We describe the development of RuleML to Version 1.02 in order to integrate a subset of features from Deliberation and Reaction RuleML into a new family of languages called Consumer RuleML, designed to be embedded into other languages such as LegalRuleML. Deliberation RuleML 1.01 increases the resolution of the language lattice of this family with re ned language options, e.g. for Datalog+ and Hornlog+. There is no corresponding Version 1.01 release of Reaction RuleML.
RuleML4 is a knowledge representation architecture designed for the interchange of the major kinds of Web rules in an XML format that is uniform across various rule logics and platforms. It has broad coverage and is speci ed as a system of extensible language families, whose modular de nition of schemas permits rule interchange with high precision. Realizing the \overarching" design in [BPS10], RuleML 1.0 spans the complementary families of Deliberation RuleML 1.0 5 and Reaction RuleML 1.0 6 [Pas14].
RuleML 1.02 encompasses updated versions of the existing Deliberation and Reaction families to Deliberation RuleML 1.02 7 and Reaction RuleML 1.02 8. The highlight of the RuleML 1.02 release is the new Consumer RuleML9 family, an initial integration of Deliberation and Reaction RuleML, as shown in Fig. 1. 2.1
Consumer RuleML obtained its name from being considered a \consumer" in two respects.
On one hand, Consumer RuleML is designed to be embedded into other languages. In particular, the absence of containers (e.g., <Rulebase>) and performatives (e.g., <Assert>, <Query>) means Consumer RuleML is a consumer, for structure and pragmatics, of containers and performatives from other languages, e.g. from FIPA, SOAP, or domain-speci c XML languages such as LegalRuleML [ABG+13].
On the other hand, Consumer RuleML consumes external resources, from other syntactic and semantic speci cations, accessed by attributes linking to external de nitions of syntactic constraints (@type), semantic pro les (@style) and extended quanti cations (@closure). The external resources may be, for example, de nitions in controlled natural language (e.g., in mathematical English) or formal speci cations expressed in Reaction RuleML.
The following example shows a Consumer RuleML 1.02 rule that uses the specialized temporal syntax of Reaction RuleML to support an inference about the temporal scope of German climate data in dbpedia.10 <fipa-message act="inform"> <sender> <agent-identifier>
<name id="RuleML"/> </agent-identifier> </sender> <receiver> <agent-identifier>
<name id="Tara-Athan"/> </agent-identifier> </receiver> <content xmlns:ruleml="http://ruleml.org/spec" xmlns:rs="http://ruleml.org/1.02/profiles/" xmlns:xs="http://www.w3.org/2001/XMLSchema" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" 7 Speci cation of Deliberation RuleML 1.02: http://deliberation.ruleml.org/1.
02/ 8 Speci cation of Reaction RuleML 1.02: http://reaction.ruleml.org/1.02/ 9 Speci cation of Consumer RuleML 1.02: http://consumer.ruleml.org/1.02/ 10 http://dbpedia.org xmlns:agg="http://aggregation-ontology.org/" xmlns:clim="http://climate-ontology.org/"> <!-- The Consumer RuleML content starts here.--> <ruleml:Rule closure="universal" style="rs:psoa"> <ruleml:if> <ruleml:Atom> <!-- A PSOA frame representation of January climate data for some location "location" with a value of "temp" using the http://dbpedia.org vocabulary.--> <ruleml:oid>
<ruleml:Var>location</ruleml:Var> </ruleml:oid> <ruleml:Rel iri="http://dbpedia.org/ontology/Place"/> <ruleml:slot> <ruleml:Ind iri="http://dbpedia.org/property/janMeanC"/> <ruleml:Var>temp</ruleml:Var> </ruleml:slot> <ruleml:slot> <ruleml:Ind iri="http://dbpedia.org/ontology/country"/> <ruleml:Ind iri="http://dbpedia.org/resource/Germany"/> </ruleml:slot> </ruleml:Atom> </ruleml:if> <ruleml:then> <ruleml:Operation type="agg:WindowedRecurring" style="agg:profile"> <!-- A complex temporal modal operation specifying an aggregation model.
