Provided by: mmorph_184.108.40.206-12.1_i386
mmorph - MULTEXT morphology tool formalism syntax
A mmorph morphology description file is divided into declaration
sections. Each section starts by a section header (`@ Alphabets',
`@ Attributes', etc.) followed by a sequence of declarations. Each
declaration starts by a name, followed by a colon (`:') and the
definition associated to the name. Here is a brief description of each
In this section the lexical and surface alphabet are declared. All
symbols forming each alphabet has to be listed. Symbols may appear in
both the lexical and surface alphabet definition in which case it is
considered a bi-level symbol, otherwise it is a lexical only or surface
only symbol. Symbols are usually letters (eg. a, b, c) , but may also
consist of longer names (beta, schwa). Symbol names consisting of one
special character (`:' or `(') may be specified by enclosing them in
double quotes (`:' or `(').
Lexical : a b c d e f g h i j k l m n o p q r s t u v w x y z
"-" "." "," "?" "!" "\"" "'" ":" ";" "(" ")" strong_e
Surface : a b c d e f g h i j k l m n o p q r s t u v w x y z
"-" "." "," "?" "!" "\"" "'" ":" ";" "(" ")" " "
In this example, the symbol strong_e is lexical only, the symbol " "
(space) is surface only. All the other symbols are bi-level.
All the strings appearing in the rest of the grammar will be made
exclusively of symbols declared in this section.
In this section, the name of attributes (sometimes called features) and
their associated value set. At most 32 different values may be
declared for an attribute.
Gender : feminine masculine neuter
Number : singular plural
Person : 1st 2nd 3rd
Transitive : yes no
Inflection : base intermediate final
In the current version of the implementation value sets of different
attributes are incompatible, even if they are defined identically. To
overcome this restriction, in a future version this section will be
split into two: declaration of value sets and declaration of
In this section, the different types of feature structures are
declared. The attributes allowed for each type are listed. Attributes
that are only used within the scope of the tool and have no meaning
outside can be listed after a bar (`|'). The values of these local
attributes ar not stored in the database or written on the final output
of the program.
Noun : Gender Number
Verb : Tense Person Gender Number Transitive | Inflection
Typed feature structures
Typed feature structures are used in the grammar and spelling rules.
It is the specification of a type and the value of some associated
attributes. The list of attribute specifications is enclosed in square
brackets (`[' and `]').
Noun[ Gender=feminine Number=singular ]
It is possible to specify a set of values for an attribute by listing
the possible valuse separated with a bar (`|'), or the complement of a
set (with respect to all possible values of that attribute) indicated
with `!=' instead of `='.
Example: Assuming the declaration of Gender as above, the following
two typed feature structures are equivalent
Noun[ Gender=masculine|neuter ]
Noun[ Gender!=feminine ]
This section contains the rules that specify the structure of words.
It has the general shape of a context free grammar over typed feature
structures. There are three basic types of rules: binary, goal and
Binary rules specify the result of the concatenation of two elements.
This is written as:
Rule_name : Lhs <- Rhs1 Rhs2
where Lhs is called the left hand side, and Rhs1 and Rhs2 the first and
second part of the right hand side. Lhs, Rhs1 and Rhs2 are specified
as typed feature structures.
Rule_1 : Noun[ Gender=feminine Number=singular ]
<- Noun[ Gender=feminine Number=singular ]
NounSuffix[ Gender=feminine ]
Variables can be used to indicate that some attributes have the same
value. A variable is a name starting with a dollar (`$').
Rule_2 : Noun[ Gender=$A Number=$number ]
<- Noun[ Gender=$A Number=$number ]
NounSuffix[ Gender=$A ]
If needed, both a variable and a value specification can be given for
an attribute (only once per attribute):
Rule_3 : Noun[ Gender=$A Number=$number ]
<- Noun[ Gender=$A Number=$number ]
NounSuffix[ Gender=$A=masculine|neuter ]
Affix rules define basic elements of the concatenations specified by
binary rules (together with lexical entries, see the section @ Lexicon
below). An affix rule consists of lexical string associated to a typed
Plural_s : "s" NounSuffix[ Number=plural ]
Feminine_e : "e" NounSuffix[ Gender=feminine ]
ing : "ing" VerbSuffix[ Tense=present_participle ]
Goal rules specify the valid results constructed by the grammar. They
consist of just a typed feature structure.
