Evolution of Semantic ReasonersTarun Bandarupally45534632Summary AThe authors in [1] discuss about Cwm, a rule-based forward chaining reasoner designed toquery, filter, examine and transform information based on RDF. In the field of Semantic Web,RDF thrives in presenting machine-readable documents with URIs as symbols. The tool isbased on RDF’s extension, Notation3 (N3) which supports quotes (to make statements aboutstatements), rules and variables. As it is more expressive and logical with features like RDFgraph filtering, it tends to reason with a first order logic. These features along with webdocument retrieval makes it a strong reasoner. In N3, a statement is stored in the form of thesubject, predicate, object, and its context where the latter enables it to be utilized for proofprocessing. Adding a –why to the tool’s command line generates a proof for the actual result.A python program, check.py validates a proof. In the reasoner query, light built-in functions(like negation of an integer) are included to speed up the process or even heavy built-ins likeremote queries are de-prioritized. Cwm views web as the interlinking of symbols and formulaei.e., it first retrieves document (N3 / RDF/XML format), then parses it, and returns the formula.Additionally, they demonstrated other uses of the tool such as filtering of calendar informationand bug tracking.Summary BThe authors in [2] propose a dynamic and flexible engine named Semantic Web KnowledgeBase (SWKB) as a technique to preserve, execute and augment knowledge in a multimediadescription framework. It includes a forward chaining knowledge base and an RDF/OWL Litecompatible reasoning engine, which could infer and retract knowledge in a non-monotonicway. They explain that real-world media knowledge isn’t always fully available, and fewconclusions based on limited knowledge need to be retracted to accommodate updatedknowledge and conclusions. Content-Based Retrieval systems address this issue to some extentbut fail to bridge the semantic gap in the content characterization. SWKB is stated to handleevolving knowledge in a dynamic framework. Other OWL-compatible reasoning engines likeDescription Logics classifiers, Theorem provers and Rule-based engines have been exploredwith the latter’s standing out as their style of reasoning is dynamic and within the OWL/RDFmodel. The first RDF/S Semantics layer is designed to assert stationary facts about vocabularyalong with a set of transitive rules for dynamic support. The OWL layer offers a robustmechanism for augmented reasoning. Their design follows the closed-world assumption: Amissing fact is false and not unknown, unlike OWL/RDF semantics. However, additionalresearch is being carried out in the OWL DL layer to introduce open and close world modes.Summary CThe authors in [3] present an effective solution to diverse Constraint Satisfaction Problems byleveraging the concept of OWL reasoning (Ocr). In this technique, the OWL classes representthe CSP constraints and its domain, and OWL: the state of variables. Also, constraint adherenceof variables and class-individual conflicts are monitored. The Ocr coupled with the idea ofEnumerating-Reasoning have been explored to solve the popular four colour problem: anyregion in a plane could be coloured with four-colours without any match between adjacentones. The variables are five regions, the domain is the colours and constraints are that adjacentregions have different colours. To transform it into an OWL reasoning problem, they built threeOWL classes. The first class was “ColorableBoard”, depicting characters of the coloured plane.Next, a class depicting each region with an identifier (Dr). Finally, an OWL class to define thecolour constraints (Cr) and OWL individuals (Vr). In their OWL code, they used OWLoperators like owl:differentFrom to demonstrate constraint complexities. The E-R algorithmintroduced a conflicts detection function R which adds a solution to the Ocr set if constraintsare met. This process could be generalised to any problem by extracting OWL classes fromthem.Summary DThe authors in [4] propose a temporal context reasoning model (TempCRM) built on OWLand RDF to create smart homes, by leveraging the concepts of context-awareness and semanticweb. The paper mainly focuses on detecting timely potential indoor hazards and alerting theoccupant. In addition to RDF and OWL representing the context and reasoning, first-orderpredicate logic (Subject, Verb, Object, Time) is introduced in the model for its ease ofapplication. The