Learning probabilistic models of link structure

Learning Probabilistic Models of Link Structure

Lise Getoor GETOOR@https://www.wendangku.net/doc/918971742.html, Computer Science Dept.

University of Maryland

College Park,MD20742

Nir Friedman NIR@CS.HUJI.AC.IL School of Computer Sci.&Eng.

Hebrew University

Jerusalem,91904,Israel

Daphne Koller KOLLER@https://www.wendangku.net/doc/918971742.html, Computer Science Dept.

Stanford University

Stanford,CA94305

Ben Taskar BTASKAR@https://www.wendangku.net/doc/918971742.html, Computer Science Dept.

Stanford University

Stanford,CA94305

Abstract

Most real-world data is heterogeneous and richly interconnected.Examples include the Web,hypertext, bibliometric data and social networks.In contrast,most statistical learning methods work with“?at”data representations,forcing us to convert our data into a form that loses much of the link structure.The re-cently introduced framework of probabilistic relational models(PRMs)embraces the object-relational nature of structured data by capturing probabilistic interactions between attributes of related entities.In this paper, we extend this framework by modeling interactions between the attributes and the link structure itself.An advantage of our approach is a uni?ed generative model for both content and relational structure.We propose two mechanisms for representing a probabilistic distribution over link structures:reference uncertainty and existence uncertainty.We describe the appropriate conditions for using each model and present learning algo-rithms for each.We present experimental results showing that the learned models can be used to predict link structure and,moreover,the observed link structure can be used to provide better predictions for the attributes in the model.

1.Introduction

In recent years,we have witnessed an explosion in the amount of information that is available to us in digital form.More and more data is being stored,and more and more data is being made accessible,through traditional interfaces such as corporate databases and,of course,via the Internet and the World Wide Web.There is much to be gained by applying machine learning techniques to these data,in order to extract useful information and patterns.

Most often,the objects in these data do not exist in isolation—there are“links”or relationships that hold between them.For example,there are links from one web page to another,a scienti?c

1

paper cites another paper,and an actor is linked to a movie by the appearance relationship.Most work in machine learning,however,has focused on“?at”data,where the instances are independent and identically distributed.The main exception to this rule has been the work on inductive logic programming(Muggleton,1992,Lavra?c and D?z eroski,1994).This work focuses on the problem of inferring link structure from a logical perspective.

More recently,there has been a growing interest in combining the statistical approaches that have been so successful when learning from a collection of homogeneous independent instances(typi-cally represented as a single table in a relational database)with relational learning methods.Slattery and Craven(1998)were the?rst to consider combining a statical approach with a?rst-order rela-tional learner(FOIL)for the task of web page classi?cation.Chakrabarti et al.(1998)explored methods for hypertext classi?cations which used both the content of the current page and informa-tion from related pages.Popescul et al.(2002)use an approach that uses a relational learner to guide in the construction of features to be used by a(statistical)propositional learner.Yang et al.(2002) identify certain categories of relational regularities and explore the conditions under which they can be exploited to improve classi?cation accuracy.

Here,we propose a uni?ed statistical framework for content and links.Our frameworks builds on the recent work on probabilistic relational models(PRMs)(Poole,1993,Ngo and Haddawy,1995, Koller and Pfeffer,1998).PRMs extend the standard attribute-based Bayesian network representa-tion to incorporate a much richer relational structure.These models allow properties of an entity to depend probabilistically on properties of other related entities.The model represents a generic de-pendence for a class of objects,which is then instantiated for particular sets of entities and relations between them.Friedman et al.(1999)adapt the machinery for learning Bayesian networks from a set of unrelated homogeneous instances to the task of learning PRMs from structured relational data.

The original PRM framework focused on modeling the distribution over the attributes of the ob-jects in the model.It took the relational structure itself—the relational links between entities—to be background knowledge,determined outside the probabilistic model.This assumption implies that the model cannot be used to predict the relational structure itself.A more subtle yet very im-portant point is that the relational structure is informative in and of itself.For example,the links from and to a web page are very informative about the type of web page(Craven et al.,1998),and the citation links between papers are very informative about the paper topics(Cohn and Hofmann, 2001).

The PRM framework can be naturally extended to address this limitation.By making links?rst-class citizens in the model,the PRM language easily allows us to place a probabilistic model directly over them.In other words,we can extend our framework to de?ne probability distributions over the presence of relational links between objects in our model.The concept of a probabilistic model over relational structure was introduced by Koller and Pfeffer(1998)under the name structural uncertainty.They de?ned several variants of structural uncertainty,and presented algorithms for doing probabilistic inference in models involving structural uncertainty.

In this paper,we show how a probabilistic model of relational structure can be learned directly from data.Speci?cally,we provide two simple probabilistic models of link structure:The?rst is an extension of the reference uncertainty model of Koller and Pfeffer(1998),which makes it suitable for a learning framework;the second is a new type of structural uncertainty,called existence uncertainty.We present a clear semantics for these extensions,and propose a method for learning

2

such models from a relational database.We present empirical results on real-world data,showing that these models can be used to predict the link structure,as well as use the presence of(observed) links in the model to provide better predictions about attribute values.Interestingly,these bene?ts are obtained even with the very simple models of link structure that we use in this paper.Thus,even simplistic models of link uncertainty provide us with increased predictive accuracy.

2.Probabilistic Relational Models

A probabilistic relational model(PRM)speci?es a template for a probability distribution over a database.The template describes the relational schema for the domain,and the probabilistic depen-dencies between attributes in the domain.A PRM,together with a particular database of objects and relations,de?nes a probability distribution over the attributes of the objects and the relations.

2.1Relational Schema

A schema for a relational model describes a set of classes,.Each class is associated with a set of descriptive attributes and a set of reference slots.1The set of descriptive attributes of a class is denoted.Attribute of class is denoted,and its domain of values is denoted.For example,the class might have the descriptive attributes Gender,with domain male,female.For simplicity,we assume in this paper that attribute domains are?nite;this is not a fundamental limitation of our approach.

The set of reference slots of a class is denoted.We use to denote the reference slot of.Each reference slot is typed:the domain type of and the range type ,where is some class in.A slot denotes a function from to

.For example,we might have a class with the reference slots Actor,whose range is the class,and Movie,whose range is the class.We note that the functional nature of slots does not prevent us from having many-to-many relations between classes.We simply use a standard transformation where we introduce a class corresponding to the relationship object. This class will have an object for every pair(or tuple)of related objects,with functional reference slots to the objects it relates.The Role class above is an example of this transformation.

It is useful to distinguish between an entity and a relationship,as in entity-relationship diagrams. In our language,classes are used to represent both entities and relationships.Thus,a relationship such as Role,which relates actors to movies,is also represented as a class,with reference slots to the class Actor and the class Movie.We use to denote the set of classes that represent entities, and to denote those that represent relationships.We use the generic term object to refer both to entities and to relationships.

The semantics of this language is straightforward.An complete instantiation speci?es the set of objects in each class,and the values for each attribute and each reference slot of each object. Thus,a complete instantiation is a set of objects with no missing values and no dangling refer-ences.It describes the set of objects,the relationships that hold between the objects and all the values of the attributes of the objects.For example,Figure1shows an instantiation of our simple movie schema.It speci?es a particular set of actors,movies and roles,along with values for each of their attributes and references.

1.There is a direct mapping between our notion of class and the tables in a relational database:descriptive attributes

correspond to standard table attributes,and reference slots correspond to foreign keys(key attributes of another table).

3

ACTOR

name gender fred male ginger female bing male

MOVIE name genre m1drama m2comedy

ROLE

role movie actor role-type

r1m1fred hero

r2m1ginger heroine

r3m1bing villain

r4m2bing hero

r5m2ginger love-interest

Figure1:An instantiation of the relational schema for a simple movie domain.

As discussed in the introduction,our goal in this paper is to construct probabilistic models over instantiations.To do so,we need to provide enough background knowledge to circumscribe the set of possible instantiations.Friedman et al.(1999)assume that the entire relational structure is given as background knowledge.More precisely,they assume that they are given a relational skeleton, ,which speci?es the set of objects in all classes,as well as all the relationships that hold between them;in other words,it speci?es the values for all of the reference slots.In our simple movie example,the relational skeleton would contain all of the information except for the gender of the actors,the genre of the movies,and the nature of the role.

2.2Probabilistic Model for Attributes

A probabilistic relational model speci?es a probability distribution over a set of instantiations of the relational schema.More precisely,given a relational skeleton,it speci?es a distribution over all complete instantiations that extend the skeleton.

A PRM consists of a qualitative dependency structure,,and the parameters associated with it, .The dependency structure is de?ned by associating with each attribute a set of formal parents Pa.These correspond to formal parents;they will be instantiated in different ways for different objects in.Intuitively,the parents are attributes that are“direct in?uences”on. The attribute can depend on another probabilistic attribute of.It can also depend on attributes of related objects.2

The quantitative part of the PRM speci?es the parameterization of the model.Given a set of parents for an attribute,we can de?ne a local probability model by associating with it a conditional probability distribution(CPD).For each attribute we have a CPD that speci?es Pa. Each CPD in our PRM is legal,i.e.,the entries are positive and sum to1.

De?nition1A probabilistic relational model(PRM)for a relational schema de?nes for each class and each descriptive attribute,a set of formal parents Pa,and a conditional probability distribution(CPD)that represents Pa.

2.PRMs allow a much richer dependency model,where objects can depend on each other via longer slot chains.They

also allow dependencies via inverse slots,which generally de?ne one-to-many relations(e.g.,the set of all roles of an actor).There is no dif?culty adding these features to our language,but they complicate the presentation considerably;

we have therefore chosen to omit them for simplicity.