An assumption is made that the dbpedia information is in agreement with the most recent climate data from the offical national source.--> <ruleml:Atom> <!-- A formula representing a monthly climate datum
(temp at location).--> <ruleml:Rel iri="clim:climateMeanSurfaceAirTemperatureCelsius"/> <ruleml:Var>location</ruleml:Var> <ruleml:Var>temp</ruleml:Var> </ruleml:Atom> <ruleml:Interval> <!-- This argument of the modal operation brings in the additional knowledge that the most recent averaging period for German climate data is 1981-2010. This knowledge is not directly available from dbpedia, but is from the Deutscher Wetterdienst, which is the provenance of most Wikipedia climate data for Germany. --> <ruleml:Data xsi:type="xs:gYear">1981</ruleml:Data> <ruleml:Data xsi:type="xs:gYear">2010</ruleml:Data> </ruleml:Interval> <ruleml:Time> <!-- The month of the datum is January, expressed in the xs:gMonth datatype. --> <ruleml:Data xsi:type="xs:gMonth">--01</ruleml:Data> </ruleml:Time> <!-
This rule has validity until around 2020, when the next climatological dataset should become available.
Such a qualification could have been explicitly represented using Reaction RuleML's "qualification" feature, but this has not yet been included in Consumer RuleML.
--> </ruleml:Operation> </ruleml:then> </ruleml:Rule> </content> </fipa-message> The premise of this rule would unify with the following fact, derived from dbpedia. While the fact's elevation slot has no counterpart in the rule premise, this is taken care of by the \look-in"/\slotribution" semantics [Bol15] speci ed via "rs:psoa". <ruleml:Ind iri="http://dbpedia.org/resource/Germany"/> </ruleml:slot> </ruleml:Atom> </content> </fipa-message>
As shown in Fig. 1, while Consumer RuleML 1.02 introduces no new language features, it does integrate some features that are not held in common by Reaction and Deliberation RuleML, and thus a particular valid Consumer RuleML fragment may not validate against either Reaction or Deliberation RuleML schemas. 2.2
The Deliberation RuleML modular schema design was created to allow additions to the language lattice (while maintaining backward compatibility). Sublanguages in the Deliberation RuleML language lattice are de ned by a set of schema modules, and are identi ed by a myng-code, a compact representation of the MYNG REST query11, which may be used to obtain the driver schema that includes the de ning set of schema modules into one schema. The myngcode is composed of ten components, representing collections of modules that have some common characteristics. Each myng-code component corresponds to a REST query parameter. For example, the supremum language of the lattice, containing all other Deliberation RuleML 1.02 sublanguages, is called and its Relax NG schema is available from12 http://deliberation.ruleml.org/1.02/
myng-b3f-d7-a7-l1-p3ff-i7f-tf3f-q7-ef-s4f.rnc which is redirected to the REST call http://deliberation.ruleml.org/1.02/relaxng/schema_rnc.php? backbone=x3f& default=x7& termseq=x7& lng=x1& propo=x3ff& implies=x7f& terms=xf3f& quant=x7& expr=xf& serial=x4f 11 The PHP script at http://deliberation.ruleml.org/1.02/relaxng/schema_rnc. php implements the MYNG REST interface to the parameterized schema for Deliberation RuleML 1.02. 12 .rnc is the lename extension for Relax NG mode scope event, situation, action, fluent, algebra CEP CA EA Allen style spatial, temporal, interval generalized If-then rules generic negation import answer test
V III foreign
II
I
Consumer RuleML safety arity
qualification message consult extended quantification
= orientation cardinality Reaction RuleML
Deliberation RuleML
metadata performatives entailment direction rulebase weight
per node
FOL IV
VI
VII terms generic operation key/keyref type
IRI
Horn logic
= closure negation material => uncertainty Fig. 1. The syntactic containment relationships between the Reaction, Deliberation and Consumer families are shown in this Venn diagram. The lled ellipses indicate sets of language constructs. The stars indicate language features, e.g. rules, rst-order logic expressivity, or performatives, that may be combined to generate language constructs. The diagram indicates that there are no language features exclusive to Consumer RuleML { these would appear in region VII. However, because Consumer RuleML integrates some features that are not held in common by Reaction and Deliberation RuleML (regions III and IV), there are constructs (in region VII) in Consumer RuleML that are neither Deliberation nor Reaction RuleML constructs.
