-
Notifications
You must be signed in to change notification settings - Fork 34
Grammar Format Specification
A grammar file must start with the grammar name and tier in the following format:
; grammar name MyGrammar
; grammar tier High
Accepted values for the tier are Easy, Moderate, and High.
After defining the grammar's name and tier, other grammar files can be imported with the following directives:
; load path/to/grammar/file.txt
; import path/to/grammar/file.txt
; import path/to/grammar/file.txt as $nonterminal
The first directive loads all the non-terminals in the referred
grammar file as if they were explicitly in the file being loaded
with the same effect of concatenating both files. The second
directive acts like the first one but skipping the $Main non-
terminal of the second grammar, making non-terminals available for
re-use, but not being attached to the production graph. In
contrast the third directive makes accessible only the $Main
rule of the imported grammar via $nonterminal (other imported
rules are internally renamed).
Remarks:
-
$nonterminalcan be any valid non-terminal. - The path to the grammar file may be enclosed within double
quotes and be a relative path. If the file cannot be found,
a
{void}wildcard (see below) will be used, specifying the nature of the error as meta-data.
The root rule S of the grammar is the $Main Non-Terminal. In
order for the grammar to be valid, it must contain at least one
$Main rule.
A valid Non-Terminal identifier is a sequence of one or more
letters, digits, or underscore characters (_). Spaces, punctuation
marks, and symbols cannot be part of a Non-Terminal identifier.
Wilcards are strings within braces which are used during random sentence generation, replacing them by random values. The same random value may be specified for a set of wildcards by specifying an ID. Also, the type may be optionally specified. Syntax is as follows (elements enclosed within brackets are optional):
{wildcard [type] [id] [where condition] [meta:[metadata]]}
When the where literal string is found within a wildcard, the
condition following it is evaluated while looking for replacements
(see Filtering Results).
When the meta: literal string is found within a wildcard, all
text following it is addressed as metadata. While metadata is not
part of the generated sentence, it can be used to display
additional information.
The following wildcards are defined:
{category} An object category.
Types are ignored.
{gesture} A gesture name.
Types are ignored.
{location} Any location.
Types may be [room|placement|beacon]
{name} A person name.
Types may be [male|female]
{object} An object name.
Types may be [known|alike]
{question} The "question" literal string.
Types are ignored. Metadata contains a question/answer
from the predefined questions set.
{void} Void construct used to inject metadata.
Types are ignored.
Also, the following alias are defined:
{beacon} Alias for {location beacon}
{aobject} Alias for {object alike}
{female} Alias for {name female}
{kobject} Alias for {object known}
{male} Alias for {name male}
{placement} Alias for {location placement}
{room} Alias for {location room}
{sobject} Alias for {object special}
Replacements can be filtered by adding conditions after the
where literal string. Such replacements must target an
specific property of the type used for wildcard replacement.
For instance, consider the following examples:
{kobject where Category="Food"}
{object where Category="Food" and Type="Known"}
Both wildcards are equivalent and will be replaced by a random known object from the Food category, if there is such category. Conditions support wildcard nesting, for which the following example is also valid.
$Main = {object where Category="{category 1}""} belongs to {category 1 where Name!="Cleaning Stuff"}
In cases when the query returns no results, the condition can't be met and another rule will be attempted. This is the case when a property is not present, so it can't be evaluated. Exceptions to this rule are 1) evaluating a property against literal "null" (without quotes), which is evaluated as true when the property is absent; and 2) evaluating an non-existing property against a boolean (true or false), which will implicitly assume that the property exists and has its value set to false.
A wildcard ended with a question mark will be replaced by an obfuscated value instead of the random value chosen by that wildcard, which will be included as metadata. The obfuscation is performed as follows (wildcards not shown can not be obfuscated):
{category?} The "objects" literal string
{location?} The "room" literal string for rooms, for placements
and beacons, the room to which the location belongs.
{object?} The category to which the object belongs.
Wildcards can be referenced with the {pron} construct.
By default, the {pron} construct refers to the last wilcard found
in the following order:
-
{name}wildcards -
{object}wildcards - last wildcard found
By default the {pron} construct is replaced with a Personal Pronoun (OBJECTIVE CASE), namely: him, her, it. Subjective cases of the personal pronoun (he, she, it) are also supported by setting the type of the {pron} construct to the sub value. Namely:
{pron} Reference with pronoun, objective case.
{pron obj} Reference with pronoun, objective case (explicit).
{pron sub} Reference with pronoun, subjective case.
Likewise, possessive pronouns can be specified in both cases, absolute (mine, yours, theirs) and adjective (my, your, their) which is default. Shortcuts are defined for convenience as shown below:
{pron pos} Reference with possessive pronoun, adjective case.
{pron pab} Reference with possessive pronoun, absolute case.
{pron paj} Reference with possessive pronoun, adjective case.
{pron posabs} Reference with possessive pronoun, absolute case.
{pron posadj} Reference with possessive pronoun, adjective case.
C/C++/Java comments are permitted. In addition, any words, spaces, or marks after # or % or ; are ignored