4

Given a relational skeleton,a PRM speci?es a distribution over a set of instantiations

consistent with.This speci?cation is done by mapping the dependencies in the class-level PRM

to the actual objects in the domain.For a class,we use to denote the objects,as speci?ed by the relational skeleton.(In general we will use the notation to refer to the

set objects of each class as de?ned by any type of domain skeleton.)Let be an attribute in

the schema,and let be some object in.Recall that the PRM allows two types of formal parents:and.For a formal parent of the form,the corresponding actual parent of is.For a formal parent of the form,the corresponding formal parent of is, where in.Thus,the class-level dependencies in the PRM are instantiated according to the relational skeleton,to de?ne object-level dependencies.The parameters speci?ed by the PRM are used for each object in the skeleton,in the obvious way.

Thus,for a given skeleton,the PRM basically de?nes a ground Bayesian network.The qualitative

structure of the network is de?ned via an instance dependency graph,whose nodes correspond

to descriptive attributes of entities in the skeleton.These are the random variables in our model.

We have a directed edge from to if is an actual parent of,as de?ned above.The

quantitative parameters of the network are de?ned by the CPDs in the PRM,with the same CPD

used multiple times in the network.This ground Bayesian network leads to the following chain rule

which de?nes a distribution over the instantiations compatible with our particular skeleton:

Pa(1)

For this de?nition to specify a coherent probability distribution over instantiations,we must ensure that our probabilistic dependencies are acyclic.In particular,we must verify that each random variable does not depend,directly or indirectly,on its own value.In other words,must be acyclic.We say that a dependency structure is acyclic relative to a relational skeleton if the directed graph is acyclic.

Theorem2(Friedman et al.,1999)Let be a PRM with an acyclic instance dependency graph, and let be a relational skeleton.Then,Eq.(1)de?nes a coherent distributions over instances that extend.

The de?nition of the instance dependency graph is speci?c to the particular skeleton at hand:the existence of an edge from to depends on whether,which in turn depends on the interpretation of the reference slots.Thus,it allows us to determine the coherence of a PRM only relative to a particular relational skeleton.When we are evaluating different possible PRMs as part of our learning algorithm,we want to ensure that the dependency structure we choose results in coherent probability models for any skeleton.

For this purpose,we use a class dependency graph,which describes all possible dependencies among attributes.In this graph,we have an(intra-object)edge if is a parent of .If is a parent of,and,we have an(inter-object)edge.

Theorem3(Friedman et al.,1999)If the class dependency graph of a PRM is acyclic,then the instance dependency graph for any relational skeleton is also acyclic.Hence,de?nes a legal model for any relational skeleton.

5

Figure2:Reference uncertainty in a simple citation domain.

In the model described here,all relations between attributes are determined by the relational skele-ton;only the descriptive attributes are uncertain.Thus,Eq.(1)determines the probabilistic model of the attributes of objects,but does not provide a model for the relations between objects.In the subsequent sections,we extend our framework to deal with link uncertainty,by extending our prob-abilistic model to include a distribution over the relational structure itself.

3.Reference Uncertainty

In the previous section,we assumed that the relational structure was not a part of our probabilistic model,that it was given as background knowledge in the relational skeleton.But by incorporating the links into the probabilistic model,we can both predict links and more importantly use the links to help us make predictions about other attributes in the model.

Consider a simple citation domain illustrated in Figure2.Here we have a document collection. Each document has a bibliography that references some of the other documents in the collection. We may know the number of citations made by each document(i.e.,it is outside the probabilistic model).By observing the citations that are made,we can use the links to reach conclusions about other attributes in the model.For example,by observing the number of citations to papers of various topics,we may be able to infer something about the topic of the citing paper.

Figure3(a)shows a simple schema for this domain.We have two classes,Paper and Cites.The Paper class has information about the topic of the paper and the words contained in the paper. For now,we simply have an attribute for each word that is true if the word occurs in the page and false otherwise.The Cites class represents the citation of one paper,the Cited paper,by another paper,the Citing paper.(In the?gure,for readability,we show the Paper class twice.)In this model,we assume that the set of objects is pre-speci?ed,but relations among them,i.e.,reference slots,are subject to probabilistic choices.Thus,rather than being given a full relational skeleton, we assume that we are given an object skeleton.The object skeleton speci?es only the objects in each class,but not the values of the reference slots.In our example,the object skeleton speci?es the objects in class Paper and the objects in class Cites,but the reference slots of the Cites relation,Cites.Cited and Cites.Citing are unspeci?ed.In other words,the probabilistic model does not provide a model of the total number of citation links,but only a distribution over their“endpoints”.Figure3shows an object skeleton for the citation domain.

6

(a)(b)

Figure3:(a)A relational schema for the citation domain.(b)An object skeleton for the citation domain.

3.1Probabilistic model

In the case of reference uncertainty,we must specify a probabilistic model for the value of the reference slots.The domain of a reference slot is the set of keys(unique identi?ers)of the objects in the class to which refers.Thus,we need to specify a probability distribution over the set of all objects in.For example for Cited,we must specify a distribution over the objects in class Paper.

A naive approach is to simply have the PRM specify a probability distribution directly over the objects in.For example for Cited,we would have to specify a distribution over the primary keys of Paper.This approach has two major?aws.Most obviously,this distribution would require a parameter for each object in,leading to a very large number of parameters.This is a problem both from a computational perspective—the model becomes very large,and from a statistical perspective—we often would not have enough data to make robust estimates for the parameters.More importantly,we want our dependency model to be general enough to apply over all possible object skeletons;a distribution de?ned in terms of the objects within a speci?c object skeleton would not apply to others.

In order to achieve a general and compact representation,we use the attributes of to de?ne the probability distribution.In this model,we partition the class into subsets labeled ac-cording to the values of some of its attributes,and specify a probability for choosing each partition, i.e.,a distribution over the partitions.We then select an object within that partition uniformly.

For example,consider a description of movie theater showings as in Figure4(a).For the foreign key Shows.Movie,we can partition the class Movie by Genre,indicating that a movie theater?rst selects the genre of movie it wants to show,and then selects uniformly among the movies with the selected genre.For example,a movie theater may be much more likely to show a movie which is a thriller in comparison to a foreign movie.Having selected,for example,to show a thriller,the theater then selects the actual movie to show uniformly from within the set of thrillers.In addition, just as in the case of descriptive attributes,the partition choice can depend on other attributes in our model.Thus,the selector attribute can have parents.As illustrated in the?gure,the choice of movie genre might depend on the type of theater.Consider another example,in our citation domain. As shown in Figure4(b),we can partition the class Paper by Topic,indicating that the topic of a

7

(a)(b)

Figure 4:(a)An example of reference uncertainty for a movie theater’s showings.(b)A simple

example of reference uncertainty in the citation domain

citing paper determines the topics of the papers it cites;and then the cited paper is chosen uniformly among the papers with the selected topic.

We make this intuition precise by de?ning,for each slot ,a partition function .We place several restrictions on the partition function which are captured in the following de?nition:

De?nition 4Let

be a reference slot with domain .Let be a function where

is a ?nite set of labels.We say that is a partition function for if there is a subset of the attributes of ,

,such that for any and any ,if the values of the attributes

of and are the same,i.e.,for each ,,then .We refer to as the partition attributes for .

Thus,the values of the partition attributes are all that is required to determine the partition to which an object belongs.

In our ?rst example,Movie Foreign Thriller and the partition attribute is Movie Genre .In the second example,Cited AI Theory and the partition attribute is

Cited Topic .There are a number of natural methods for specifying the partition function.It can be de?ned simply by having one partition for each possible combination of values of the partition attributes,i.e.,one partition for each value in the cross product of the partition attribute values.Our examples above take this approach.In both cases,there is only a single partition attribute,so specifying the partition function in this manner is not too unwieldy,but for larger collections of partition attributes or for partition attributes with large domains,this method for de?ning the partitioning function may be problematic.A more ?exible and scalable approach is to de?ne the partition function using a decision tree built over the partition attributes.In this case,there is one partition for each of the leaves in the decision tree.

Each possible value determines a subset of from which the value of (the referent)will be

selected.For a particular instantiation of the database,we use

to represent the set of objects in

that fall into the partition .8

We now represent a probabilistic model over the values of by specifying a distribution over

possible partitions,which encodes how likely the reference value of is to fall into one partition

versus another.We formalize our intuition above by introducing a selector attribute,whose

domain is.The speci?cation of the probabilistic model for the selector attribute is the same as that of any other attribute:it has a set of parents and a CPD.In our earlier example,the

CPD of Movie might have as a parent Type.For each instantiation of the parents,we

have a distribution over.3The choice of value for determines the partition from which the reference value of is chosen;the choice of reference value for is uniformly distributed within this set.

De?nition5A probabilistic relational model with reference uncertainty has the same compo-nents as in De?nition1.In addition,for each reference slot with,we have:

a partition function with a set of partition attributes;

a new selector attribute within which takes on values in the range of;

a set of parents and a CPD for.

To de?ne the semantics of this extension,we must de?ne the probability of reference slots as well as descriptive attributes:

Pa

Pa

(2)

where refers to—the partition that the partition function assigns.Note that the last term in Eq.(2)depends on in three ways:the interpretation of,the values of the attributes within the object,and the size of.

There is one small problem with this de?nition:the probability is not well-de?ned if there are no objects in a partition,i.e.,.In this case,we will perform a renormalization of the distribution for the selector attribute,removing all probability mass from the empty partition. We then treat this term,Pa,as in the above product.In other words,an instantiation where and is an empty partition necessarily has probability zero.4

3.In the current work,we treat this distribution as a simple multinomial distribution over this value space.In gen-

eral,however,we can represent such a distribution more compactly,e.g.,using a Bayesian network.For example, the genre of movies shown by a movie theater might depend on its type(as above).However,the language of the movie can depend on the location of the theater.Thus,the partition will be de?ned by Movie Genre Language,and its parents would be Type and Location.We can rep-resent this conditional distribution more compactly by introducing a separate variable Genre,with a parent Type,and another Language,with a parent Location.

