary published by
Swinburne Institute of Technology with the kind p ermission of the authors and several
hundred more entries were added by the author from lists compiled during Japanese
reading and study
Since the release of this initial version of EDICT in mid there has b een steady
ow of entries from p ersons interested in adding to the coverage of the EDICT le Several
individuals had compiled extensive p ersonal vocabulary lists which they contributed The
Sony company in Japan p ermitted the inclusion of a signi cant in house le Many entries
were derived from a Japanese German dictionary le in the EDICT format compiled by
Helmut Goldenstein from the Langenscheidt edited by Hadamitzky As Goldenstein
had p ermission to make this le freely available the entries which were not already in the
EDICT le were translated into English
A ma jor task in developing such a dictionary le is the identi cation and entry of the
tens of thousands of kanji comp ounds which make up the written language along with
their yomikata Fortunately this task has b een greatly assisted by the prior work of the
two main public domain FEP systems in Japan the ab ove mentioned SKK system and
8
the even larger WNN FEP system developed jointly by the Kyoto University Research
Institute for Mathematical Sciences OMRON Corp oration and ASTEC Corp oration As
part of the WNN pro ject a public pro ject was undertaken in which academics researchers
and students throughout Japan submitted kanji kana entries to go into the large henkan
le now known as the WNN Pub dic MOKE s main henkan le was compiled from the
SKK and WNN les and the existence of these les in the public domain has b een a
ma jor factor in the development of EDICT
Many thousands of entries in EDICT have b een derived directly fom the SKK and
WNN sources These entries are
loanwords , p gairaigo As one feature of these systems is the ability to
generate a loanword from its English original there are many English loanword
pairs which only required reformatting to b e included in EDICT
prop er nouns The WNN Pub dic has separate les of jinmei h p erson name
and chimei h place name words which were suitable for immediate conversion
into the EDICT format In the case of prop er nouns the English translation eld
is used for the romanized transliteration of the words e g
Lg X Hiroshima pl
Note that a slightly mo di ed version of the common Hepburn system of romaniza
tion has b een used throughout the EDICT le In Japanese the lengthened form
of the vowel o is almost always written ou whereas in roma ji it is usu
ally written o o or o and on o ccasions oh In order to re ect the Japanese
orthography more accurately and to b e compatible with the usage of common FEP
systems the ou convention has b een followed throughout Thus B U will
app ear as Touhoku not Toohoku or T ohoku The only exceptions to this are se
lected entries for Tokyo Osaka and Kyoto which are well known in that form and
may not b e recognized in the more orthographically correct versions of Toukyou
Oosaka and Kyouto
8
An acronym formed from Japanese sentence Watashi no Namae wa Nakano desu my name is
Nakano as a design goal of the WNN pro ject was to b e able to enter such a sentence and have the
correct kanji and kana PHhPIB A selected automatically
Chapter
The Character Dictionary File
The second le necessary for a complete dictionary system is the character le containing
information ab out each of the kanji commonly used in Japanese Such a le ful ls a
number of roles in a dictionary system
to record against each character the key items which can b e used to identify the
character e g radical stroke count or reading
to record other information of interest in scholarship or translation such as the
English meaning s asso ciated with the character references to ma jor dictionaries
co ding schemes etc
to provide a starting p oint for dictionary software once a character has b een iden
ti ed to search the main dictionary le and display comp ounds containing that
character
Character File Structure
In keeping with the goals of the main dictionary le the character le was compiled as
a simple text le KANJIDIC with one line for each of the kanji characters in the
JIS X standard
Each line consists of a number of space separated elds The elds in the lines of
information are
the kanji itself
the byte ASCI I representation of the hexadecimal co ding of the two byte JIS
enco ding of the character This eld is to facilitate editing and manipulation of the
le
information elds b eginning with a unique identifying letter or pair of letters At
the time of writing the eld types in use are
U hexnum The Unico de enco ding of the kanji one p er line See App endix
C for information on the Unico de system
N num the index number in the Nelson dictionary There will b e one p er
line unless the character is not in Nelson or is considered to b e a non standard
version in which case there will b e a fsee Nnnng cross reference app ended to
the meanings elds
B num the radical bushu number one p er line As far as p ossible this
is the radical number used in Nelson In a number of cases Nelson classi es
the character di erently from the classical or historical usage In these cases
the classical or historical radical numbers follow as a separate C num entry or
entries
C num the historical or classical radical number where this di ers from the
B num entry
S num the stroke count at least one entry p er line If there is more than
