KEx:a Peer-to-Peer solution for Distributed Knowledge
Management
M.Bonifacio1,2,P.Bouquet1,2,G.Mameli2,and M.Nori2
1Dept.of Information and Communication Tech.–University of Trento(Italy)
2Istituto per la Ricerca Scientifica e Tecnologica,Trento(Italy) Abstract.Distributed Knowledge Management is an approach to Knowledge
Management bad on the principle that the multiplicity(and heterogeneity)of
perspectives within complex organizations should not be viewed as an obstacle
to knowledge exploitation,but rather as an opportunity that can foster innovation
and creativity.Despite a wide agreement on this principle,most current KM sys-
tems are bad on the idea that all perspectival aspects of knowledge should be
eliminated in favor of an objective and general reprentation of knowledge.In
this paper we propo a peer-to-peer architecture(called KEx),which embodies
the principle above in a quite straightforward way:(i)each peer(called a K-peer)
provides all the rvices needed to create and organize“local”knowledge from
an individual’s or a group’s perspective,and(ii)social structures and protocols
of meaning negotiation are defined to achieve mantic coordination among au-
tonomous ,when arching documents from other K-peers).
1Introduction
Distributed Knowledge Management(DKM),as described in[6],is an approach to KM bad on the principle that the multiplicity(and heterogeneity)of perspectives within complex organizations should not be viewed as an obstacle to knowledge exploitation, but rather as an opportunity that can foster innovation and creativity.
朱雀The fact that different individuals and communities may have very different per-spectives,and that th
e perspectives affect their reprentation of the world(and there-fore of their work)is widely discusd–and generally accepted–in theoretical rearch on the nature of knowledge.Knowledge reprentation in artificial intelligence and cog-nitive science have produced many theoretical and experimental evidences of the fact that what people know is not a mere collection of facts;indeed,knowledge always pre-suppos some(typically implicit)interpretation schema,which provide an esntial component in n-making(e,for example,the notions of context[18,7,13],mental space[12],partitioned reprentation[10]);studies on the social nature of knowledge stress the social nature of interpretation schemas,viewed as the outcome of a special kind of“agreement”within a community of knowing(e,for example,the notions of scientific paradigm[16],frame[15]),thought world[11],perspective[3]).
大鹏赋Despite this large convergence,it can be obrved that the high level architecture of most current KM systems in fact does not reflect this vision of knowledge(e[5,6,4] for a detailed discussion of this claim).The fact is that most KM systems embody the assumption that,to share and exploit knowledge,it is necessary to implement a process
2
of knowledge-extraction-and-refinement,who aim is to eliminate all subjective and contextual aspects of knowledge,and create an objective and general reprentation that can then be reud by other people in a variety of situations.Very often,this process isfinalized to build a central knowledge ba,where knowledge can be accesd via a knowledge portal.This centralized approach–and its underlying objectivist episte-mology–is one of the reasons why so many KM systems are derted by urs,who perceive such systems either as irrelevant or oppressive[9].
In this paper we propo a peer-to-peer(P2P)architecture,called KEx,which is coherent with the vision of DKM.Indeed,P2P systems em particularly suitable to implement the two core principles of DKM,namely the principle of autonomy(commu-nities of knowing should be granted the highest possible degree of mantic autonomy to manage their local knowledge),and the principle of coordination(the collaboration between autonomous communities must be achieved through a process of mantic co-ordination,rather than through a process of mantic homogenization)[6].In KEx,each community of knowing(or Knowledge Nodes(KN),as they are called in[4])is repre-nted by a peer,and the two principles above are implemented in a quite straightfor-ward way:(i)each peer provides all the rvices needed by a knowledge node to create and organize its own local knowledge(autonomy),and(ii)by defining social structures and protocols of meaning negotiation in order to achieve mantic , when arching documents from other peers).
The paper goes as follows.In ction2,we describe the main features of KEx,and argue why they provide a uful support to DKM;in3,we describe its implementation in a peer-to-peer platform called JXTA;finally,we draw some conclusions and future work.
