Review Semantic web; towards Ontology based Web Application
Editor(s): Name Surname, University, Country Solicited review(s): Name Surname, University, Country Open review(s): Name Surname, University, Country
Zeeshan Ahmed*, Saman Majeed Department of Bioinformatics, Biocenter, University of Wuerzburg, Am Hubland 97074, Wuerzburg Germany
Abstract. In this review we address the importance of field i.e. Semantic Web; a mechanism of presenting information over the web in a format so that human being as well as machines can understand the semantic of context. Describing the techno- logical contributions of Semantic Web in detail, we present its main innovation i.e. Ontology, and development supporting technologies i.e. XML (eXtensible Mark-up Language), RDF (Resource Description Framework) and OWL (Web Ontology Language), offering different ways of more explicitly structuring and richly annotating Web pages. Furthermore, in this review paper, we discuss how ontology is contributing to semantic web development with the presentation of some Semantic Web application and conclude with some existing limitations need to be overcome.
Keywords: Web, Ontology, Semantic Web
1. Introduction easy to go for scattered extensive information by looking into bookmarked web pages but quite diffi- Targeting the challenge of implementing a web cult to extract a piece of needed information. Al- based system capable of performing semantic based though some search engines and screen scrapers are search to extract desired information from attached invented, search engine uses full text query to search repositories over the web, the field of Web and Se- information but can only return unstructured contents mantic Web is explored, as it promotes the imple- not the actual structured information stored in data- mentation of semantic based web applications by base on web where as screen scrapers extracts and providing the concept of structuring of data over the repurpose fragments from web pages but insufficient web to take advantage in extracting semantic based in creating a rich multi domain information environ- information. World Wide Web is a global informa- ment [2]. Most of the search engines are not satisfac- tion sharing and communication system made up of tory because they require excessive manual pre- three standards Uniform Resource Identifier (URL), processing e.g. designing a schema, cleaning raw Hypertext Transfer Protocol (HTTP) and Hypertext data, manually classifying documents into taxonomy Mark-up Language (HTML) by Tim Berners-Lee to and manual post processing e.g. browsing through effectively store, communicate and share different large result lists with too many irrelevant items forms of information. The Information is provided [11].To increase the integration and interoperability over the web in text, image, audio and video formats over the web the concept of “Web Service” was in- using HTML, considered unconventional in defining troduced [27, 28]. Due to the dynamic nature web and formalizing the meaning of the context. services became very famous in industry in short Most of the data is structured only inside the avail- time but with the passage of time due to the heavily able databases over the web and due to this it is quite increase in number of web services end-to-end ser-
* Zeeshan Ahmed, E-mail: [email protected] vice authentication, authorization, data integrity and collect, manipulate and annotate information inde- confidentiality problems were identified which are pendently by providing effective access to the infor- still alive and not handled by existing web technolo- mation. Semantic web provides categorization and gies [1]. uniform access to resources, promoting the transfor- HTML documents are formatted such that these mation of World Wide Web into semantically mod- cannot be processed semantically because these are elled knowledge representation systems and common only available in a readable format. This deficiency framework which allows data to be shared and reused leads to the problems of searching, extracting, main- [7]. Semantic web also gives the concept of semantic taining, uncovering and viewing the knowledge based web services to provide solutions to the prob- based information over the web. More over this for- lems of dynamically composed service based appli- mat deficiency becomes the major cause of some cations. semantic problems and the need of some other ap- Research in the field of Semantic Web research proach which will publish data over the web in not depends on a number of key methodologies i.e. only the readable but also in a processable format. knowledge representation languages or reasoning Because if data will be available in Meta data (read- algorithms [32]. Currently, semantic web is standing able and processable data format) then it will im- on a very important building block Ontology [6]. prove the process of search, extraction and mainte- Moreover semantic web aims at providing informa- nance of data over the web. To take advantage of tion in machine processable semantic models which interactive information sharing, interoperability and assigns information resources to classes whose mean- user centred design web application development ing is defined in ontologies [8], a collection of inter- Web 2 was introduced, which then improved to the related semantic concepts. Ontology is the explicit concept of Web 3 to include transformation of the representation and description of already available Web into a database to provide accessibility of the finite sets of terms and concepts used to make the contents by multiple non browser applications [19]. abstract model of a particular domain, described. Then continuing the streak of advancement in exist- Moreover, along with the processing ability