4.While an important technical detail,in practice,for the majority of our experimental domains,we did not encounter

empty partitions.

9

3.2Coherence of the Probabilistic Model

As in the case of PRMs with attribute uncertainty,we must be careful to guarantee that our proba-bility distribution is in fact coherent.In this case,the object skeleton does not specify which objects are related to which,and therefore the mapping of formal to actual parents depends on probabilistic choices made in the model.The associated ground Bayesian network will therefore be cumbersome and not particularly intuitive.We de?ne our coherence constraints using an instance dependency graph,relative to our PRM and object skeleton.

De?nition6The instance dependency graph for a PRM and an object skeleton is a graph with the nodes and edges described below.For each class and each,we have the following nodes:

a node for every descriptive attribute;

a node and a node,for every reference slot

The dependency graph contains?ve types of edges:

Type I edges:Consider any attribute(descriptive or selector)and formal parent.

We de?ne an edge,for every.

Type II edges:Consider any attribute(descriptive or selector)and formal parent

where.We de?ne an edge,for every and.

Type III edges:Consider any attribute and formal parent,where, and.We de?ne an edge,for every.In addition,for ,we add an edge for for every and for every.

Type IV edges:Consider any slot and partition attribute for.

We de?ne an edge,for every and.

Type V edges:Consider any slot.We de?ne an edge,for every.

We say that a dependency structure is acyclic relative to an object skeleton if the directed graph is acyclic.

Intuitively,type I edges correspond to intra-object dependencies and type II edges to inter-object dependencies.These are the same edges that we had in the dependency graph for regular PRMs, except that they also apply to selector attributes.Moreover,there is an important difference in our treatment of type II edges.In this case,the skeleton does not specify the value of,and hence we cannot determine from the skeleton on which object the attribute actually depends.Therefore, our instance dependency graph must include an edge from every attribute.

Type III edges represent the fact that the actual choice of parent for depends on the value of the slots used to de?ne it.When the parent is de?ned via a slot-chain,the actual choice depends on the values of all the slots along the chain.Since we cannot determine the particular object from the skeleton,we must include an edge from every slot potentially included in the chain.

Type V edges represent the dependency of a slot on the attributes de?ning the associated partition. To see why this dependence is required,we observe that our choice of reference value for depends on the values of the partition attributes of all of the different objects in.Thus,

10

these attributes must be determined before is determined.Finally,type V edges represent the fact that the actual choice of parent for depends on the value of the selector attributes for the slots used to de?ne it.In our example,as Movie Genre,the genres of all movies must be determined before we can select the value of the reference slot Movie.

Based on this de?nition,we can specify conditions under which Eq.(2)speci?es a coherent prob-ability distribution.

Theorem7Let be a PRM with reference uncertainty whose dependency structure is acyclic relative to an object skeleton.Then and de?ne a coherent probability distribution over instantiations that extend via Eq.(2).

Proof:The probability of an instantiation is the joint distribution over a set of random variables de?ned via the object skeleton.We have two types of variables:

We have one random variable for each and each.Note that this also includes variables of the form that correspond to selector variables.

We have one random variable for each reference slot and.This variable denotes the actual object in that the slot points to.

Let be the entire set of variables de?ned by the object skeleton.Clearly,there is a one-to-one mapping between joint assignments to these variables and instantiations of the PRM that are consistent with.Because the instance dependency graph is acyclic,we can assume without loss of generality that the ordering is a topological sort of the instance dependency graph;thus,if,then all ancestors of in the instance dependency graph appear before it in the ordering.

Our proof will use the following argument.Assume that we can construct a non-negative function which has the form of a conditional distribution,i.e.,for any assignment to,we have that

(3) Then,by the chain rule for probabilities,we can de?ne

If satis?es Eq.(3),then is a well-de?ned joint distribution.

All that remains is to de?ne the function in a way that it satis?es Eq.(3).Speci?cally,the function will be de?ned via Eq.(2).We consider the two types of variables in our distribution. Suppose that is of the form,and consider any parent of.There are two cases: If the parent is of the form,then by the existence of type I edges,we have that

precedes in.Hence,the variable precedes.

11

If the parent is of the form,then by the existence of type III edges,for each along precedes in and hence,by the existence of type V edges all of the also precede.Furthermore,by the existence of type II edges,for any,we have that precedes.

In both cases,we can de?ne

Pa

as in Eq.(2).As the right-hand-side is simply a CPD in the PRM,it speci?es a well-de?ned condi-tional distribution,as required.

Now,suppose that is of the form.In this case,the conditional probability depends on, and the value of the partition attributes of all objects in.By the existence of type V edges,precedes.Furthermore,by the existence of type IV edges,we have that for every and.Consequently,the assignment to determines the number of objects in each partition of values of and hence the set for every. Finally,we set

if

otherwise

which is a well de?ned distribution on the objects in.

This theorem is limited in that it is very speci?c to the constraints of a given object skeleton.As in the case of PRMs without relational uncertainty,we want to learn a model in one setting,and be assured that it will be acyclic for any skeleton we might encounter.We accomplish this goal by extending our de?nition of class dependency graph.We do so by extending the class dependency graph to contain edges that correspond to the edges we de?ned in the instance dependency graph. De?nition8The class dependency graph for a PRM with reference uncertainty has a node for each descriptive or selector attribute and each reference slot,and the following edges: Type I edges:For any attribute and formal parent,we have an edge.

Type II edges:For any attribute and formal parent where,we have an edge.

Type III edges:For any attribute and formal parent,where,and ,we de?ne an edge.In addition,each,we add an edge

.

Type IV edges:For any slot and partition attribute for,we have an edge.

Type V edges:For any slot,we have an edge.

Figure5shows the class dependency graph for our extended movie example.

It is now easy to show that if this class dependency graph is acyclic,then the instance dependency graph is acyclic.

12

(a)(b)

Figure5:(a)A PRM for the movie theater example.The partition attributes are indicated using dashed lines.(b)The dependency graph for the movie theater example.The different

edge types are labeled.

Lemma9If the class dependency graph is acyclic for a PRM with reference uncertainty,then for any object skeleton,the instance dependency graph is acyclic.

Proof:Assume by contradiction that there is a cycle

Then,because each of these object edges corresponds to an edge in the class dependency graph,we have the following cycle in the class dependency graph:

This contradicts our hypothesis that the class dependency graph is acyclic.

The following corollary follows immediately:

Corollary10Let be a PRM with reference uncertainty whose class dependency structure is acyclic.For for any object skeleton,and de?ne a coherent probability distribution over instantiations that extend via Eq.(2).

4.Existence Uncertainty

The second form of structural uncertainty we introduce is called existence uncertainty.In this case, we make no assumptions about the number of links that exist.The number of links that exist and the identity of the links are all part of the probabilistic model and can be used to make inferences about other attributes in our model.In our citation example above,we might assume that the set of papers is part of our background knowledge,but we want to provide an explicit model for the presence or absence of citations.Unlike the reference uncertainty model of the previous section,we do not assume that the total number of citations is?xed,but rather that each potential citation can be present or absent.

13

(a)(b)

Figure6:(a)An entity skeleton for the citation domain.(b)A CPD for the Exists attribute of Cites.

4.1Semantics of relational model

The object skeleton used for reference uncertainty assumes that the number of objects in each rela-tion is known.Thus,if we consider a division of objects into entities and relations,the number of objects in classes of both types are?xed.Existence uncertainty assumes even less background infor-mation than speci?ed by the object skeleton.Speci?cally,we assume that the number of relationship objects is not?xed in advance.

We assume that we are given only an entity skeleton,which speci?es the set of objects in our domain only for the entity classes.Figure6(a)shows an entity skeleton for the citation example.Our basic approach is to allow other objects within the model—those in the relationship classes—to be undetermined,i.e.,their existence can be uncertain.In other words,we introduce into the model all of the objects that can potentially exist in it;with each of them,we associate a special binary variable that tells us whether the object actually exists or not.We call entity classes determined and relationship classes undetermined.

To specify the set of potential objects,we note that relationship classes typically represent many-many relationships;they have at least two reference slots,which refer to determined classes.For example,our Cite class has the two reference slots Citing and Cited.Thus the potential domain of the Cites class in a given instantiation is.Each“potential”object in this class has to form.Each such object is associated with a binary attribute that speci?es whether paper did or did not cite paper.

De?nition11Consider a schema with determined and undetermined classes,and let be an entity skeleton over this schema.We de?ne the induced relational skeleton,,to be the relational skeleton that contains the following objects:

If is a determined class,then.

Let be an undetermined class with reference slots whose range types are

respectively.Then contains an object for all tuples

.

14

The relations in are de?ned in the obvious way:Slots of objects of determined classes are taken from the entity skeleton.Slots of objects of undetermined classes are induced from the object de?nition:is.

To ensure that the semantics of schemas with undetermined classes is well-de?ned,we need a few tools.Speci?cally,we need to ensure that the set of potential objects is well de?ned and?nite.It is clear that if we allow cyclic references(e.g.,an undetermined class with a reference to itself),then the set of potential objects is not?nite.To avoid such situations,we need to put some requirements on the schema.

De?nition12A set of classes is strati?ed if there exists a partial ordering over the classes such that for any reference slot with range type,.

Lemma13If the set of undetermined classes in a schema is strati?ed,then given any entity skeleton the number of potential objects in any undetermined class is?nite.

Proof:We prove this by induction on the strati?cation level of the class.Let be the set of reference slots of,and let.Then If is the?rst level in the strati?cation,it cannot refer to any undetermined classes.Thus,the number of potential objects in is simply the product of the number of objects in each determined class as speci?ed in the entity skeleton.Each term is?nite,so the product of the terms will be?nite. Next,assume by induction that all of the classes at strati?cation levels less than are?nite.If is at level in the strati?cation,the constraints imposed by the strati?cation constraints imply that the range type of all of its reference slots are at strati?cation levels.By our induction hypothesis, each of these classes is?nite.Hence,the cross product of the classes of the reference slots is?nite and the number of potential objects in is?nite.