Goal_1 : Noun
Goal_2 : Verb[ inflection=final ]
In addition to these three basic rule types, there are prefix or suffix
composite rules and unary rules. A unary rule consist of a left hand
side and a right hand side.
Rule_4 : Noun[ gender=$G number=plural ]
<- Noun[ gender=$G number=singular invariant=yes]
Prefix and suffix composite rules have the same shape as binary rules
except that one part of the right hand side is an affix (i.e. has an
Append_e : Noun[ Gender=feminine Number=$number ]
<- Noun[ Gender=feminine Number=$number ]
"e" NounSuffix[ Gender=feminine ]
anti : Noun[ Gender=$gender Number=$number ]
<- "anti" NounPrefix
Noun[ Gender=$gender Number=$number ]
This optional section contains the definition of symbol classes. Each
class is defined as a set of symbols, or other classes. If the class
contains only bi-level elements it is a bi-level class, otherwise it is
a lexical or surface class.
Dental : d t
Vowel : a e i o u
Vowel_y : Vowel y
Consonant: b c d f g h j k l m n p q r s t v w x z
This optional section contains the definition of pair disjunctions.
Each disjunction is defined as a set of pairs. Explicit pairs specify
a sequence of surface symbols and a sequence of zero or one lexical
symbol, one of them possibly empty. A sequence is enclosed between
angle brackets `<' and `>'. The empty sequence is indicated with `<>'.
In the current implementation only the surface part of a pair can be a
sequence of more than one element. The special symbol `?' stands for
the class of all possible symbols, including the morpheme and word
s_x_z_1 : s/s x/x z/z
VowelPair1: a/a e/e i/i o/o u/u
ie.y: <i e>/y
DoubleConsonant: <b b>/b <d d>/d <f f>/f <g g>/g <k k>/k <m m>/m
<p p>/p <s s>/s <t t>/t <v v>/v <z z>/z
Note that VowelPair1 and VowelPair2 don't specify the same thing:
VowelPair2 would match a/o but VowelPair1 would not.
Implicit pairs are specified by the name of a bi-level symbol or a bi-
Examples: the following s_x_z_2 and VowelPair3 are equivalent to the
above s_x_z_1 and VowelPair2 (assuming that s, x, z and Vowel are bi-
level symbols and classes).
s_x_z_2 : s x z
VowelPair3 : Vowel
In a pair disjunction all lexical parts should be disjoint. This means
you cannot specify for the same pair disjunction a/a and o/a or a/a and
In a future version this section will be split in two: simple pair
disjunctions and pair sequences.
In this section are declared the two level spelling rules. A spelling
rule consist of a kind indicator followed by a left context a focus and
a right context. The kind indicator is `=>' if the rule is optional,
`<=>' if it is obligatory and `<=' if it is a surface coercion rule.
The contexts may be empty. The focus is surrounded by two `-'. The
contexts and the focus consist of a sequence of pairs or pair
disjunctions declared in the `@ Pairs section. A morpheme boundary is
indicated by a `+' or a `*', a word boundary is indicated by a `~'.
Sibilant_s: <=> s_x_z_1 * - e/<> - s
Consonant Vowel - DoubleConsonant - * Vowel
i_y_optionnel: => a - i/y - * ?/e
Constraints may be specified in the form of a list of typed feature
structures. They are affix-driven: the rule is licensed if at least
one of them subsumes the closest corresponding affix. The morpheme
boundary indicated by a star (`*') will be used to determine which
affix it is. If there is no such indication, then the affix adjacent
to the morpheme where the first character of the focus occurs is used.