2KEx:a P2P architecture for DKM
收据模板手写KEx is a P2P system which allows a collection of KNs to arch and provide documents on a mantic basis without presupposing a beforehand agreement on how documents should be categorized,or on a common language for reprenting mantic information within the system.In the following ctions,we describe the high-level architecture of KEx,and explain what role each element plays in a DKM vision.
2.1K-peers
KEx is defined as a collection of peers,called knowledge peers(K-Peers),each of which reprents a KN,namely an individual’s or a group’s perspective on a given body of knowledge.Each K-peer can play two main roles:provider and eker.A K-peer acts as a provider when it“publishes”in the system a body of knowledge,together with an explicit perspective on it(called a a topic hierarchy ud to categorized local documents[8]);a K-peer acts as a eker when it arches for information
by making explicit part of its own perspective,and negotiates it with other K-peers.
Each K-peer has the structure shown in Figure1.Below we illustrate the main modules and functionalities.
3
Fig.1.The KEx’s main components
Document Repository.A Document Repository is where each KN stores its own local knowledge.We can imagine a private space in which the KN maintains its document and data,possibly using a local mantic ,afile-system structure,or a databa schema),or a document management system in order to organize and access them.
Context Repository.Following[2],we define a context as a partial and approximate reprentation of the world from an individual’s or a group’s perspective.The reason why we adopt this notion of context is that it provides a robust formal framework(called Local Models Semantics[13])for modeling both contexts and their relationships.
In order to u contexts in KEx,we adopted a web-oriented syntax for contexts, called CTXML.It provides an XML-Schema specification of context for document organization and classification3.
In KEx,each context plays the role of a category system for organizing and classi-fying documents,or any other kind of digital information identifiable by a URI,stored in a document repository.Each peer can u more than one context to classify local knowledge;a K-peer’s contexts are stored in a context repository.
From the standpoint of DKM,contexts are relevant in two distinct ns:
3Currently,contexts are trees,who nodes are labelled with words defined in some name space.
Arcs are Is-A,Part-Of or generic relations between nodes.Details can be found in[8].
教室里的掌声作文4
–on the one hand,they have an important role within each KN,as they provide a dynamic and incremental explicitation of its mantic perspective.Once contexts are reified,they become cognitive artifacts that contribute to the process of perspec-tive making[3],namely the consolidation of a shared view in a KN,continuously subject to revision and internal negotiation among its members;
–on the other hand,contexts offer a simple and direct way for a KN to make public its perspective on the information that that KN can provide.Therefore,as we will e,contexts are an esntial tool for mantic coordination among different KN.
It is important to obrve that contexts provide only a common syntax for classifica-tion structures.Indeed,we could e them as a language for wrapping any classification ,like directory systems,databas schemas,web directories).This means that in principle people can conti
写家的作文
nue to work with their preferred document manage-ment system,provided it can be wrapped using CTXML.
Context management module.The context management module allows urs to cre-ate,manipulate,and u contexts in KEx.The module has two main components:
–Context editor:provides urs with a simple interface to create and edit contexts, and to classify information with respect to a context.This happens by allowing urs to create links from a resource(identified by a URI)to a node in a con-text.Examples of resources are:documents in local directories,the address of a databa access rvices,address of other K-peers that provide information that
a KN wants to explicitly classify in its own context.韩国历史
–Context browr:is part of Seeker component(GUI Ur in Figure1)and allows urs to navigate contexts in the context repository.The main reasons for navigat-ing a context in KEx are two.Thefirst is obviously tofind document in the local knowledge repository by navigating the mantic structure.The cond,and more important reason,is to build queries.The intuitive idea is that urs can make con-text dependent queries(namely,from their perspective)by lecting a category in one o
f the available contexts.Once a category is lected,the context browr builds
a focus4–namely a contextual interpretation of the ur’s query–by automatically
extracting the relevant portion of the context to which the category belongs.The focus is then ud as a basis for meaning coordination and negotiation with other K-peers during the arch.
