Presentation on theme: "A Semantic Web Browser for Supporting Open-Corpus Linking and Adaptive Hypermedia Melike Şah Intelligence, Agents and Multimedia Group School of Electronics."— Presentation transcript:
A Semantic Web Browser for Supporting Open-Corpus Linking and Adaptive Hypermedia Melike Şah Intelligence, Agents and Multimedia Group School of Electronics and Computer Science University of Southampton email@example.com Supervisors: Prof Wendy Hall, Prof David C De Roure SemWeB –
2 Outline Background – Semantic Web technologies – Adaptive Hypermedia SemWeB – Semantic Web Browser Conclusions
3 Semantic Web – an extension of the current Web, in which information is given well-defined meaning –is an extension of Web principles from documents to data – technology for creating and sharing data – Powerful knowledge representation formalisms – Inferencing mechanisms – Interoperability – Is a global information space for inter-linked data (linked data) Semantic metadata is on the Web now!
4 Semantic Web is a reality (Linked Data) Linked data - for exposing, sharing and connecting pieces of data on the Semantic Web (available in RDF) Linking Open Data Community Project - extends Web by publishing various open datasets as RDF on the Web and setting RDF links between data items from different data sources.
5 Adaptive Hypermedia One page fits all! No! Different users have different browsing needs. Page content and hyperlinks should be adapted accordingly. Adaptive Hypermedia is a solution. Most of apps are in educational hypermedia domain (AHA, InterBook, …) Early adaptive hypermedia systems use controlled vocabularies. Semantic Web is a solution.
6 Adaptive Hypermedia Standards IEEE PAPI and IMS LIP are well known user modelling standards. They mainly developed for learners in educational hypermedia. Ordinary users will not enter such information to a Web site or may not need that kind of personalization. How about browsing interests, goals, strategies of a user?
7 Personalization Mechanisms Existing approaches are obstructive – Users need to log in to multiple websites – Users have to enter personal information and preferences many times – Profiles are different for each site There is a need for generic user profiles and personalization architectures, which can achieve adaptive hypermedia on diverse websites
SemWeb: a Personalized Semantic Web Browser Şah, M., Hall, W., De Roure, D. C.: Designing a Personalized Semantic Web Browser. Accepted to Adaptive Hypermedia and Adaptive Web-Based Systems, 2008. Şah, M., Hall, W., De Roure, D. C.: SemWeB: A Semantic Web Browser for Supporting the Browsing of Users using Semantic and Adaptive Links. Accepted to Doctoral Consortium of Adaptive Hypermedia and Adaptive Web-Based Systems, 2008.
9 Background There are linked data browsers – Tabulator, Disco, OpenLink RDF Browser, …. – Separation between metadata and Web content – Our intention is not to create a linked data browser, but to create a semantic layer to a browser. Interfaces for supporting browsing of users – Magpie, COHSE (No Adaptive Hypermedia support and they use databases for linking) – Our aim is to adapt information to the needs of the users. Besides, we will use Web as source for linking (open-corpus linking).
13 Information Extraction and Semantic Annotation Information extraction using ontologies and ontology-driven lexicon based on the modified GATE framework. For demonstration ECS ontology is used. We crawl RDF files from ECS domain and created gazetteers and their URI mappings for semantic annotation.
14 Information Extraction (Cont.) Extend GATE with a mapping service, which matches named entity URIs to lexicons or lexicons to named entity URIs. – i.e. Wendy Hall lexicon is matched to http://id.ecs.soton.ac.uk/person/1650 Semantic annotations are created using JAPEC pattern matching rules. Annotation storage unit stores the created annotations as XML files at the server-side.
19 Semantic Hyperlink Creation Semantic Hyperlinks are requested asynchronously using AJAX request to server with resource URI, goal(s), title userid. Possible link anchors and targets can be found by analyzing RDF description of the resource. In addition, more useful information shown to the users according to their goals – For example, persons recent DBLP publications – Wikipedia definition and links to broader or special topics
20 Algorithm 1) Dereference URI x of resource 2) Match triple patterns (x any any) and (any any x) 3) Match triple patterns (x rdfs:seeAlso y) 4) Match triple patterns (x owl:sameAs y) 5) If the user has a goal, use Goal Services for finding related information 6) If the user is logged in annotate links with visual cues 6) Create a response XML file using a presentation vocabulary and return this to the clients browser
25 Personalization We want to personalize browsing of users using metadata obtained from Web page and user profile. First we need a generic user model which also expresses browsing needs of the users.
26 Browsing Browsing can be categorized in three groups ( Bawden, D.: Information Systems and the Stimulation of Creativity (1986) and Cove, J., Walsh, B.: Online Text Retrieval via Browsing (1988)): – Purposive / search browsing (looking for a definite piece of information) – directed – exploratory / general purpose browsing (deliberately searching for inspiration) – semi- directed – Capricious / serendipity browsing (randomly examining material) – undirected
27 Proposed User Model In the user model, at present we use seven categories: Identification, Preference, Security, Browsing Goal, Interest, Expertise and Browsing Behavior. Later the user model can be extended with more information (i.e. portfolio) from existing standards. In addition, our model can applied to existing standards.
29 A Part of Proposed User Model (Cont.) Interest: Low, Medium, High Expertise: Novice, Intermediate, Expertise Goal: Will be automatically provided by browser based on semantic context. Browsing Level: Inactive, Passive, Active, Very Active Browsing Type: Conditional – If the user has a browsing goal, then directed – If the user has a browsing interest, then semi-directed – If the user does not have a browsing goal or interest, then undirected
31 User Modeling To start personalization, users need to register and log in from their browsers. Profiles are kept at server-side triple store. Additionally, a profile editor is developed, where users can update profiles from their browsers. Users can be explicitly assigned to expertise, interests and goals from SemWeB. Browsing level and browsing type are automatically updated by SemWeB.
33 Adaptive Navigation Support and Adaptive Presentation Based on different browsing types – Directed browsing, show related links according to short-term browsing goals. – Semi-directed browsing or un-directed browsing, use interests to recommend links that are relevant to the users interests.
36 Creating Adaptive Links (Cont.) Adaptation based on expertise – When a link is requested by a novice user, provide links to Wikipedia pages. – When the user is an expert, provide detailed semantic links. Personalized Homepages Also, link sorting and link annotation can be done based on interest ratings, goal priorities, expertise and browsing levels.
42 Conclusions We presented a personalized Semantic Web browser architecture, which uses a novel behavior-based user model for adaptation. In our approach, AH and context-based linking can be achieved on different Web sites. SemWeB is not an application specific software and tested on ECS, DBpedia and DBLP linked data domains.