With the exception of the d component (default query parameter), each myng-code component corresponds to a GUI facet. In earlier versions, the d component served to con gure options regarding attributes with default values, in the sense of XSD [AB14]. In RuleML 1.02, attributes no longer have syntactic default values13, so this facet of the MYNG GUI has been omitted in Version 1.02. However, it is retained in the myng-code, with a restricted range of values, for backward compatibility. 2.3
Deliberation RuleML 1.02 is equipped with an improved MYNG GUI 1.02 for con guring a Deliberation sublanguage with a set of desired features. The dynamic output of the MYNG GUI (in Versions 1.01+) includes several components: { The myng-code and possibly an anchor14 of the form selection { The MYNG REST interface URL for a Relax NG schema of the selected language { An abbreviated URL for the Relax NG schema using the myng-code { The URL for the XSD schema of an anchor { A display of the generated Relax NG schema driver (on demand only)
Relax NG schemas con gured using MYNG (online through the MYNG REST interface URL or as local copy after download) may be used outside of MYNG for schema-aware authoring, instance validation, or parser generation through XML tools such as oXygen XML Editor and JAXB. New MYNG GUI functionality has been added in Version 1.02, including { Fill and Clear buttons for the form as a whole as well as expressivity-speci c
Fill/Clear and Check/Uncheck All for a number of facets 2.4
Reaction RuleML is a standardized rule markup/serialization language and semantic interchange format for reaction rules and rule-based event processing [BPS10, Pas14, PB09]. Reaction rules include spatial/temporal Derivation Rules (DR), Knowledge Representation (KR) calculi / logics for temporal/event/action/situation reasoning, as well as Event-Condition-Action (ECA) rules, Production (CA) 13 See http://consumer.ruleml.org/1.02#Semantic_Variant_Attributes for a mechanism of specifying semantic default values through semantic pro les. 14 A few Deliberation RuleML sublanguages of special signi cance have been designated as anchor sublanguages (or simply anchors); XSD schemas (automatically generated in advance by o ine scripts) are made available for anchors. An anchor of a Deliberation RuleML sublanguage is a minimal Deliberation RuleML anchor that contains the sublanguage. The anchor of myng-b3f-d7-a7-l1-p3 -i7f-tf3f-q7-ef-s4f is na ologeq. rules, and Trigger (EA) rules, and distributed rule-based Complex Event Processing (CEP). Reaction RuleML 1.02 incorporates this reactive spectrum of rules into RuleML, employing a system of step-wise extensions of the Deliberation RuleML 1.02 foundation. 2.5
In Fig. 1, the overlapping feature sets of the three families create seven disjoint feature regions, which will be unpacked below. The union of those sets constitutes the complete feature set of RuleML 1.02.
{ Region I (Deliberation, Reaction, and Consumer) Deliberation, Reaction, and Consumer RuleML 1.02 have a number of features in common: Terms, including individuals, data terms, variables, skolem constants, rei ed terms, generalized lists, and functional expressions Atomic formulas, including polyadic positional arguments, slots, rest variables, and object identi ers Horn logic expressivity, including logical connectives and quanti cation over variables Equations Negations, including classical (strong) and negation-as-failure (weak) forms, with, e.g., well-founded or answer-set semantics Degree of uncertainty modi ers, which may be used, e.g., for fuzzy or probabilistic logics Type modi ers, which may be used, e.g., for sorted logics or as an abbreviation for classi cation through unary predicates Material or non-material implications IRIs as constants Implicit quanti cation (closure), including reference to externally de ned extended quanti ers such as numerical quanti cation
Mechanism for distributed de nitions (@key and @keyref) { Region II (Reaction and Deliberation) Some features common to Reaction and Deliberation RuleML 1.02 were intentionally kept out of Consumer RuleML 1.02 in order to minimize its complexity and maintain Consumer RuleML's focus on semantics over pragmatics:
Reasoning direction modi ers Metadata annotations Basic performatives (Assert, Retract, Query) Rulebase containers Metalogical entailment assertions Equation orientation modi ers Functional expression interpretation modi ers (per), which may be used, e.g., to specify that a built-in functional expression should be replaced by its value Cardinality modi ers, which may be used, e.g., to specify that a functional expression or slot has a certain multiplicity of its values
Slot weights { Region III (Reaction and Consumer) Some advanced features from Reaction RuleML have been made available in Consumer RuleML 1.02, although they have not (yet) been incorporated into Deliberation RuleML:
Rules (<Rule>) generalizing <Implies>), which may be used, e.g., for defeasible rules Specialized syntax dedicated to spatial, temporal, and interval terms and formulas Allen interval operators Generic negations Generic operations, e.g. for modal logic Syntax for indicating the semantic pro le (@style)
Extensibility through foreign namespaces (content) { Region IV (Deliberation and Consumer) Consumer RuleML 1.02 follows Deliberation RuleML in regards to expressivity:
First-order logic expressivity { Region V (Reaction only) Reaction RuleML 1.02 contains advanced features that are not available in either Deliberation or Consumer RuleML: Event, situation, action, and uent terms, formulas, and algebra operators Production (CA) rules Trigger (EA) and Event-Condition-Action (ECA) rules Distributed rule-based Complex Event Processing (CEP) Actions/performatives for semantic import/consult, updates, answers, VVI tests Message descriptors Interface de nitions and attributes for ∗ scope, ∗ mode ∗ safety ∗ arity ∗ cardinality ∗ quali cations ∗ quanti cation { Region VI (Deliberation only) Deliberation RuleML 1.02 has one feature that is not contained in either Reaction or Consumer RuleML:
IRI labeling denoting abstract propositions and rules (@node) 3
Deliberation RuleML 1.02 is largely backward compatible with Version 1.01, with the exception of the following: { The content model of attributes whose values may denote Web links (i.e. @iri, @node, and @type) no longer includes relative IRI references. Instead, the CURIE syntax is used to abbreviate IRIs. The upgrader XSLT transformation15 to Deliberation RuleML 1.02 expands relative IRI references to IRIs.