one eld the rst is considered the accepted count while subsequent ones are
common variants
G num The joyo grade level i e the elementary school grades in which
the characters are taught in Japan There is at most one p er line with the
following co ding
G through G indicate joyo grades
G indicates the remaining seondary school general use characters
G indicates jinmeiyo h Y for use in names characters These
kanji lie outside the joyo but are p ermitted in names
H num the index number in Halp ern There is at most one p er line
F num the frequency of use ranking as rep orted in At present the
most used characters have a ranking Those characters that lack this
eld are not ranked
P co de the SKIP pattern co de similar to that used in Halp ern The co de
is of the form P num num num See Halp ern for a description of his SKIP
1
pattern co de
Q num num the four corner enco ding of the kanji This is a Chinese classi
cation system which categorizes the stroke combinations in the corners of the
kanji Some older Japanese character dictionaries such as Morohashi include
a four corner index
MN num the index number in the Morohashi daikanwajiten In the few cases
where JIS kanji are not in Morohashi the index indicates the last number with
the same Bushu and stroke count
MP num num the volume and page in Morohashi on which the entry is to
b e found for the kanji
E num the index number in Henshall Only the joyo kanji are covered by
this b o ok
1
At the time of writing this eld is not included in the distributed version of the le as Halp ern has
claimed patent and copyright protection for the system
Y xxxxx the PinYin Mandarin Chinese reading of the character including
the tone There may b e several PinYin elds for each kanji however the kokuji
!f characters developed in Japan do not have these elds
the readings used for the characters As is a common practice in character dictio
naries the on readings are in katakana and the kun readings in hiragana Readings
used as pre xes and su xes are indicated by a trailing and leading resp ectively
and a is used to separate a reading from its okurigana Note that the readings
are not in any particular order and are based on common usage rather than any
o cially approved lists
the English translations and or notes Each eld is encapsulated by a pair of braces
f g
Here are examples of the entries for a number of common kanji in the KANJIDIC le
k b U e N B S G H F P Q MP
MN E Yyu Yyu X fraing
N U c b N B S G H F P Q MP
MN E Ywu fro ofg fhouseg fshopg fdealerg fsellerg
U N B B S G H F P Q MP
MN E Yying Yang e I e J fre ectg fre ectiong
fpro jectiong
Character File Compilation
The compilation of the KANJIDIC le b egan in late when a contributor to the main
le obtained and provided the basic bushu stroke count and reading information for the
JIS Level kanji This was extended shortly afterwards to include the JIS Level kanji
by the provision of a le of the bushu and stroke counts from a public source in Japan
The Sony dictionary le provided many of the Nelson indices readings and initial English
translations
In the following eighteen months there was a large amount of painstaking editorial
work carried out by several contributors to verify and expand the information elds
During this time the Nelson indices and the Grade co des were completed and veri ed the
Halp ern and Henshall indices the usage fequencies and the SKIP co des compiled and
the readings and translations established and veri ed for all but a small number of the
characters
In early the author b ecame aware of a similar le which had b een compiled by
Christian Wittern at Kyoto University with some material provided by Urs App from
Hanazono University Wittern made this le available to the pro ject and it is from
this source that the Morohashi four corner and PinYin elds are derived A further
veri cation of the Nelson indices was also made with this source
Chapter
Copyright Issues
Dictionary copyright is a complex issue b ecause clearly the rst lexicographer who pub
lished v inu means dog could not claim a copyright violation by all subsequent
Japanese dictionaries What makes each dictionary unique and copyrightable is the
particular selection of words the phrasing of the meanings the presentation of the con
tents a very imp ortant p oint in the case of these les and the means of publication
Advice to the author from p eople with exp erience in copyright matters is that EDICT
and KANJIDIC are in no di erent a p osition than any other new dictionary While there
is a p opular view that material in the public domain keyed in by volunteers is somehow
exempt from copyright law the more considered opinion is that the same restrictions
apply as for any other form of publication
Copying material from published dictionaries has b een discouraged during the devel
opment of these les but with a voluntary e ort b eing carried out simultaneously in many
countries it has b een di cult to p olice this matter It has also b een the established prac
tice for lexicographers to use previous works as reference material in the development of
their own dictionaries Nelson in the foreword to his character dictionary acknowledges
his indebtedness to many other dictionaries and reference works the list of which covers
more than a page While there is a certain amount of honour among thieves in commer
cial lexicography dictionary compilers frequently include b ogus entries in order to trap