As discussed,each undetermined has a special existence attribute whose values are

true false.For uniformity of notation,we introduce an attribute for all classes;for classes that are determined,the value is de?ned to be always true.We require that all of the reference slots of a determined class have a range type which is also a determined class.

For a PRM with strati?ed undetermined classes,we de?ne an instantiation to be an assignment of values to the attributes,including the Exists attribute,of all potential objects.

4.2Probabilistic model

We now specify the probabilistic model de?ned by the PRM.By treating the Exists attributes as standard descriptive attributes,we can essentially build our de?nition directly on top of the de?nition of standard PRMs.

Speci?cally,the existence attribute for an undetermined class is treated in the same way as a descriptive attribute in our dependency model,in that it can have parents and children,and has an associated CPD.Figure6(b)illustrates a CPD for the Cites.Exists attribute.In this example,the existence of a citation depends on the topic of the citing paper and the topic of the cited paper;e.g., it is more likely that citations will exist between papers with the same topic.5.

5.This is similar to the‘encyclopedic’links discussed by(Ghani et al.,2001).Our exists models can capture each of

the types of links(encyclopedic,co-referencing,and partial co-referencing)de?ned in(Ghani et al.,2001).Moreover our exists models are much more general,and can capture much richer patterns in the existence of links.

15

Using the induced relational skeleton,treating the existence events as descriptive attributes,we have set things up so that Eq.(1)applies with minor changes.There are two minor changes to the de?nition of the distribution:

We want to enforce that false if false for one of the slots of.Suppose that has the slots,we de?ne the effective CPD for as follows.Let Pa

Pa,and de?ne

Pa Pa if true

otherwise

We want to“decouple”the attributes of non-existent objects from the rest of the PRM.Thus, if is a descriptive attribute,we de?ne Pa Pa,and

Pa Pa if true

otherwise

It is easy to verify that in both cases Pa is a legal conditional distribution.

In effect,these constraints specify a new PRM,in which we treat as a standard descriptive attribute.For each attribute(including the Exists attribute),we de?ne the parents of in to be Pa and the associated CPD to be Pa.

Given an entity skeleton,a PRM with exists uncertainty speci?es a distribution over a set of instantiations consistent with:

Pa(4) We can similarly de?ne the instance dependency graph and the class dependency graph for a PRM with existence uncertainty using the corresponding notions for the standard PRM.As there, we require that the class dependency graph is acyclic.One immediate consequence of this requirement is that the schema is strati?ed.

Lemma14If the class dependency graph is acyclic,then there is a strati?cation of the unde-termined classes.

Proof:The strati?cation is given by an ordering of Exists attributes consistent with the class de-pendency graph.Because the class dependency graph has an edge from to for every slot whose range type is,will precede in the constructed ordering.Hence it is a strati?cation ordering.

Furthermore,based on this de?nition,we can now easily prove the following result:

Theorem15Let be a PRM with existence uncertainty and an acyclic class dependency graph. Let be an entity skeleton.Then Eq.(4)de?nes a coherent distribution on all instantiations of the induced relational skeleton.

16

Proof By Lemma14and Lemma13we have that is a well-de?ned relational https://www.wendangku.net/doc/918971742.html,ing the assumption that is acyclic,we can apply Theorem2to and conclude that de?nes a coherent distribution over instances to,and hence so does.

One potential shortcoming of our semantics is that it de?nes probabilities over instances that in-clude assignment of descriptive attributes to non-existent objects.This potentially presents a prob-lem.An actual instantiation(i.e.,a database)will contain value assignments only for the descriptive attributes of existing objects.This suggests that in order to compute the likelihood of a database we need to sum over all possible values of descriptive attributes of potential objects that do not exist. (As we shall see,such likelihood computations are integral part of our learning procedure.)Fortu-nately,we can easily see that the de?nition of ensures that if,then variables of the form are independent of all other variables in the instance.Thus,we can ignore such descriptive attributes when we compute the likelihood of a database.

The situation with Exists attribute is somewhat more complex.When we observe a database,we also observe that many potential objects do not exist.The non-existence of these objects can provide information about other attributes in the model,which is taken into consideration by the correlations between them and the Exists attributes in the PRM.At?rst glance,this idea presents computational dif?culties,as there can be a very large number of non-existent objects.However,we note that the de?nition of is such that we need to compute false Pa only for objects whose slots refer to existing objects,thereby bounding the number of non-existent objects we have to consider.

5.Example:Word models

Our two models of structural uncertainty induce simple yet intuitive models for link existence. We illustrate this by showing a natural connection to two common models of word appearance in documents.Suppose our domain contains two entity classes:Document,representing the set of documents in our corpus,and Words,representing the words contained in our dictionary.Doc-uments may have descriptive attributes such as Topic;dictionary entries have the attribute Word, which is the word itself,and may also have additional attributes such as the type of word.The relationship class Appearance represents the appearance of words in documents;it has two slots InDoc and HasWord.In this schema,structural uncertainty corresponds to a probabilistic model of the appearance of words in documents.

In existence uncertainty,the class Appearance is an undetermined class;the potential objects in this class correspond to document-word pairs,and the assertion true means that the particular dictionary entry appears in the particular document.Now,suppose that has the parents Appearance.InDoc.Topic and Appearance.HasWord.Word.This implies,that,for each word and topic,we have a parameter which is the probability that a word appears in a document of topic.Furthermore,the different events

are conditionally independent given the topic.It is easy to see that this model is equivalent to the model often called binary naive Bayes model(McCallum and Nigam,1998),where the class variable is the topic and the conditionally independent features are binary variables corresponding to the appearance of different dictionary entries in the document.

17

When using reference uncertainty,we can consider several modeling alternatives.The most

straightforward model is to view a document as a bag of words.Now,Appearance also in-

cludes an attribute that designates the position of the word in the document.Thus,a document

of words has related Appearance objects.We can provide a probabilistic model of word ap-

pearance by using reference uncertainty over the slot Appearance.HasWord.In particular,if we

choose HasWord Word,then we have a multinomial distribution over the

words in the dictionary.If we set Appearance.InDoc.Topic as the parent of the selector variable HasWord,then we get a different multinomial distribution over words for each topic. The result is a model where a document is viewed as a sequence of independent samples from a

multinomial distribution over the dictionary,where the sample distribution depends on the docu-

ment topic.This document model is called the multinomial Naive Bayesian model(McCallum and

Nigam,1998).

Thus,for this simple PRM structure,the two forms of structural uncertainty lead to models that

are well-studied within the statistical NLP community.However,the language of PRMs allows us

to represent more complex structures:Both the existence and reference uncertainty can depend on

properties of words rather than on the exact identity of the word;for example they can also depend

on other attributes,such as the research area of the document’s author.

6.Learning PRMs with Structural Uncertainty

In this section,we brie?y review the learning algorithm for the basic PRM framework in Friedman

et al.(1999)and describe the modi?cations needed to handle the two extensions of PRMs proposed

in the previous sections.Our aim is to learn such models from data:given a schema and an instance,

construct a PRM that captures the dependencies between objects in the schema.We stress that basic

PRMs and both proposed variants are learned using the same type of training data:a complete

instantiation that describes a set of objects,their attribute values and their reference slots.However,

in each variant,we attempt to learn somewhat different structure from this data.For basic PRMs,

we learn the probability of attributes given other attributes;for PRMs with reference uncertainty,we

also attempt to learn the rules that govern the choice of slot references;and for PRMs with existence

uncertainty,we attempt to learn the probability of existence of relationship objects.

6.1Learning basic PRMs

We separate the learning problem into two tasks:evaluating the“goodness”of a candidate structure,

and searching the space of legal candidate structures.

Model Scoring For scoring candidate structures,we adapt Bayesian model selection Heckerman

(1998).We compute the posterior probability of a PRM given an https://www.wendangku.net/doc/918971742.html,ing Bayes

rule we have that.This score is composed of two main parts:the

prior probability of,and the probability of the instantiation assuming the PRM is.By making

fairly reasonable assumptions about the prior probability of structures and parameters,this term can

be decomposed into a product of terms Friedman et al.(1999).As in Bayesian network learning,

each term in the decomposed form of the score measures how well we predict the values of

given the values of its parents.Moreover,the term for depends only on the suf?cient

statistics C,that count the number of entities with and Pa.

18

Model Search To?nd a high-scoring structure in basic PRM framework,we use a simple search procedure that considers operators such as adding,deleting,or reversing edges in the dependency model.The procedure performs greedy hill-climbing search,using the Bayesian score to evaluate structures.As in Bayesian network search,we can take advantage of score decomposibility to perform the hill-climbing ef?ciently:after adding or deleting a parent of an attribute,the only steps that need to be re-scored are other edges that add/delete parents for this attribute.

6.2Learning with reference uncertainty

The extension to scoring required to deal with reference uncertainty is not a dif?cult one.Once we ?x the partitions de?ned by the attributes,a CPD for compactly de?nes a distribution over values of.Thus,scoring the success in predicting the value of can be done ef?ciently using standard Bayesian methods used for attribute uncertainty(e.g.,using a standard Dirichlet prior over values of).

The extension to search the model space for incorporating reference uncertainty involves expand-ing our search operators to allow the addition(and deletion)of attributes to partition de?nition for each reference slot.Initially,the partition of the range class for a slot is not given in the model. Therefore,we must also search for the appropriate set of attributes.We introduce two new op-erators re?ne and abstract,which modify the partition by adding and deleting attributes from. Initially,is empty for each.The re?ne operator adds an attribute into;the abstract operator deletes one.As mentioned earlier,we can de?ne the partition simply by looking at the cross product of the values for each of the partition attributes,or using a decision tree.In the case of a decision tree,re?ne adds a split to one of the leaves and abstract removes a split.These newly introduced operators are treated by the search algorithm in exactly the same way as the standard edge-manipulation operators:the change in the score is evaluated for each possible operator,and the algorithm selects the best one to execute.