semantic ing web and to cope with the currently existing web web agent is capable of communicating, receiving problems i.e., Information filtration, security, confi- and transferring information to different sources dentiality and augmentation of meaningful contents (agent or human). in mark-up presentation, the concept of “Semantic The main and currently not achieved goal of se- Web” was proposed by Tim Berners Lee [3]. Seman- mantic web is to structure the meaningful contents of tic Web is also renowned as the modified version of unstructured published data over web to take advan- Web3. tage in improving the search process [3] and to in- The remainder of this review is organized as fol- volve knowledge management in making some more lows: section 2 discusses Semantic Web in detail, advanced knowledge modelled management systems. section 3 presents Ontology, section 4 presents some No doubt semantic web using ontology has contrib- Ontology supporting development languages and uted in the progress of web but still there are some section 5 describes some Semantic Web. We present limitations and due to them semantic web is currently some Semantic Web (Ontology) limitations in sec- not successful in attaining the actual goal of com- tion 6 and then section 7 shortly concludes the dis- pletely structuring the information over the web cussion. which can be processed by machines and making advanced knowledge modelled system. The need is 2. Semantic Web to enhance the concept of ontology with respect to development point of view because all the theories Semantic Web is a mechanism of presenting in- can be fruitful if the implementation is possible. Se- formation over the web in a format so that human mantic Web is a mechanism of presenting informa- being as well as machines can understand the seman- tion over the web in a format so that human being as tic of context. Semantic web is a mesh of information well as machines can understand the semantic of con- which can be linked up in a way, so that it can easily text. Semantic web is a mesh of information which be processed by machines [5] and aim to produce can be linked up in a way, so that it can easily be technologies capable of reasoning on semi structured processed by machines [5] and aim to produce tech- information [4]. The semantic web is an intelligent nologies capable of reasoning on semi structured incarnation and advancement in World Wide Web to information [4]. The semantic web is an intelligent incarnation and advancement in World Wide Web to collect, manipulate and annotate information inde- what there is [38], mathematical formulation of prop- pendently by providing effective access to the infor- erties and relationships of certain entities [39]. On- mation. Semantic web provides categorization and tology is a main building block of Semantic Web to uniform access to resources, promoting the transfor- provide the information in machine processable se- mation of World Wide Web into semantically mod- mantic models and produce semantically modeled elled knowledge representation systems and common knowledge representation systems. It is playing a framework which allows data to be shared and reused vital role in solving the existing web problems by [7]. Semantic web also gives the concept of semantic producing semantic aware solutions. Ontology has based web services to provide solutions to the prob- two kinds of representation schemes i.e. information lems of dynamically composed service based appli- and formal [40]. Information representation method cations. provides broad range of entities and relations (Object, Research in the field of Semantic Web research Attribute, Value triples) whereas Formal representa- depends on a number of key methodologies i.e. tion method is based on family of Description Logics knowledge representation languages or reasoning [41]. algorithms [32]. Currently, semantic web is standing Ontology makes machines capable of understand- on a very important building block Ontology [6]. ing the semantic of languages that humans use and Moreover semantic web aims at providing informa- understand by producing the abstract modeled repre- tion in machine processable semantic models which sentation of already defined finite sets of terms and assigns information resources to classes whose mean- concepts involved in intelligent information integra- ing is defined in ontologies [8], a collection of inter- tion and knowledge management [9]. Ontology is related semantic concepts. Ontology is the explicit basically categorized in three different categories i.e., representation and description of already available Natural Language Ontology (NLO), Domain Ontol- finite sets of terms and concepts used to make the ogy (DO) and Ontology Instance (OI) to provide re- abstract model of a particular domain, described. lationships between generated lexical tokens of Moreover, along with the processing ability semantic statements based on natural language [25], knowl- web agent is capable of communicating, receiving edge of a particular domain and to generate auto- and transferring information to different sources matic object based web pages [9]. Ontologies are (agent or human). constructed and connected to each other in a decen- The main and currently not achieved goal of se- tralized manner to clearly express semantic contents mantic web is to structure the meaningful contents of and arrange semantic boundaries to find out required unstructured published data over web to take advan- needed information [10]. tage in improving the search process [3] and to in- Natural language based information is treated as volve knowledge management in making some more the input to the ontology construction process, which advanced knowledge modelled management systems. parses the text