pirates of their material Several such entries were noticed in Spahn Hadamitzky
where obscure kanji were given English meanings from Lewis Carroll s Jabb erwocky
brillig slithy b orogove etc
Two interesting p oints of copyright arise in relation to the character dictionary le
the le contains indices identifying the characters in a number of ma jor dictionar
ies Nelson Halp ern etc It is imp ortant to consider whether those indices are
protected by copyright and whether their inclusion in the le can b e viewed as a
breach of the copyright of those publications
The p osition taken with KANJIDIC is that the indices are not intrinsic comp o
nents of the information content of those publications They are p ointers to the
information in the publications and their presence in the KANJIDIC le merely
facilitates access to the relevant information in the same manner as a page number
The practice of including indices to other dictionaries is well established Nelson
includes the indices of the Fuzanbo dictionary and the author s copy of the pre war
Rose Innes character dictionary includes indices for the contemporary edition
of the Ueda daikanwajiten
a more di cult p oint arose with the SKIP co des mentioned ab ove Halp ern states
in his dictionary that patent protection has b een sought for the enco ding metho d
and sp eci cally advises that the co des must not b e used without p ermission to order
or index characters in other dictionaries While a number of users of the les have
expressed doubt to the author ab out the validity of such a patent as the enco ding
system is quite similar to others used previously by other authors it is clear that
the presence of the SKIP co des in the le could with appropriate software b e used
to identify the kanji and thus would violate a valid copyright For this reason the
distribution of the SKIP co des was discontinued until the issue could b e resolved
The copyright of the pro ject dictionary les presumably b elongs to the myriad of
contributors However with the placing of the pro ject in the public domain from the
b eginning it is clear that the protection sought is minimal The only restrictions placed
on the distribution of the les is that no charge must b e made other than a nominal charge
for the medium and that the les must not b e incorp orated in commercial pro ducts
With regard to this latter p oint there have b een discussions with several prosp ective
developers of commercial packages ab out access to the les The agreed p osition has
b een that pro ducts may access the les but they may not include the dictionary les
themselves for more than a nominal charge and they must provide the facility for users
to obtain and incorp orate up dated versions of the les
Chapter
Conclusion
The pro ject describ ed in this pap er represents a ma jor development in computer lexicog
raphy and a considerable achievement considering that its compilation has b een carried
out largely by volunteers who are employed in elds other than linguistics and language
instruction
Software packages drawing up on the les have b een compared favourably with ex
p ensive commercial pro ducts A technical translator of Japanese working in the USA
remarked in corresp ondence with the author recently that the les are the information
source increasingly b eing used in his eld
Chapter
Acknowledgements
The total number of p eople who have contributed to the pro ject is approaching a hun
dred Their work is most gratefully acknowledged and their names are recorded in the
do cumentation les of the pro ject
The author is particularly indebted to the following p eople who have made signi cant
contributions
Urs App Stephen Chung Lee Collins John Crossley Hitoshi Doi Mark Edwards
Mike Erickson Curtis Eubanks Je rey Friedl Magnus Halldorsson Ken Lunde Theresa
Martin Kevin Mo ore Yasuaki Nakano Cli ord Olling Alfredo Pino chet Harold Rowe
Iain Sinclair Rik Smo o dy Kurt Stueb er Hidekazu Tozaki Scott Trent Richard Walters
Christian Wittern
App endix A
The Japanese Writing System
The following is a very brief outline of the essential elements of Japanese orthography
The language is written in three sets of graphical symbols
kanji These are the characters invented in China and introduced to Japan during
the rst millenium AD Instead of representing a vowel or consonant a kanji charac
ter represents a word or a morpheme Of the tens of thousands of kanji the Japanese
Education Ministry has established a set of joyo u Y general use kanji for
use in domains such as schools textb o oks and newspap ers This set forms the basis
of mo dern Japanese literacy although several thousand additional kanji are in use
in sp ecial sub ject areas Each kanji has in general two sets of pronunciations or
readings The kun reading is usually derived from an archaic Japanese source and
applies when characters are used singly in nouns verb ro ots etc or are used in
combinations to form prop er nouns Such words are known as wago native
Japanese words The on readings usually derive from Chinese sources and are
b elieved to b e the result of the adoption of many Chinese words and concepts into
Japanese during the assimilation of the writing system centuries ago On readings
are used when two or more characters are used together to form a new word or com
p ound These words are known as kango Chinese word As an example of
the di erent types of readings the character has the meaning of east and is used