We note that,as usual,the decomposition of the score can be exploited to substantially speed up the search.In general,the score change resulting from an operator is re-evaluated only after applying an operator that modi?es the parent or partition set of an attribute that modi?es.This is also true when we consider operators that modify the parent of selector attributes.

6.3Learning with Existence Uncertainty

The extension of the Bayesian score to PRMs with existence uncertainty is straightforward;the ex-ists attribute is simply a new descriptive attribute.The only new issue is how to compute suf?cient statistics that include existence attributes without explicitly enumerating all the non-existent entities.We perform this computation by counting,for each possible instantiation of Pa, the number of potential objects with that instantiation,and subtracting the actual number of ob-jects with that parent instantiation.Let be a particular instantiation of Pa.To compute C true,we can use standard database query to compute how many objects have Pa.To compute C false,we need to compute the number of potential entities.We can do this without explicitly considering each by decomposing

the computation as follows:Let be a reference slot of with.Let Pa be the subset of parents of along slot and let be the corresponding instantiation.We count the number of consistent with.If Pa is empty,this count is simply.The product

19

Figure7:,the PRM learned using reference uncertainty.The reference slots are Role.Movie, Role.Actor,Vote.Movie,and Vote.Person.Dashed lines indicate attributes used in de?n-

ing the partition.

of these counts is the number of potential entities.To compute C false,we simply subtract C true from this number.

No extensions to the search algorithm are required to handle existence uncertainty.We simply introduce the new attributes,and integrate them into the search space.Our search algorithm now considers operators that add,delete or reverse edges involving the exist attributes.As usual, we enforce coherence using the class dependency graph.In addition to having an edge from to for every slot whose range type is,when we add an edge from to,we add an edge from to and an edge from to.

7.Experimental Results

We evaluated our learning algorithms on several real-world data sets.In this section,we describe the PRM learned for a domain using both of our models for representing structural uncertainty. In each case,we compare against a simple baseline model.Our experiments used the Bayesian score with a uniform Dirichlet parameter prior with equivalent sample size,and a uniform distribution over structures.The partitions are de?ned using the cross-product of values of the partition attributes;they are not with using decision trees.

7.1Predictive ability

We?rst tested whether the additional expressive power allows us to better capture regularities in the domain.Toward this end,we evaluated the likelihood of test data given our learned model. Unfortunately,we cannot directly compare the likelihood of the EU and RU models,since the

20

知识管理和信息管理之间的联系和区别_[全文]

引言：明确知识管理和信息管理之间的关系对两者的理论研究和实际应用都有重要意义。从本质上看,知识管理既是在信息管理基础上的继承和进一步发展,也是对信息管理的扬弃,两者虽有一定区别,但更重要的是两者之间的联系。 知识管理和信息管理之间的联系和区别 By AMT 宋亮 相对知识管理而言,信息管理已有了较长的发展历史,尽管人们对信息管理的理解各异,定义多样,至今尚未取得共识,但由于信息对科学研究、管理和决策的重要作用,信息管理一词一度成为研究的热点。知识管理这一概念最近几年才提出,而今知识管理与创新已成为企业、国家和社会发展的核心,重要原因就在于知识已取代资本成为企业成长的根本动力和竞争力的主要源泉。 信息管理与知识管理的关系如何?有人将知识管理与信息管理对立起来,强调将两者区分开;也有不少人认为知识管理就是信息管理;多数人则认为,信息管理与知识管理虽然存在较大的差异,但两者关系密切,正是因为知识和信息本身的密切关系。明确这两者的关系对两者的理论研究和实际应用都有重要意义。从本质上看,知识管理既是在信息管理基础上的继承和进一步发展,也是对信息管理的扬弃,两者虽有一定区别,但更重要的是两者之间的联系。 一、信息管理与知识管理之概念 1、信息管理 “信息管理”这个术语自本世纪70年代在国外提出以来,使用频率越来越高。关于“信息管理”的概念,国外也存在多种不同的解释。1>.人们公认的信息管理概念可以总结如下:信息管理是实观组织目录、满足组织的要求,解决组织的环境问题而对信息资源进行开发、规划、控制、集成、利用的一种战略管理。同时认为信息管理的发展分为三个时期:以图书馆工作为基础的传统管理时期,以信息系统为特征的技术管理时期和以信息资源管理为特征的资源管理时期。 2、知识管理 由于知识管理是管理领域的新生事物，所以目前还没有一个被大家广泛认可的定义。Karl E Sverby从认识论的角度对知识管理进行了定义，认为知识管理是“利用组织的无形资产创造价值的艺术”。日本东京一桥大学著名知识学教授野中郁次郎认为知识分为显性知识和隐性知识，显性知识是已经总结好的被基本接受的正式知识，以数字化形式存在或者可直接数字化，易于传播；隐性知识是尚未从员工头脑中总结出来或者未被基本接受的非正式知识，是基于直觉、主观认识、和信仰的经验性知识。显性知识比较容易共享，但是创新的根本来源是隐性知识。野中郁次郎在此基础上提出了知识创新的SECI模型:其中社会化(Socialization)，即通过共享经验产生新的意会性知识的过程；外化(Externalization)，即把隐性知识表达出来成为显性知识的过程；组合（Combination），即显性知识组合形成更复杂更系统的显性知识体系的过程；内化（Internalization），即把显性知识转变为隐性知识，成为企业的个人与团体的实际能力的过程。

e-Learning 分享：2015年elearning发展趋势

2015年elearning发展趋势 1.大数据 elearning涉及的数据量越来越大，而传统的数据处理方式也变得越来越难以支撑。 仅在美国就有46%的大学生参加在线课程，而这也仅仅是一部分elearning用户而已。虽然大数据的处理是一个难题，但同时也是改善elearning的一个良好的契机。 下面是大数据改善elearning的一些案例。 大数据能够使人们更深入的了解学习过程。例如：对于（课程）完成时间及完成率的数据记录。 大数据能够有效跟踪学习者以及学习小组。例如：记录学习者点击及阅读材料的过程。 大数据有利于课程的个性化。例如：了解不同的学习者学习行为的不同之处。 通过大数据可以分析相关的学习反馈情况。例如：学习者在哪儿花费了多少时间，学习者觉得那部分内容更加有难度。 2.游戏化 游戏化是将游戏机制及游戏设计添加到elearning中，以吸引学习者，并帮助他们实现自己

的学习目标。 不过，游戏化的elearning课程并不是游戏。 elearning的游戏化只是将游戏中的元素应用于学习环境中。游戏化只是利用了学习者获得成功的欲望及需求。 那么，游戏化何以这么重要？ 学习者能够回忆起阅读内容的10%，听讲内容的20%。如果在口头讲述过程中能够添加画面元素，该回忆比例上升到30%，而如果通过动作行为对该学习内容进行演绎，那么回忆比例则高达50%。但是，如果学习者能够自己解决问题，即使是在虚拟的情况中，那么他们的回忆比例则能够达到90%！（美国科学家联合会2006年教育游戏峰会上的报告）。近80%的学习者表示，如果他们的课业或者工作能够更游戏化一些，他们的效率会更高。（Talent LMS 调查） 3.个性化 每位学习者都有不同的学习需求及期望，这就是个性化的elearning如此重要的原因。个性化不仅仅是个体化，或差异化，而是使学习者能够自由选择学习内容，学习时间以及学习方式。 相关案例： 调整教学进度使教学更加个性化； 调整教学方式使教学更加与众不同； 让学习者自己选择适合的学习方式； 调整教学内容的呈现模式（文本、音频、视频）。

信息管理与知识管理典型例题讲解

信息管理与知识管理典型例题讲解 1.（）就是从事信息分析，运用专门的专业或技术去解决问题、提出想法，以及创造新产品和服务的工作。 2.共享行为的产生与行为的效果取决于多种因素乃至人们的心理活动过程，（）可以解决分享行为中的许多心理因素及社会影响。 3.从总体来看，目前的信息管理正处（）阶段。 4.知识员工最主要的特点是（）。 5.从数量上来讲，是知识员工的主体，从职业发展角度来看，这是知识员工职业发展的起点的知识员工类型是（）。 6.人类在经历了原始经济、农业经济和工业经济三个阶段后，目前正进入第四个阶段——（），一种全新的经济形态时代。 1.对信息管理而言，( )是人类信息活动产生和发展的源动力，信息需求的不断 变化和增长与社会信息稀缺的矛盾是信息管理产生的内在机制，这是信息管理的实质所在。 2.（）就是从事信息分析，运用专门的专业或技术去解决问题、提出想法，以及创造新产品和服务的工作。 3.下列不属于信息收集的特点（） 4.信息管理的核心技术是（） 5.下列不属于信息的空间状态（） 6.从网络信息服务的深度来看，网络信息服务可以分 1.知识管理与信息管理的关系的（）观点是指，知识管理在历史上曾被视为信息管理的一个阶段，近年来才从信息管理中孵化出来，成为一个新的管理领域。 2.著名的Delicious就是一项叫做社会化书签的网络服务。截至2008年8月，美味书签已经拥有（）万注册用户和160万的URL书签。 3.激励的最终目的是（）

4.经济的全球一体化是知识管理出现的（） 5.存储介质看，磁介质和光介质出现之后，信息的记录和传播越来越（） 6.下列不属于信息的作用的选项（） 1.社会的信息化和社会化信息网络的发展从几方面影响着各类用户的信息需求。（） 2.个性化信息服务的根本就是“（）”，根据不同用户的行为、兴趣、爱好和习惯，为用户搜索、组织、选择、推荐更具针对性的信息服务。 3.丰田的案例表明：加入( )的供应商确实能够更快地获取知识和信息，从而促动了生产水平和能力的提高。 4.不属于信息资源管理的形成与发展需具备三个条件（） 5.下列不属于文献性的信息是（） 6.下列不属于（）职业特点划分的