model architecture contains an OWL-based smart home ontology with entitieslike location/person and their relationships, a first-order predicate, a probability densityfunction which models the device usage patterns, RDF-based temporal contexts likeLocation(Peter, in, kitchen, 22:24:39) and reasoning rules. As an example, they illustrated howthe reasoning engine alerts the user with RDF-based rules when they forget to turn off gas aftercooking. The average usage behaviour of the stove with RDF-context and timestamp is fed tothe TempCRM to update the knowledge base. Their engine offers dual functionality viz, itinvokes with the entry of an RDF-temporal context and it also invokes periodically. As thehazard level could increment with time, the second function adjusts the threshold. As a part offuture research, they are planning to add a semi-automatic functionality to adapt to new kindsof scenarios in the knowledge base.DiscussionA systematic review of literature revealed that semantic reasoning engines make newrules/logical inferences from the contents like asserted facts in the knowledge base. Rules areusually If-Then statements in RDF and OWL-based reasoners and they have a predefined setof rules to ensure deduction, evaluate conjunctions and disjunctions. The literature has beencarefully selected to visualise the evolution of the reasoners. The first paper highlighted theconcept of a forward chain rule-based semantic reasoner with the help of a tool named Cwm.It follows a simple mechanism of computing all the possible implications from the given set ofasserted facts based on the RDF’s extension, Notation 3. Additionally, light built-in functions(like negation of an integer) were included in the reasoner queries to speed up the inferenceprocess. OWL and RDF-based reasoners (Cwm) have been programmed under an open-worldassumption where a what a missing fact is treated as unknown unlike false in closed-world.This contrast was addressed in the second paper with a dynamic and flexible engine namedSemantic Web Knowledge Base (SWKB) which could infer and retract knowledge in a nonmonotonic way. In practise, as complete knowledge is not always available viz, the SKWB hasthe capacity to reason with insufficient knowledge and deduce ad-hoc conclusions. With theaccumulation of knowledge, a few of the inferences are retracted and the process ensues.SKWB being rule-based, has the core structure of an RDF/OWL based reasoner where the twolayered architecture asserts stationary facts about vocabulary and performs augmentedreasoning. To contextualise these reasoners, the third paper proposed a global solution toseveral Constraint Satisfaction Problems with the help of the OWL reasoning (Ocr) andEnumerating-Reasoning technique. It also explored the process of rule generation for thereasoner to implement. For instance, it begins with a constructor (Ex: owl:allValuseFrom) andcorresponding axioms to define the properties of OWL classes (owl:isSubClassOf), and finallythe rules are generated with If-Then statements. These reasoners outperformed theconventional technique by a decent margin. As technology is swiftly evolving, context-awarenetworks and devices are becoming more prominent. They generate the desired/ higher-levelcontexts and provide automated services. The OWL/RDF based reasoners find a directapplication in creating smart homes as proposed in the fourth paper. First-order predicate logics(Ex: ContextType(, , , ) were used in conjunction with theOWL/RDF reasoners for better efficiency. Backward chain inference, which originates froman unfamiliar fact to sequence of entailments to the root known fact was briefly discussed. Eachpaper turned out to be a precursor to the next, covering the journey of semantic reasoners.References[1] Berners-Lee, Tim, et al. A Reasoner for the Web. w3. [Online] January 1, 2003.w3.org/2000/10/swap/doc/paper/index.pdf.[2] Jelmini, Carlo and Marchand-Maillet, Stephane. OWL-based reasoning with retractableinference. ResearchGate. [Online] January 2004.[3] Xiong, Haoyi and Jiang, Ying. CONSTRAINT SATISFACTION PROBLEM SOLVINGBASED ON OWL REASONING. ResearchGate. [Online] October 2008.https://www.researchgate.net/publication/266066367_CONSTRAINT_SATISFACTION_PROBLEM_SOLVING_BASED_ON_OWL_REASONING.[4] Liao, Hsien-Chou and Chien-Chih, Tu. A RDF and OWL-based temporal context reasoningmodel for smart home. ResearchGate. [Online] August 2007.https://www.researchgate.net/publication/26556985_A_RDF_and_OWLbased_temporal_context_reasoning_model_for_smart_home.