alone to write the word meaning east higashi Thus the kun reading is higashi
When used in a comp ound with the character 6 which means capital a new word
with the meaning eastern capital is created with the pronunciation tokyo Thus the
on reading is to
hiragana and katakana These are two syllabaries collectively known as kana which
were developed in Japan during the second half of the rst millenium AD Both
were developed from kanji characters which were traditionally used phonetically
Hiragana is more cursive in shap e and katakana more angular
Example Lc D hiragana katakana
In their mo dern forms each syllabary has basic symbols representing the ve
vowels a i u e o consonant vowel syllables ka ki sa shi and a single
consonant n Other symbols are created through the use of diacritic marks to
indicate voiced or labial consonants Several small kana are also used as vowel and
consonant mo di ers bringing the total number of distinct syllables in mo dern use
to over one hundred
Although either syllabary can b e used e ectively to write any Japanese text in mo dern
Japanese the three sets of symbols are used together in distinctly di erent roles
kanji are used to write the base forms of most nouns verbs adjectives adverbs etc
hiragana is used to write most of the grammatical elements such as the in ecting
p ortions of verbs and adjectives connectives and particles as well as demonstratives
many p ersonal pronouns and a number of other words such as auxiliary verbs
katakana is used primarily for the many thousands of loanwords , p gairaigo
which have entered Japanese from other languages in recent centuries Most of
these have come from English although a number have also come from Chinese
French Portuguese Dutch and German It is also used for emphasis and for many
onomatop o eic words
The following simple sentence demonstrates a combination of all three
I A A l 7 X Meaning Yukiko b ought a kimono at the
department store
The kanji comp onents of this sentence are
yukiko prop er name the characters mean snow child
A kimono the characters mean wearing object
7 ka the ro ot of the the verb kau to buy
The hiragana comp onents are
I wa particle marking the sentence topic
A de the particle indicating lo cation of an action
l wo the particle indicating the preceding phrase is the ob ject of the verb
X imashita the past tense p olite in ection appropriate for the verb kau
The katakana comp onent is
depaato the abbreviated loanword derived from the English department
store
Japanese can also b e written using the Latin alphab et The representation of Japanese
syllables in this alphab et is called romaji E f of which there are several forms In this
pap er the Hepburn system is used Roma ji is used in Japan where it may assist foreigners
such as on railway station signs and in much advertising and shop signage where it has
a certain chic app eal Apart from some elementary language texts for foreign students it
has virtually no role in written Japanese language
App endix B
Japanese Text Pro cessing
Part of the information in this app endix was drawn from the do cument Electronic
Handling of Japanese Text by Ken Lunde which is in a freely available public domain
le JAPAN INF Lunde has greatly expanded this do cument in pro ducing his recent
text
The advent of computers and the relatively limited co desets which prevailed in the
early decades of computerization represented a considerable challenge to would be Japan
ese computer users Until the s most Japanese computer systems were restricted to
using roma ji or one of the kana usually katakana for input and output Some commenta
tors even expressed the view that the incapability of computers to handle the large symbol
sets used in Chinese and Japanese would hasten the demise of those writing systems
The invention of output devices such as laser printers capable of handling a large vari
ety of symbols with su cient precision combined with the reduction in the manufacturing
costs of programmable graphics display devices and micro electronics in general enabled
by the early s a progressive incorp oration of traditional Japanese orthography into
Japanese computer systems
One of the essential elements in the computerization of any symbol set is the stan
dardization of the co ding system used to represent the symbols After considerable ex
p erimentation and development by Japanese computer companies a Japanese Industrial
Standard JIS C Co de of the Japanese Graphic Character Set for Information
Interchange was released in with further revisions in and to track the
introduction and up date of the j oy o kanji set and its extensions The current standard
is JIS X In this standard a two byte co de is used for each symbol The co de
ranges were selected to enable them to b e incorp orated as graphical extensions to the
common compatible co ding systems used world wide ASCI I CCITT No Alphab et
AS etc Also the co des were selected to lie within the printable characters of the
ASCI I set and thus they could with the insertion of suitable shift in shift out sequences
b e mixed in text and electronic mail les with characters from the other co des The total
number of symbols capable of enco ding by this system is by However this
is ample to provide for all the characters in common Japanese use
The co de standard covers characters kanji in levels level
kanji arranged by on reading level kanji arranged by radical katakana