2.2Roles of K-peers in KEx
Each K-peer can play two main roles:eker and provider.Their interactions are repre-nted in Figure2,and described in detail in the following two ctions.
体质健康Seeker As a eker,a K-peer allows urs to arch for documents(and other infor-mation)from other K-peers and federations(e Section2.3).The eker supports the 4See[17]for a formal definition of focus.
5
Fig.2.The KEx system:interaction between Seeker and Provider roles
ur in the definition of context-dependent queries through the context browr.A query is compod by a query expression and a focus.A query expression is a list(possibly empty)of one or more keywords provided by a ur;a focus is a portion of a context determined by the category that the ur has lected.Moreover,the eker provides the discovery mechanism,ud tofind resources to which the query is to be nt.The ur decides to nd the query to some of the available K-peers and federations.When the ur submits the query,the eker activates a ssion associated to that query(there can be only one active ssion for each eker).In a ssion,a eker can receive veral asynchronous replies from the providers which resolved the query(through the mean-ing negotiation protocol,e below)and called back the eker.The results returned to the ur are compod by the aggregation of all the results received from the providers; each result is made up of a list of document ,name of the document, short description,and so on).Each result is prented together with the part of context that the provider has matched against the current query.This relationship between con-texts can be ud as an opportunity for learning relationships across contexts of different KNs that the eker can store and reu for future queries(e ction2.3).Finally,if one or more interesting documents are found,the eker can contact the K-peers that have the documents and,if possible,download them.
关于古诗的作文
Provider.The provider is the cond main role in the KEx system.It contains the func-tionalities required to take and resolve a query,and to identify the results that must to
6
be returned to the eker.When a K-peer receives a context-dependent query(keywords +focus),it instantiates a provider(which is configured to u a t of contexts and to provide documents in a given portion of the knowledge repository),and tries to resolve the query in two ways:
–Semantic resolution:using a context matching algorithm[17],the provider arches for relations between the locally available contexts and the query’s focus.More specifically,the matching algorithm arches categories who associated contex-tual information in the providers contexts matches(in a n defined in[17])with the querys focus.If a match is found,the URIs of the resources associated to the provider’s context are returned to the eker,together with a short information on the reason why a mantic match was found.If the matched category contains also links to resources in other K-peers,the provider propagates the query to tho K-peers.
–Lexical resolution:using a keyword-bad indexer,the provider arches for the occurrence of specific keywords into the t of documents of the local repository.
If the query contains only keywords,the provider will u only the lexical arch; if it contains only a focus,the provider will u only the mantic arch;if both are available,the outcome will be the result of intercting the mantic and lexical results.
2.3K-Services
KEx provides a collection of rvices which have an important role in supporting knowledge exchange(that’s why they are called K-rvices).The main K-rvices are described in the following ctions.
Context normalization and enrichment.This rvice allows to perform a linguistic ,deleting stop words,tokenizing,part-of-speech tagging,etc.)on ur defined contexts,and to u knowledge from an external linguistic , WordNet)to add mantic information to the categories in a context.
Normalization us pretty standard NLP techniques,so we do not discuss it here.As to enrichment,it is applied offline to a context defined by a ur(e[17]for details). It takes a ur-defined ,a context built with the context editor)as input and returns a mantically enriched context as output.In our current implementation,the result is that linguistic ,ns,sy
nonyms,hierarchical relations with other categories,and so on)is extracted from WordNet and is“attached”to each context node label.
It is important to say why enrichment is not equivalent to introduce a shared(“ob-jective”)mantics in KEx.Indeed,the intuition is that the meaning of a label in each context node has two components:
–thefirst is the linguistic component,which means that the words ud as labels have a meaning(or,better,a t of meanings)in a“dictionary”.This is ud to record that many words have different ,“apple”as a fruit,“apple”as
a tree,and“apple”as a computer brand),even if only one of them is likely to be
the relevant one in a given context;