in nouns and verbs. Nouns are repre- No doubt semantic web using ontology has contrib- sented as “Classes” and verbs as “Properties” con- uted in the progress of web but still there are some taining values, relationships with other properties and limitations and due to them semantic web is currently some constraints. Classes are further divided in main not successful in attaining the actual goal of com- and sub class categories maintained in taxonomical pletely structuring the information over the web hierarchy. The size of ontology varies due to the in- which can be processed by machines and making crease in number of classes and instances. Ontologies advanced knowledge modelled system. The need is can be made manually from scratch, by extracting to enhance the concept of ontology with respect to information from web and by merging already exist- development point of view because all the theories ing ontologies into new ontologies. But this manual can be fruitful if the implementation is possible. process sometimes becomes very complex and time consuming especially when dealing with the large 3. Ontology amount of data. Moreover, to support the process of semantic enrichment reengineering for the building Due to the verities in use of Ontology, it has dif- of web consisting of meta data depends on the prolif- ferent definitions with respect to the different field eration of ontologies and relational meta data. This e.g. in Computer Science it defines as the combina- requires high production of meta data at high speed tion of concepts and relationships for domain model- and low cost. So in these cases machine learning ap- ling [37], in Philosophy it renowned as the study of proaches can be very helpful in generating ontologies automatically because they provide real time schemes like classification rules, instance based Definitions (DTD) and depends on data types, attrib- learning, numeric predictions, clustering, Bayesian utes, both internal and external elements structure networks and decision trees which can be very help- documents and provides syntax serialization and ab- ful in the generation of ontologies. breviation for data modelling [17]. The XML schema restricts the syntax to be only used for the structured documents, because of this XML has two main prob- lems in process of information extraction; first it is without semantic and second is the arbitrary naming and structuring of elements [20]. Figure. 1. Ontology development activities [12] Figure Legend. Six Ontology development activities i.e. De- termine Scope, Enumerate Terms, Classify Ontology, Define 4.2. RDF Classes, Define Properties and Create Instances. RDF a URL based syntax data representation pro- Ontology development is an iterative process vides a secure and reliable mechanism for the ex- based on six main activities i.e., Determine Scope, change of metadata between web applications. RDF Enumerate Terms, Classify Ontology, Define Classes, processes meta data by making an abstract data Define Properties and Create Instances as shown in model based on three object types attributes Resource, Figure 1. In the beginning of an ontology develop- Property and Statement [24]. Resource is an expres- ment process it is very important to determine the sion; Property is an attribute to de-scribe a resource scope otherwise it will be very time and effort con- where as the statement is a re-source having some suming [29]. Then enumerated terms need to be iden- property and value. RDF uses three containers, Ob- tified to classify ontologies within their respective ject bag, Sequence, and Alternative, to keep multiple types. Classes and their respective properties along available and alternative values arranged in an order with their relationships and constraints are defined in resources and properties. Bag contains resources, using identified enumerated terms. In the end only Sequence contain resources along with their proper- the instances are created and used. To implement ties having single or multiple values arranged in or- ontology development process some experience, a der and Alternative contains resources having alter- powerful user friendly ontology supporting tool and nate value(s) of a property [18]. RDF provides syntax communication between domain experts and devel- serialization and abbreviation for RDF data modeling. opers is required e.g. example LibraryWine etc [12]. Serialized syntax expresses the full capabilities of data modeling in a very regular fashion and abbrevi- 4. Ontology Supporting Languages ated syntax includes additional constructs to provide a more compact form in representing a subset of the First step in building ontologies is to create the data model. RDF is more useful than XML in ontol- nodes and edges. Once the concepts and relationships ogy construction because it provides semantic based of graph based ontology are constructed then next features for data including domain independency, step is to quantify the strengths of semantic relation- vocabulary and privileges in defining terminologies ships [11]. Ontologies can be constructed manually used in schema language. Furthermore it also pro- and automatically by using some ontology supporting vides syntax based on reification (statements about languages i.e., XML (eXtensible Mark-up Language), statements), data types, attributes, nesting, elements, RDF (Resource Description Framework) [5] and element types, element container and no restrictions OWL (Web Ontology Language) offering ways of in structuring document like XML. RDF has its own more explicitly structuring and richly annotating grammar but not complete, relies on the support of Web pages. XML to fulfill the need. Moreover RDF modeling mechanism is insufficient in expressing various logi- 4.1. XML cal statements [17].