连锁咖啡厅顾客满意度涉入程度对忠诚度影响之研究以高雄市星巴克为例

连锁咖啡厅顾客满意度涉入程度对忠诚度影响之研究以高雄市星巴克 为例 Document number【SA80SAB-SAA9SYT-SAATC-SA6UT-SA18】

连锁咖啡厅顾客满意度对顾客忠诚度之影响－以高雄 市星巴克为例 The Effects of Customer Satisfaction on Customer Loyalty—An Empirical study of Starbucks Coffee Stores in Kaohsiung City 吴明峻 Ming-Chun Wu 高雄应用科技大学观光管理系四观二甲 学号：06 中文摘要 本研究主要在探讨连锁咖啡厅顾客满意度对顾客忠诚度的影响。 本研究以高雄市为研究地区，并选择8间星巴克连锁咖啡厅的顾客作为研究对象，问卷至2006年1月底回收完毕。 本研究将顾客满意度分为五类，分别是咖啡、餐点、服务、咖啡厅内的设施与气氛和企业形象与整体价值感；将顾客忠诚度分为三类，分别是顾客再购意愿、向他人推荐意愿和价格容忍度，并使用李克特量表来进行测量。 根据过往研究预期得知以下研究结果： 1.人口统计变项与消费型态变项有关。 2.人口统计变项与消费型态变项跟顾客满意度有关。 3.人口统计变项与消费型态变项跟顾客忠诚度有关。 4.顾客满意度对顾客忠诚度相互影响。 关键词：连锁、咖啡厅、顾客满意度、顾客忠诚度 E-mail

一、绪论 研究动机 近年来，国内咖啡消费人口迅速增加，国外知名咖啡连锁品牌相继进入台湾，全都是因为看好国内咖啡消费市场。 在国内较知名的连锁咖啡厅像是星巴克、西雅图极品等。 本研究针对连锁咖啡厅之顾客满意度与顾客忠诚度间关系加以探讨。 研究目的 本研究所要探讨的是顾客满意度对顾客忠诚度的影响，以国内知名的连锁咖啡厅星巴克之顾客为研究对象。 本研究有下列五项研究目的： 1.以星巴克为例，探讨连锁咖啡厅的顾客满意度对顾客忠诚度之影响。 2.以星巴克为例，探讨顾客满意度与顾客忠诚度之间的关系。 3.探讨人口统计变项与消费型态变项是否有相关。 4.探讨人口统计变项与消费型态变项跟顾客满意度是否有相关。 5.探讨人口统计变项与消费型态变项跟顾客忠诚度是否有相关。 二、文献回顾 连锁咖啡厅经营风格分类 根据行政院（1996）所颁布的「中华民国行业标准分类」，咖啡厅是属於九大行业中的商业类之饮食业。而国内咖啡厅由於创业背景、风格以及产品组合等方面有其独特的特质，使得经营型态与风格呈现多元化的风貌。 依照中华民国连锁店协会（1999）对咖啡产业调查指出，台湾目前的咖啡厅可分成以下几类：

elearning时代：企业培训大公司先行

e-learning时代：企业培训大公司先行 正如印刷书籍造就16世纪后的现代大学教育，互联网的普及，将使学习再次发生革命性的变化。通过e-learning，员工可以随时随地利用网络进行学习或接受培训，并将之转化为个人能力的核心竞争力，继而提高企业的竞争力。 从个人到企业，从企业到市场，e-learning有着无限的生长空间，跃上时代的浪尖是必然，也是必须。 企业培训:e-learning时代已经来临 约翰-钱伯斯曾经预言“互联网应用的第三次浪潮是e-learning”，现在，这个预言正在变为现实。 美国培训与发展协会(ASTD)对2005年美国企业培训情况的调查结果显示，美国企业对员工的培训投入增长了16.4%，e-learning培训比例则从24%增长到28%，通过网络进行学习的人数正以每年300%的速度增长，60%的企业已使用网络形式培训员工;在西欧，e-learning市场已达到39亿美元规模;在亚太地区,越来越多的企业已经开始使用e-learning…… 据ASTD预测，到2010年，雇员人数超过500人的公司，90%都将采用e-learning。 e-learning:学习的革命 正如印刷书籍造就16世纪后的现代大学教育，互联网的普及，将使学习再次发生革命性的变化。 按照ASTD给出的定义,e-learning是指由网络电子技术支撑或主导实施的教学内容或学习。它以网络为基础、具有极大化的交互作用，以开放式的学习空间带来前所未有的学习体验。 在企业培训中，e-learning以硬件平台为依托，以多媒体技术和网上社区技术为支撑，将专业知识、技术经验等通过网络传送到员工面前。通过e-learning，员工可以随时随地利用网络进行学习或接受培训，并将之转化为个人能力的核心竞争力，继而提高企业的竞争力。 据IDC统计，自1998年e-learning概念提出以来，美国e-learning市场的年增长率几乎保持在80%以上。增长速度如此之快，除了跟企业培训本身被重视度日益提高有关，更重要的是e-learning本身的特性: 它大幅度降低了培训费用，并且使个性化学习、终生学习、交互式学习成为可能。而互联网及相关产品的开发、普及和商业化是e-learning升温最强悍的推动力。 数据表明，采用e-learning模式较之传统模式至少可节约15%-50%的费用，多则可达

国内外知识共享理论和实践述评

●储节旺,方千平(安徽大学　管理学院,安徽　合肥　230039) 国内外知识共享理论和实践述评 3 摘　要:知识共享是知识管理的重要环节之一。本文评析了知识共享的内涵、类型、过程、模式、实践。对知识共享内涵的认识有信息技术、沟通、组织学习、市场、系统等角度;对知识共享的类型的认识有共享的主体、共享的知识类型等角度;对知识共享的过程的认识有交流、知识主体等角度;知识共享的模式包括SEC I 模式、行动—结果的模式、基于信息传递的模式、正式和非正式的模式。对知识共享的实践主要总结了一些著名公司的做法和成绩。 关键词:知识共享;知识管理;知识共享模式;理论研究 Abstract:Knowledge sharing is the i m portant link in the knowledge manage ment chain .This paper revie ws the meaning,type,p r ocess,model and p ractice of knowledge sharing .The meaning of knowledge sharing is su mma 2rized fr om the pers pectives of I T,communicati on,organizati on learning,market and syste m.The type of knowl 2edge sharing is su mmarized fr om the pers pectives of the main body and the type of knowledge sharing .The p r ocess of knowledge sharing is su mmarized fr om the pers pectives of communicati on and the main body of knowledge .The 2model of knowledge sharing is su mmarized fr om the pers pectives of SEC I,acti on -result,inf or mati on -based deliv 2ery and for mal and inf or mal type .The p ractice of knowledge sharing is su mmarized in s ome famous cor porati ons ’method of work and their achieve ments . Keywords:knowledge sharing;knowledge manage ment;model of knowledge sharing;theoretical study 3本文为国家社会科学基金项目“国内外知识管理理论发展与学科体系构建研究”(项目编号:06CT Q009)及安徽大学人才建设计划项目“企业应对危机的知识管理问题研究”研究成果之一。 知识管理的目标是为了有效地开发利用知识,实现知识的价值,其前提与基础是知识的获取与积累,其核心是知识的交流与学习,其追求的目标是知识的吸收与创造,从而最终达到充分利用知识获得效益、获得先机、获得竞争优势的目的。在这个过程中最根本的前提是知识的存在,包括知识的质与量。而如何拥有足够质与量的知识在于知识的积累与共享,因为个体的经验、知识和智慧都是相对有限的,而一个组织团体的知识是相对无限的,如何使一个组织的所有成员交流共享知识是知识管理达到最佳状态的关键。知识共享是发挥知识价值最大化的有效途径,知识共享是知识管理的基点,是知识管理的优势所在。因此,知识共享的理论研究和实施运营都显得十分必要。 保罗?S ?迈耶斯于1997年详细介绍了适合于知识共享的组织系统的设计方案[1]。自此以后,专家学者们开始从不同角度对知识共享展开研究,研究领域从企业不断扩 大到政府、教育等社会组织。有关知识共享的研究逐渐成为知识管理的重点。从笔者掌握的知识共享研究的相关文献来看,其研究内容主要体现在知识共享的定义类型、过程、模式等方面。 1　知识共享的定义 由于知识内涵的开放性和知识分类的复杂性,专家学者们对知识共享的内涵理解也很难形成一致的看法,但他们还是从不同的视角对知识共享的内涵提出了许多颇有见地的观点。 111　观点简述 1)信息技术角度。Ne well 及Musen 从信息技术的角 度来理解知识共享,认为必须厘清知识库中符号和知识之间存在的差异。知识库本身并不足以用来捕捉知识,而是赋予知识库意义的解释过程。因此,对知识共享而言必须包括理性判断与推理机制,才能赋予知识库生命[2]。 2)信息沟通角度。Hendriks 和Botkin 等人认为知识 共享是一种人与人之间的联系和沟通的过程。 3)组织学习角度。Senge 认为,知识共享不仅仅是一 方将信息传给另一方,还包含愿意帮助另一方了解信息的内涵并从中学习,进而转化为另一方的信息内容,并发展