hiragana numerals English characters symbols Russian characters
Greek characters and line elements The separation of the kanji into levels was
to accommo date the systems which did not want the exp ense of storing the graphical
representation of many rarely used characters The adoption of the on reading order for
the level kanji was to assist the users of the very early word processor systems where
character entry was by ordinal index The kana with diacritic marks are included as
separate characters as are a selection of small kana
A supplemental character set JIS X has also b een established which
contains a further rarely used kanji
The raw JIS co des cannot b e used in text do cuments as they coincide with normal
ASCI I co des Three co de mo di cation systems are commonly employed to distinguish
b etween the enco ded Japanese and normal bit co des
JIS co de This is a bit encapsulation metho d recommended in the JIS standard
do cument In it a byte kanji in shift sequence is prep ended to a segment of
Japanese text and a byte kanji out sequence app ended The raw JIS co des are
used b etween the two shift sequences This metho d is commonly used for do cument
les which are b eing transmitted over communications networks as only bits of
each byte are used but is less often used within computer systems as the shift
sequences complicate the text handling
EUC Extended Unix Co de also known as AT T JIS This is an bit co de in
which is added to the two raw JIS co des thus setting the most signi cant bit
of each of the bytes to EUC is used in Unix systems and in some MS DOS PC
systems
Shift JIS also known as MS KANJI This is also an bit co de and also has the
most signi cant bit of the rst byte set to one but distributes the information
content of the raw JIS co des unevenly over the two bytes This is done to maintain
compatibility with the earlier half width enco ding system for kana Shift JIS is
used in Macintosh systems the NEC series of PCs and DOS V the Japanese
version of MS DOS PC op erating system
App endix C
The Unico de Co ding System
The following information ab out Unico de was provided by Lee Collins at Taligent
The Unico de sequences are the nal o cial mapping to JIS of the CJK JRG s Chinese
Japanese Korean Joint Research Group Uni ed Rep ertoire and Ordering Version
which is the uni ed Han character set of ISO and Unico de All of the Unico de
companies Apple IBM Microsoft NeXT Taligent etc are now using this mapping
There has b een some confusion b ecause of di erence in nomenclature Unico de p eople
call it UniHan the Chinese sometimes call it HCS Han Character Set and ISO calls
it Ideographic CJK Character Uni ed Rep ertoire and Ordering ISO cannot use the
term Han character b ecause Japan was very sensitive to this even though it is a direct
translation of Kanji and it cannot b e called a character set b ecause only ISO WG is
emp owered with the authority to enco de characters Problems of naming aside they are
all the same thing
The CJK JRG was formed under the aegis of ISO in to investigate and prop ose a
uni ed Han character set for inclusion in ISO It brought together various exp erts
on Han characters from China Hong Kong Japan Korea Taiwan and the United States
selected by the national b o dies participating in ISO WG Including the initial work in
the US on Unico de and in China on GB which were merged and b ecame the basis
for the URO the task spanned ab out years The work was completed in April
It contains Han characters from all of the ma jor standards used in East Asia
including JIS X and JIS X
The Unico de consortium provides a cross reference le for all of the source sets
For further details ab out the URO UniHan refer to the The Unico de Standard Ver
sion Vol I I published by Addison Wesley ISBN For a slightly di erent
presentation of the characters a copy of or of the Ideographic CJK Character
Uni ed Rep ertoire and Ordering Version might b e available through the lo cal repre
sentative on ISO WG
Biblio graphy
J V Neustupn y Introduction to Japanese Writing Melb ourne Japanese Studies
Centre
P G O Neill and S Yanada An Introduction to Written Japanese London English
Universities Press
J M Unger The Fifth Generation Fallacy New York Oxford University Press
C Seeley The japanese script since Visible Language pp July
K R Lunde Understanding Japanese Information Processing Sebastap ol CA
O Reilly Asso ciates
I Shinmura Kojien Tokyo Iwanami Shoten
T Morohashi Daikanwajiten Tokyo Taishukan Shoten
T Iwasaki and J Kawamura New English Japanese Dictionary Tokyo Kenkyusha
M Shimizu and S Narita Japanese English Dictionary Tokyo Ko dansha
M Takahashi Romanized English Japanese Dictionary Tokyo Taiseido
F Tamamura Practical Japanese English Dictionary Tokyo Asso ciation for Over
seas Technical Scholarship
Kenkyusha s Furigana English Japanese Dictionary Tokyo Kenkyusha
A N Nelson The Modern Reader s Japanese English Character Dictionary Tokyo
Tuttle
M Spahn and W Hadamitzky Japanese Character Dictionary Tokyo Nichigai
Asso ciates
A Skoutarides and G Peters Nihongo Dictionary for Japanese I Melb ourne Swin
burne Institute of Technology
W Hadamitzky Lehrbuch und Lexikon der japanischen Schrift Ko eln Langen
scheidt
J Halp ern New Japanese English Character Dictionary Tokyo Kenkyusha
K G Henshall A Guide to Remembering Japanese Characters Tokyo Tuttle
A Rose Innes Beginners Dictionary of Chinese Japanese Characters Yokohama
Yoshikawa Shoten
Z Ueda Kanwajiten Tokyo Ko dansha