Reaction RuleML 1.02 is also largely backward compatible with Version 1.0, with the exception of the following: { Operator has been renamed to Operation { truth-valued generics and operations are only used as formulas, but are no longer allowed as functional terms { act edge has been renamed and generalized to a do edge interpreting performatives uniformly as (knowledge) actions
Consumer RuleML 1.02 is the rst version of this family of RuleML sublanguages, and so has no backward compatibility properties. 4
The RuleML Wiki hosts the RuleML issue tracking system16, which is used to manage the development of the RuleML language. Users are encouraged to post bug reports and enhancement requests. Only a selection of enhancements will thus be given here.
In Deliberation RuleML Version 1.03 and beyond, we expect to see the following enhancements.
{ Incorporate declarative features developed in Reaction RuleML, including spatial, temporal, and interval terms and formulas, semantic pro les, Allen operators and modal operators (or, moving them into a common `root' family from where they can be inherited by the speci c families). { Generalize the (RNC and XSD) schemas, MYNG, and the formatters (normalizers and compacti ers) according to the PSOA syntax and semantics in a manner combinable with the current POSL syntax and semantics [Bol15]. { Work upward, beyond rst-order logic, in the deliberation expressivity hierarchy [BPS10], providing schemas etc. for higher-order and modal logics.
Of highest priority in the development of Reaction RuleML is the update of the translator service framework [PBZ+12] to Reaction RuleML 1.02 so that it can be used, e.g., in Rule Responder [PB14, ZTP+12]. The initial goal is to have the Prova translators updated for 1.02, followed by JAXB translators, e.g. into Drools and other rule engine languages. Furthermore, the development of semantic pro les for reaction rules, events, and actions [Pas14], semantic pro les 15 RuleML Upgraders: http://deliberation.ruleml.org/1.02/xslt/upgrader/ 16 http://wiki.ruleml.org/index.php/Category:Issues for modularization, distribution, and imports [PB14], as well as veri cation, validation, and integrity testing [Pas05]. Support for generalized RNC schema, MYNG, and formatters are in the focus of Reaction RuleML 1.03.
In Consumer RuleML Version 1.03 and beyond, we expect to see the following enhancements.
{ Broaden the coverage of Deliberation and Reaction RuleML.
In MYNG Version 1.03 and beyond, we expect to see the following enhancements. { The MYNG coverage should be extended to Reaction and Consumer RuleML. { New anchor languages should be added, based on user suggestions. { The MYNG functionality should be improved, including:
Provide a visual indicator of the (lattice) distance from an anchor to the RuleML sublanguage selected by GUI options, especially the zerodistance case of exact match between an anchor and the selected options. Implement the inverse functionality of determining the options corresponding to a given anchor or myng-coded language selected by the user. { Progress will continue towards making MYNG a language-lattice development environment :
Add validation of an instance against the con gured Relax NG schema, using, e.g., Validator.nu.17 Develop automatic determination of the lub schema for a valid instance of the supremum language, performing test validation against the penultimate Relax NG schemas containing all modules but one, and exploiting (monotonicity) properties of the language lattice.18 Add Schematron validation, e.g. using Validator.nu, to impose additional language restrictions, such as the `minus' restrictions of Datalog , which restore decidability lost in Datalog+.
Enable on-the- y conversion of Relax NG schemas, con gured using MYNG, into monolothic XSD schemas. 5
RuleML 1.02 has been speci ed in terms of its families of languages { Deliberation, Reaction, and Consumer RuleML. Consumer RuleML is seen to be a nexus of integration between Deliberation and Reaction RuleML, paving the way for an expansion of their common core of syntactic features. We expect that feedback from the further integration of Consumer RuleML into other languages, such as LegalRuleML, will foster additional development of the RuleML architecture. 17 http://validator.nu 18 http://wiki.ruleml.org/index.php/MYNG_Checker