In case there is no affix, the typed feature structure of the lexical
stem is used.
s_x_z_1 * - e/<> - s NounSuffix[ Number=plural ]
This section is optional and can also be repeated. This section lists
all the lexical entries of the morphological description. Unlike the
other sections, definitions do not have a name. A definition consist
of a typed feature strucure followed by a list of lexical stems that
share that feature structure. A lexical stem consists of the string
used in the concatenation specified by the grammar rules followed by
`=' and a reference string. The reference string can be anything and
usually is used to indicate the canonical form of the word or an
identifier of an external database entry.
Noun[ Number=singular ] "table" = "table" "chair" = "chair"
Verb[ Transitive=yes|no Inflection=base ] "bow" = "bow1"
Noun[ Number=singular ] "bow" = "bow2"
If the stem string and the reference strings are identical, only one
needs to be specified.
Noun[ Number=singular ] "table" "chair"
The formal syntax description below is in Backus Naur Form (BNF). The
following conventions apply:
<id> is a non-terminal symbol (within angle brackets).
ID is a token (terminal symbol, all uppercase).
<id>? means zero or one occurrence of <id> (i.e. <id> is optional).
<id>* is zero or more occurrences of <id>.
<id>+ is one or more occurrences of <id>.
::= separates a non-terminal symbol and its expansion.
| indicates an alternative expansion.
; starts a comment (not part of the definition).
The start symbol corresponding to a complete description is named
<Start>. Symbols that parse but do nothing are marked with
`; not operational'.
<Start> ::= <AlphabetDecl> <AttDecl> <TypeDecl> <GramDecl>
<ClassDecl>? <PairDecl>? <SpellDecl>? <LexDecl>*
<AlphabetDecl> ::= ALPHABETS <LexicalDef> <SurfaceDef>
<LexicalDef> ::= <LexicalName> COLON <LexicalSymbol>+
<SurfaceDef> ::= <SurfaceName> COLON <SurfaceSymbol>+
<LexicalSymbol> ::= <LexicalSymbolName> ; lexical only
| <BiLevelSymbolName> ; both lexical and surface
<SurfaceSymbol> ::= <SurfaceSymbolName> ; surface only
| <BiLevelSymbolName> ; both lexical and surface
<AttDecl> ::= ATTRIBUTES <AttDef>+
<AttDef> ::= <AttName> COLON <ValName>+
<TypeDecl> ::= TYPES <TypeDef>+
<TypeDef> ::= <TypeName> COLON <AttName>+ <NoProjAtt>?
<NoProjAtt> ::= BAR <AttName>+
<LexDecl> ::= LEXICON <LexDef>+
<LexDef> ::= <Tfs> <Lexical>+
<Lexical> ::= LEXICALSTRING <BaseForm>?
<BaseForm> ::= EQUAL LEXICALSTRING
<Tfs> ::= <TypeName> <AttSpec>?
<VarTfs> ::= <TypeName> <VarAttSpec>?
