Application Of Facet Classification To Knowledge Management

Since the amount of available information in today's society is growing enormously, it becomes fundamental to develop efficient and easy to use systems to access and navigate information.In the last decades, a lot of techniques are been developed in order to support searching and navigation processes.Among them, hypertext and hypermedia systems have gathered a discrete success. As known, they provide a way to explore texts and other kind of media in a not sequential order. In fact, the main characteristic of these systems is the capability of linking and browsing linked information. Unfortunately, the provided structure of links is usually static and typically reflects the creator mental model rather than user model. Database systems traditionally provide and alternative way to store and access data. A typical way to search information here is by using sophisticated queries and keyword search. Another time, they do not provide an easy and transparent access to information simply because system interface is usually too far from user language and level of expertise. For these and other reasons, many researchers claim that these techniques, usually referred as classical IR techniques, are becoming inadequate. In response to these needs, in the last years classifications and in particular faceted systems are becoming very popular. They tend to naturally1 group objects with respect to their distinctive properties, i.e. their attributes, categorizing them in different hierarchies; each hierarchy corresponds to a different facet.In faceted systems, all objects are grouped in categories, the facets, according to their distinctive properties. For this reason facets describe the set of objects from orthogonal different perspectives. Instead of forcing users following a unique static classification, they can orthogonally browse through facets and combine them in different ways. Subsequently, among systems that use facets for improving searching, we distinguish between homogeneous faceted systems and heterogeneous faceted systems: Homogeneous faceted systems(systems that manage objects having the same attributes); Heterogeneous faceted systems(systems that manage objects that can have different attributes). Objects managed by our system are assumed to be heterogeneous and indexed in classifications.SWEB is a system of knowledge management which bases on semantic web, and being developed by knowdive group, University of Trento. The goal of SWEB is to provide a set of technological tools to support the user in the creation, acquisition, adaptation, evolution, management, and sharing of knowledge and data within the global infrastructure of the emerging Web 2.0. At the moment these are the main sub-systems which are part of the KnowDive project: ClassDB builds up on top of the simple but commonly used data structures for encoding knowledge– classification hierarchies; DocDB is a hybrid document management system which combines the benefits of three main approaches to information access: classification schemes, faceted classifications, and direct search. SMatch resolves the problem of heterogeneity among (formal) classifications which inevitably arises when classifications are designed independently by different users. SP2P develops a robust Peer-to-Peer technological infrastructure with the goal to support completely decentralized knowledge discovery and sharing, transient user participation and community building, as well as knowledge adaptation through semantic convergence of peer vocabularies.The more intuitive way to represent a facet hierarchy is through a rooted tree where the root label is the name of the facet and internal and leaf nodes'labels represent single allowed values. Objects are represented in the system by meta-documents. A meta-document is an internal representation that describes a document, or generically an object. And then we give two fundamental principles by which the facet based get-specific algorithm is motivated:"context"of a node and the"get-specific"rule, then we do classify operations according to these two principles. After this, we also list some principles for all implementation choices and discuss the main consequences of the implementation principles: automatical modality and interactive modality. At last we analyse the operations on meta-documnet and discuss some cases to get a clear idea about these operations.Future work is as follows:1. Semantic CRUD operationsWe have also to define the behavior of the system with respect to the CRUD operations on nodes and documents. These operations include for example"Cut & Paste"of nodes/documents, modifying node labels, modifying document attributes, issues on adopted children.2. Personalization mode:Operations related to the"personalization mode"should allow the user to personalize theclassification in order to obtain a"view"of the original one.3. In future versions facets could be inferred directly from labels by processing Natural Language.