XML is one of the fundamental contributions to- 4.3. OWL wards middleware technologies [13]. It is a markup Meta language which allows sharing of information The Web Ontology Language OWL is proposed between different applications through markup, by W3C proposal in 2004 [35]. OWL is derived from structure and transformation. As the major contribu- American DARPA Agent Markup Language tion towards semantic web, XML uses Data Type (DAML) [22]. OWL is based on ontology, inference find out the related documents of the project. The and European Ontology Interchange Language (OIL) proposed architecture mainly consists of three main [23], claims to be an extension in RDF in expressing components: Receiver, Interpreter and Analyzer. Re- logical statements [21]. The OWL provides an Ap- ceiver is used to provide index services and obtain plication Programming Interface (API) for the devel- the information about the structure of indexed files opment of Semantic Web application using Ontology with the help of so called brainFiller . Interpreter first [31], influenced by XML Document Object Model retrieves Information (structure / unstructured) using (DOM) [36], can easily be used in any OWL support- full text search and then uses so called LiveLink to ing language editor e.g. Protégé 4. OWL API pro- structure the contents of obtained information with vides number of classes and interface for OWL based the help of manual annotation and meta data, to store ontology modeling [34]. It is rich in vocabulary be- and retrieve contents based on their properties and cause it not only describes classes and properties but preferences. At the last step the Analyzer queries also provides the concept of namespace, import, car- using Jena inference engine on created RDF models dinality relationship between the classes and enumer- to infer runs and also uses F-Logic to integrate rules. ated classes. OWL has some limitations like only one To take the advantage of proposed approach by shar- “Namespace” per project is allowed, “Import” is not ing the information from search, four case scenarios currently supported, no database backend and Multi- are designed: Local search, group search, closed User support and a few OWL Language features are community and open community. Local search sce- missing [12] nario only deals with the search mechanism and can only be applied to a personal desktop, group search 5. Semantic Web Applications can be applied with n a particular network domain, closed community consists a number of users having Residing in the domain of Semantic Web many different roles but same topic where as open search products are available and several approaches have consists of users with different roles and different been introduced by many researchers which are pro- topics. viding lots of values in the implementation of seman- tic based applications [26] with use of Ontology and 5.2. Reisewissen providing structured data over the web to take advan- tage in implementing efficient web based information Reisewissen is proposed to provide quality ser- retrieval search mechanism e.g. Semantic Desktop, vices by semantically connecting, organizing and Reisewissen and Meta Data Search Layer Etc. In this sharing the isolated pieces of information by trans- section, without going in to much product detail, we piercing to data sources, caching & fetching of data, will only present some semantic web based ap- transforming data from heterogeneous to RDF mod- proaches, to take advantage in having an idea about els, mapping of ontologies between database and semantic based system development using Ontology. triples, matching RDF and non RDF based informa- tion [30]. Moreover Reisewissen is implemented us- 5.1. Semantic desktop- Personal Information Model; ing semantic web technology based Ontology (RDF), by producing RDF models and manually mapping The authors promoted the idea of stepping into ontologies during the implementation of search user’s mental model by implementing Personal In- mechanism. Reisewissen identifies potential relevant formation Model (PIM). PIM is designed to improve knowledge sources and provide quality services by the process for the identification of documents and semantically connecting, organizing and sharing the retrieval of no unnecessary document. The design is currently isolated pieces of information in an online based on ontologies and classes, the relationships of portal to anticipating customer behavior. Proposed classes and ontologies are predefined and the infor- approach is implemented using semantic web tech- mation can be accessed using RDF graphs. Four rules nologies in a project Reisewissen, a hotel recommen- based on forward changing principle are defined to dation engine and travel information system. The retrieve the information, this information is divided design of Reisewissen is composed of three main in three parts Author (single or team), Relevant pro- components Data Connectors (DC), Evaluation ject, and Relevant solution. The system works in the Framework (EF) and Evaluation Engine (EE). Data following way. First query runs aiming to find out connectors are used to provide transparency to data the project and if project is found then it moves to sources and transformation of data from heterogene- ous to common data format (RDF and Java