知识管理系统：目标与策略

知识管理：目标与策略 摘要：知识管理是社会经济发展的主要驱动力和提高组织竞争力的重要手段。其基本内容是运用集体的智慧提高应变和创新能力。本文旨在界定知识经济的概念，探讨知识管理的目标，比较分析知识管理的两种策略之异同，以促进我国管理的创新，有利于引导我国企业步入知识经济时代。 关键词：管理；组织；创新 在人类社会的发展进程中，管理创新和技术进步可以说是推动经济增长的两个基本动力源。随着知识社会的到来，知识将成为核心和具有柔性特点的生产要素，而对知识的管理更是社会经济发展的主要驱动力和提高组织竞争力的重要手段。对组织而言，知识和信息正在取代资本和能源成为最主要的资源，知识经济迫切要求管理创新。适应此要求，近几年来，一种新的企业管理理念——知识管理（Knowledge management）正在国外一些大公司中形成并不断完善。其中心内容便是通过知识共享、运用集体的智慧提高应变和创新能力。知识管理的实施在于建立激励雇员参与知识共享的机制，设立知识总监，培养组织创新和集体创造力。总结和研究知识管理的做法和成功经验将有利于我国企业管理的创新，有利于引导我国企业步入知识经济时代。 一、概念的界定 什么是知识管理？一个定义说：“知识管理是当企业面对日益增

长着的非连续性的环境变化时，针对组织的适应性、组织的生存及组织的能力等重要方面的一种迎合性措施。本质上，它嵌涵了组织的发展过程，并寻求将信息技术所提供的对数据和信息的处理能力以及人的发明和创新能力这两者进行有机的结合。”笔者认为，知识管理虽然广泛运用于企业管理的实践，但作为具有一般管理的共同性质的公共管理同样也面临着知识管理的问题。对于公共部门而言，知识管理的目标与核心就是通过提高人的发明和创新能力来实现组织创新。 知识管理为组织实现显性和隐性知识共享提供了新的途径。显性知识易于整理和进行计算机存储，而隐性知识是则难以掌握，它集中存储在雇员的脑海里，是雇员所取得经验的体现。知识型组织能够对外部需求作出快速反应、明智地运用内部资源并预测外部环境的发展方向及其变化。虽然要做到这一点需要从根本上改变组织的发展方向和领导方式，但是其潜在回报是巨大的。要了解知识管理，首先要把它同信息管理区分开来。制定一个有效的信息管理战略并不意味着实现了知识管理，这正如不能单纯从一个组织的设备硬件层面来衡量其办公自动化水平一样。要想在知识经济中求得生存，就必须把信息与信息、信息与人、信息与过程联系起来，以进行大量创新。库珀认为：“正是由于信息与人类认知能力的结合才导致了知识的产生。它是一个运用信息创造某种行为对象的过程。这正是知识管理的目标。”实行有效知识管理所要求的远不止仅仅拥有合适的软件系统和充分的培训。它要求组织的领导层把集体知识共享和创新视为赢得竞争优势的支柱。如果组织中的雇员为了保住自己的工作而隐瞒信息，如果组

(完整版)信息管理与知识管理

信息管理与知识管理 （）是第一生产力,科学技术 （）是物质系统运动的本质特征，是物质系统运动的方式、运动的状态和运动的有序性信息按信息的加工和集约程度分（）N次信息源 不是阻碍企业技术创新的主要因素主要包括（）市场需求薄弱 不属于 E.M.Trauth博士在20世纪70年代首次提出信息管理的概念，他将信息管理分内容为（）网络信息管理 不属于W.Jones将个人信息分为三类（）个人思考的 不属于信息的经济价值的特点（）永久性 不属于信息的社会价值特征的实（）广泛性 不属于信息资源管理的形成与发展需具备三个条件（）信息内容 存储介质看，磁介质和光介质出现之后，信息的记录和传播越来越（）方便快捷 国外信息管理诞生的源头主要有（）个1个 即信息是否帮助个人获得更好的发展，指的是（）个人发展 社会的信息化和社会化信息网络的发展从几方面影响着各类用户的信息需求。（）2个 社会价值的来源不包括（）社会观念 图书馆层面的障碍因素不去包括：（）书籍复杂无法分类 下列不属于信息收集的过程的（）不提供信息收集的成果 下列不属于信息收集的特点（）不定性

下列不属于用户在信息获取过程中的模糊性主要表现（）用户无法准确表达需求 下列哪种方式不能做出信息商品（）口口相传 下列现象不是五官感知信息为（）电波 下列属于个人价值的来源（）个人需求 信息information”一词来源于哪里（）拉丁 信息管理的核心技术是（）信息的存储与处理 技术信息管理观念是专业技术人员的（基本要素 信息认知能力是专业技术人员的（基本素质） 信息是通信的内容。控制论的创始人之一维纳认为，信息是人们在适应（外部） 世界并且使之反作用于世界的过程中信息收集的对象是（信息源）由于（）是一种主体形式的经济成分，它们发展必将会导致社会运行的信息化。（信息经济） 在网络环境下，用户信息需求的存在形式有（）4种 判断题 “效益”一词在经济领域内出现，效益递增或效益递减也是经济学领域的常用词。（对） “知识经济”是一个“直接建立在知识和信息的生产、分配和使用之上的经济”，是相对于“以物质为基础的经济”而言的一种新型的富有生命力的经济形态。（对） 百度知道模块是统计百度知道中标题包含该关键词的问题，然后由问

星巴克swot分析

6月21日 85度C VS. 星巴克SWOT分析 星巴克SWOT分析 优势 1.人才流失率低 2.品牌知名度高 3.熟客券的发行 4.产品多样化 5.直营贩售 6.结合周边产品 7.策略联盟 劣势 1.店內座位不足 2.分店分布不均 机会 1.生活水准提高 2.隐藏极大商机 3.第三空间的概念 4.建立电子商务 威胁 1.WTO开放后，陆续有国际品牌进驻 2.传统面包复合式、连锁咖啡馆的经营 星巴克五力分析 １、供应商：休闲风气盛，厂商可将咖啡豆直接批给在家煮咖啡的消费者 ２、购买者：消费者意识高涨、资讯透明化（比价方便） ３、同业內部竞争：产品严重抄袭、分店附近必有其他竞争者 ４、潜在竞争者：设立咖啡店连锁店无进入障碍、品质渐佳的铝箔包装咖啡 ５、替代品：中国茶点、台湾小吃、窜红甚快的日本东洋风...等 ８５度ｃ市場ｓｗｏｔ分析： 【Strength优势】 具合作同业优势 产品精致 以价格进行市场区分，平价超值 服务、科技、产品、行销创新，机动性强 加盟管理人性化 【Weakness弱势】 通路品质控制不易 品牌偏好度不足 通路不广 财务能力不健全 85度c的历史资料，在他们的网页上的活动信息左邊的新聞訊息內皆有詳細資料，您可以直接上網站去查閱，皆詳述的非常清楚。

顧客滿意度形成品牌重於產品的行銷模式。你可以在上他們家網站找找看！【Opportunity机会】 勇于變革变革与创新的经营理念 同业策略联盟的发展、弹性空间大 【Threat威胁】 同业竞争对手(怡客、维多伦)门市面对面竞争 加盟店水准不一，品牌形象建立不易 直，间接竞争者不断崛起(壹咖啡、City Café．．．) 85度跟星巴克是不太相同的零售，星巴客应该比较接近丹堤的咖啡厅，策略形成的部份，可以从 1.产品线的宽度跟特色 2.市场区域与选择 3.垂直整合 4.规模经济 5.地区 6.竞争优势 這6點來做星巴客跟85的区分 可以化一个表，來比较他們有什么不同 內外部的話，內部就从相同产业來分析(有什麼优势跟劣势) 外部的話，一樣是相同产业但卖的东西跟服务不太同，来与其中一个产业做比较(例如星巴客) S(优势):點心精致平价,咖啡便宜,店面设计观感很好...等等 W(劣势):对于消费能力较低的地区点心价格仍然较高,服务人员素质不齐,點心种类变化較少 O(机会):对于点心&咖啡市场仍然只有少数的品牌独占(如:星XX,壹XX...等),它可以透过连锁店的开幕达成高市占率 T(威协):1.台湾人的模仿功力一流,所以必须保持自己的特色做好市场定位 2.消費者的口味变化快速,所以可以借助学者"麥XX"的做法保有主要的點心款式外加上几样周期变化的點心 五力分析 客戶讲价能力（the bargaining power of customers）、 供应商讲价能力（the bargaining power of suppliers）、 新进入者的竞争（the threat of new entrants）、 替代品的威协（the threat of substitute products）、 现有厂商的竞争（The intensity of competitive rivalry）

知识管理

知识管理?技术平台构建_建立知识管理系统时应该 注意的几个方面 著者:于涛部门:泛微大连项目管理部 实施知识管理，不但需要推行知识管理的理念，建立知识管理组织及创建知识共享的文化，还要通过知识管理系统来使之变为现实。但在利用知识管理系统的时候一定要注意以下几个方面： 1、服从知识管理目标 构建知识管理系统，首先要服从知识管理的整体目标，即能够“将最恰当的知识在最恰当的时间传给最恰当的人”，使他们能够做出最好的决策，只有达到了这个目标，才能成为真正的知识管理系统。其次要清楚知识管理的具体目标，究竟是为了更快地解答客户的咨询，是避免公司一旦因员工流失而出现青黄不解，还是为了降低运作成本？在知识管理系统实施的每一步开始前，都应该十分清楚该阶段预期达到的成效和项目的成果是什么。 2、尊重“以人为本” 知识管理系统的规划必须“以人为本”，而不要以文字知识为核心。实施时应当清楚的定义企业使用知识管理系统的用户是谁，他们的具体要求是什么，用户应当能够通过知识管理系统定义知识、发现知识、整合企业内外众多的知识源，并对其分类和组织。 3、鼓励知识共享行为 一般，鼓励知识共享更多地依赖于组织的内部管理，但知识管理系统能够为组织实施各种知识共享措施提供支持。例如：提供意见反馈、评分与建议等用户之间交互的工具，增加用户参与感。从系统提供的机制中，管理层可以清楚地得知用户的参与及贡献度，并能由用户反馈、使用频率明确地判别其知识价值。 4、符合知识创新和转化的需求 知识管理系统能够将企业每个成员纳入管理范围之内，为其提供创新所必须的知识库，通过规定的知识创新步骤、设计好的知识表达的规范格式来引导成员将其隐性知识显性化。 5、注意系统安全问题 由于知识管理系统可能涉及到组织的核心知识资产，所以系统要有一定的技术手段保护核心知识的流失，例如对相关的分析数据或文档信息有一套权限管理机制，并满足用户个性化需求。 案例： 埃森哲的知识管理系统被誉为“全球知识共享网络”（lotus notes），系统利用KX（knowledge xchange）黄页提供多种查询途径。全球47个国家110个办事处共有75000个用户。员工通过黄页可以迅速找到所需资料库加入到个人的工作站中。资料库种类繁多、内容丰富，全球同步随时更新。资料库包括图书库/实务帮助、论坛、行业分析等，可以按照不同市场/行业/或服务领域分门别类。