<AttSpec> ::= LBRA <AttVal>* RBRA
<VarAttSpec> ::= LBRA <VarAttVal>* RBRA
<AttVal> ::= <AttName> <ValSpec>
<VarAttVal> ::= <AttName> <VarValSpec>
<ValSpec> ::= EQUAL <ValSet>
| NOTEQUAL <ValSet>
<VarValSpec> ::= <ValSpec>
| EQUAL DOLLAR <VarName>
| EQUAL DOLLAR <VarName> <ValSpec>
<ValSet> ::= <ValName> <ValSetRest>*
<ValSetRest> ::= BAR <ValName>
<GramDecl> ::= GRAMMAR <Rule>+
<RuleDef> ::= <RuleName> COLON <RuleBody>
<RuleBody> ::= <VarTfs> LARROW <Rhs>
| <Tfs> ; goal rule
| LEXICALSTRING <Tfs> ; lexical affix
<Rhs> ::= <VarTfs> ; unary rule
| <VarTfs> <VarTfs> ; binary rule
| LEXICALSTRING <Tfs> <VarTfs> ; prefix rule
| <VarTfs> <Tfs> LEXICALSTRING ; suffix rule
<ClassDecl> ::= CLASSES<ClassDef>+
<ClassDef> ::= <LexicalClassName> COLON <LexicalClass>+
| <SurfaceClassName> COLON <SurfaceClass>+
| <BiLevelClassName> COLON <BiLevelClass>+
<LexicalClass> ::= <LexicalSymbol>
<SurfaceClass> ::= <SurfaceSymbol>
<BiLevelClass> ::= <BiLevelSymbolName>
<PairDecl> ::= PAIRS <PairDef>+
<PairDef> ::= <PairName> COLON <PairDef>+
<PairDef> ::= <PairName> COLON <Pair>+
<Pair> ::= <SurfaceSequence> SLASH <LexicalSequence>
SurfaceSequence ::= LANGLE <SurfaceSymbol>* RANGLE
LexicalSequence ::= LANGLE <LexicalSymbol>* RANGLE
<SpellDecl> ::= SPELLING <SpellDef>+
<SpellDef> ::= <SpellName> COLON <Arrow> <LeftContext> <Focus>
<LeftContext> ::= <Pattern>*
<RightContext> ::= <Pattern>*
<Focus> ::= CONTEXTBOUNDARY <Pattern>+ CONTEXTBOUNDARY
<Pattern> ::= <Pair>
<Constraint> ::= <Tfs>
<Arrow> ::= RARROW
<AttName> ::= NAME
<BiLevelClassName> ::= NAME
<BiLevelSymbolName> ::= NAME | SYMBOLSTRING
<LexicalClassName> ::= NAME
<LexicalName> ::= NAME
<LexicalSymbolName> ::= NAME | SYMBOLSTRING
<PairName> ::= NAME
<RuleName> ::= NAME
<SpellName> ::= NAME
<SurfaceClassName> ::= NAME
<SurfaceName> ::= NAME
<SurfaceSymbolName> ::= NAME | SYMBOLSTRING
<TypeName> ::= NAME
<ValName> ::= NAME
<VarName> ::= NAME
Simple tokens of the BNF above are defined as follow: The token name on
the left correspond to the literal character or characters on the
In the section header tokens above, spaces may separate the `@' from
the reserved word.
is any sequence of letter, digit, underline (`_'), period (`.')
is a string of lexical symbols
is a string of surface symbols
is a string of just just one character (used only in alphabet
A string consist of zero or more characters within double quotes (`"').
Characters preceded by a backslash (`\') are escaped (the usual C
escaping convention apply). Symbols that have a name longer than one
character are represented using a SGML entity like notation:
`&symbolname;'. The maximum number of symbols in a string is 127.
"double quote is \" and backslash is \\"
"escape like in C : \t is ASCII tab"
"escape with octal code: \011 is ASCII tab"
Tokens can be separated by one or many blanks or comments.
A blank separator is space, tab or newline.
A comment starts with a semicolon and finishes at the next newline
(except when the semicolon occurs in a string.
Inclusion of files can be specified with the usual `#include'
will splice in the content of the file verb.entries at the point where
this directive occurs.
The `#' should be the first character on the line. Tabs or spaces may
separate `#' and `include'. The file name must be quoted. Only tabs
or spaces may occur on the rest of the line. Inclusion can be nested
up to 10 levels.
G. Russell and D. Petitpierre, MMORPH - The Multext Morphology Program,
Version 2.3, October 1995, MULTEXT deliverable report for task
Dominique Petitpierre, ISSCO, <email@example.com>
The parser for the morphology description formalims above was written
using yacc (1) and flex (1). Flex was written by Vern Paxson,
<firstname.lastname@example.org>, and is distributed in the framework of the GNU
project under the condition of the GNU General Public License
Version 2.3, October 1995 MMORPH(5)