objects), If number of nodes is more than equal to for- moreover it also provides the caching and fetching of warding degree selects subset of nodes. data. Evaluation Framework is a workbench to test For each selected node, if node contains target the quality of data and rules by providing functions document then update connectivity and if it and filters to map resources and return result in deci- doesn’t then continue search using nodes. sive format (Boolean or float value) and Evaluation Authors have explicitly mentioned that this Engine combines individual filters to rank and filter search mechanism is good but still there is a room for information by weighting and yielding the overall improvement in examining the path length of score. Information is obtained using Simple Object searches for different and same users characterized Access Protocol (SOAP) based web services and by their different query distributions. Moreover time stored in both RDF and non RDF formats, which are to converge to a stable network can be an ambiguity then matched to find out the desired result. Data and need to have more realistic simulation using pa- stored in RDF format is based on developed ontolo- rameters and distributions. gies mapped between database and RDF triples. Moreover Reisewissen uses Prolog to capture ex- pert’s knowledge which can be formalized and can 6. Limitations of Semantic Web; Ontology generate new data by implementing the customer request in evaluator encapsulated rules. Data is The development of ontology driven applica- matched semantically by combining data properties tions is difficult because of some limitations and to ontology and similarities between two concepts are principal problems which are as follows; determined by distance reflecting their respective Natural language parsers used to parse the positions in hierarchy. As a result list of selected re- in-formation to construct the ontologies are sults are generated to customer. limited because they can only work over a single statement at a time [13]. 5.3. Meta Data Layer Existing methodologies of structuring on- tologies are insufficient and need to be im- The authors have discussed a successful process proved because now it is quite impossible to for meta data search based on three questions for define the boundaries of ontology based par- information extraction: What user needs, Where it ticular domain’s abstract model and auto- lies, and How it can be retrieved. The targeted objec- matically handle the increase in size of on- tive is to identify the location from set of locations tology due to the increase in number of contained by a document and avoid looking into non- classes and instances. specific document. Scalability and efficiency of this Creating ontologies manually is a time con- approach is determined using simulation of docu- suming process which becomes very com- mented meta data keywords, location pointers, node plex when there is a large amount of data to connections and node knowledge. The whole process create large number of ontologies. To take of identifying target location and search consists of advantage in creating large number of on- nine procedural steps. tologies by reducing the complexity and Select target document from network having at time, an automatic ontology creation least one keyword. mechanism is required. Some mechanisms Use keywords contained in document for the are already proposed and implemented to construction of a search query. create ontologies automatically but they are Start with free node (not already containing any in-sufficient and less qualitative. While cre- target document). ating nouns based classes using existing Make a record of start node. automatic ontology creation mechanism, it Start node’s knowledge treated as base knowl- is quite impossible to identify the possible edge for the selection of other sub nodes. existing relationships between classes to If number of nodes is equal to forwarding degree draw the taxonomical hierarchy [14]. Fur- select those nodes. thermore it is also quite impossible to per- If number of nodes is less than forwarding de- form automatic emergence of ontologies to gree select additional nodes. create new ontologies [16]. Currently available ontology validators are References restricted and not capable of validating all kind of ontologies e.g. based on complex [1] Grit Denker, Son Nguyen, and Andrew Ton, “OWL-S Se- inheritance relationship [1]. mantics of Security Web Services: a Case Study”, J. Davies et al. (Eds.): ESWS 2004, LNCS 3053, pp. 240–253, 2004. Domain specific ontologies are highly de- sSpringer-Verlag Berlin Heidelberg 2004 pendent on the domain of the application [2] David Huynh, Stefano Mazzocchi, and David Karger, “Piggy and because of this dependency domain spe- Bank: Experience the Semantic Web Inside Your Web Browser”, Y. Gil et al. (Eds.): ISWC 2005, LNCS 3729, pp. cific ontologies contain specific senses 413 – 430, 2005. which are not possible to find in general [3] Tim Berners-Lee, James Hendler and Ora Lassila, “The Se- purpose ontology [15]. mantic Web, A new form of Web content that is mean-ingful The process of semantic enrichment reengi- to computers will unleash a revolution of new possi-bilities”, May 2001, Viewed February 2007, neering for web development consists of re-