知识管理案例之一Marconi的知识共享

知识管理案例之一——Marconi的知识共享 https://www.wendangku.net/doc/918971742.html,./ewkArticles/Category111/Article14375.htm 来源：.cio. 翻译特约撰稿人：王晨 2003-6-6 投稿 Marconi进入了一场并购狂潮，在3年的时间内先后收购了10家通信公司，随之而来的便是严峻的挑战：这个30亿美元通信设备的生产巨头如何保证它的技术支持人员了解最新收购的技术并在中向客户快速而准确的提供答案呢？Marconi如何培养熟悉公司所有产品的新人员？ Marconi的技术支持人员（遍布全球14个呼叫中心的500个工程师）每个月要回复大约10，000个有关公司产品的问题。在并购之前，技术支持人员依赖于公司的外部网（Extranet）TacticsOnline，他们和客户在那里可以查询最近的问题和文本文档。当新的技术支持人员和产品加入公司以后，Marconi希望用一个更加全面的知识管理系统来补充。而新加盟公司的工程师却不愿将他们一直支持的产品的知识共享。“他们觉得他们的知识就象一X安全的毯子，确保他们不会失去工作”，负责管理服务技术与研发的主管DaveBreit说，“进行并购之后，最为重要的就是避免知识囤积，而要使它们共享”。 与此同时，Marconi还想通过把更多的产品和系统信息呈现给客户以及缩短客户的长度来提高客户服务部门的效率。“我们希望在提供客户自助服务的Web页和增加技术支持人员之间进行平衡”，Breit说，“我们还希望能够为我们的一线工程师（直接与客户打交道的）更加迅速的提供更多的信息，使得他们能够更快的解决客户的问题”。 建立知识管理的基础 当Marconi1998年春开始评估知识管理技术时，在技术支持人员中共享知识已经不是什么新鲜的概念了。技术支持人员已经习惯了在3到4人的团队中工作，以作战室的方式集合起来解决客户的技术问题。而且一年前，Marconi已经开始将技术支持人员季度奖的一部分与他们提交给TacticsOnline的知识多少、他们参与指导培训其他技术支持人员的情况进行挂钩。“每位技术服务人员需要教授两次培训课，写10条FAQ，才能获得全奖”，Breit说，“当新公司加入以后，新的技术支持人员将实行同样的奖金计划，这个办法使我们建立了一个相当开放的知识共享的环境”。 为了增强TacticsOnline，Marconi从ServiceWareTechnologies那里选择了软件，部分是因为他们的技术很容易与公司的排障客户关系管理系统（RemedyCRMSystem）集成，技术支持人员用这个系统来记录客户请求，跟踪客户交流。另外，Breit指出，Marconi希望它的技术支持人员能够利用已有的产品信息的Oracle数据库。 Breit的部门花了六个月的时间实施新系统并培训技术支持人员。这个被称为知识库（KnowledgeBase）的系统可以与公司的客户关系管理系统相连接，由强大的Oracle数据库所支持。Marconi客户与产品集成的观点为技术服务人员提供了全面的交流历史纪录。例如，技术服务人员可以将制造者加入数据库，立即接过上一位技术服务人员与客户进行交流。 在第一线 TacticsOnline补足了这个新系统。“知识库中的数据是关于我们不同产品线的特定的排障技巧和线索”，知识管理系统的管理员ZehraDemiral说，“而TacticsOnline更多的则是客户进入我

2019年温州一般公需科目《信息管理与知识管理》模拟题

2019年温州一般公需科目《信息管理与知识管理》模拟题 一、单项选择题(共30小题，每小题2分) 2.知识产权的时间起点就是从科研成果正式发表和公布的时间，但有期限，就是当事人去世（）周年以内权利是保全的。 A、三十 B、四十 C、五十 D、六十 3.（）国务院发布《中华人民共和国计算机信息系统安全保护条例》。 A、2005年 B、2000年 C、1997年 D、1994年 4.专利申请时，如果审察员认为人类现有的技术手段不可能实现申请者所描述的结果而不给予授权，这体现了（）。 A、技术性 B、实用性 C、创造性 D、新颖性 5.根据本讲，参加课题的研究策略不包括（）。 A、量力而行 B、逐步击破 C、整体推进 D、有限目标 6.本讲认为，传阅保密文件一律采取（）。 A、横向传递 B、纵向传递 C、登记方式 D、直传方式 7.本讲提到，高达（）的终端安全事件是由于配置不当造成。 A、15% B、35% C、65% D、95% 8.本讲认为知识产权制度本质特征是（）。 A、反对垄断 B、带动创业 C、激励竞争 D、鼓励创新 9.本讲指出，以下不是促进基本 公共服务均等化的是（）。 A、互联网+教育 B、互联网+医 疗 C、互联网+文化 D、互联网+工 业 10.根据（）的不同，实验可以分 为定性实验、定量实验、结构分析实 验。 A、实验方式 B、实验在科研中 所起作用C、实验结果性质D、实验 场所 11.本讲认为，大数据的门槛包含 两个方面，首先是数量，其次是（）。 A、实时性 B、复杂性 C、结构化 D、非结构化 12.2007年9月首届()在葡萄牙里 斯本召开。大会由作为欧盟轮值主席 国葡萄牙的科学技术及高等教育部 主办，由欧洲科学基金会和美国研究 诚信办公室共同组织。 A、世界科研诚信大会 B、世界学术诚信大会 C、世界科研技术大会 D、诺贝尔颁奖大会 13.（）是2001年7月15日发现 的网络蠕虫病毒，感染非常厉害，能 够将网络蠕虫、计算机病毒、木马程 序合为一体，控制你的计算机权限， 为所欲为。 A、求职信病毒 B、熊猫烧香病 毒 C、红色代码病毒 D、逻辑炸弹 14.本讲提到，（）是创新的基础。 A、技术 B、资本 C、人才 D、知 识 15.知识产权保护中需要多方协 作，（）除外。 A、普通老百姓 B、国家 C、单位 D、科研人员 17.世界知识产权日是每年的（）。 A、4月23日 B、4月24日 C、4月25日 D、4月26日 18.2009年教育部在《关于严肃 处理高等学校学术不端行为的通知》 中采用列举的方式将学术不端行为 细化为针对所有研究领域的()大类行 为。 A、3 B、5 C、7 D、9 19.美国公民没有以下哪个证件 （）。 A、护照 B、驾驶证 C、身份证 D、社会保障号 20.关于学术期刊下列说法正确 的是（）。 A、学术期刊要求刊发的都是第 一手资料B、学术期刊不要求原发 C、在选择期刊时没有固定的套 式 D、对论文的专业性没有限制 22.知识产权的经济价值（），其 遭遇的侵权程度（）。 A、越大，越低 B、越大，越高 C、越小，越高 D、越大，不变 23.（）是一项用来表述课题研究 进展及结果的报告形式。 A、开题报告 B、文献综述 C、课题报告 D、序论 25.1998年，（）发布《电子出版 物管理暂行规定》。

Elearning平台未来发展新趋势

Elearning 平台未来发展新趋势

随着e-Learning概念被大众逐渐所接受，如今许多e-Learning公司纷纷涌现，现基本形成三大氛围趋势。一类提供技术（提供学习管理平台，如上海久隆信息jite公司、Saba公司、Lotus公司）、一类侧重内容提供（如北大在线、SkillSoft公司、Smartforce公司、Netg公司）、一类专做服务提供（如Allen Interactions公司）。不过随着e-Learning发展，提供端到端解决方案的公司将越来越多，或者并不局限于一个领域，如目前北大在线除提供职业规划与商务网上培训课件外，还提供相应的网络培训服务，e-Learning作为企业发展的添加剂，必将成为知识经济时代的正确抉择。 e-Learning的兴起本身与互联网的发展应用有着密切关联，它更偏重于与传统教育培训行业相结合，有着扎实的基础及发展前景，因而在大部分互联网公司不景气之时，e-Learning公司仍然被许多公司所看好。 权威的Taylor Nelson Sofresd对北美的市场调查表明，有94%的机构认识到了e-Learning的重要性，雇员在10000人以上的公司62.7%都实施了e-Learning，有85%的公司准备在今后继续增加对e-Learning的投入。而54%的公司已经或预备应用e-Learning来学习职业规划与商务应用技能。 面对如此庞大的e-Learning市场，全世界e-Learning公司将面临巨大机遇，但竞争也更加激烈。优胜劣汰成为必然。将来只有更适应市场需求，确保质量，树立信誉，建设自有品牌的公司才更具备有竞争力。那么未来e-Learning到底将走向何方？ 如今很多平台开发厂商以及用户都很关心这个问题。其实从e-Learning诞生到现在，如今大家有目共睹，移动终端上的3-D让我们的体验飘飘欲仙，该技术放大了我们的舒适度；云计算使得软件的访问轻而易举；全息图像为学习环境提供了逼真的效果；个性化的LMS使得用户可以学会更多专业人员的从业经验；更多的终端、更低廉的带宽连接为现实向世界提供解决方案的能力。 个性化学习是e-Learning平台共同追求的目标，力求解决单个群体的日常需求，将成为elearning平台下一阶段发展的新追求。那么个性化的学习将为何物呢？它又将通过何途径来实现呢？ 移动学习 移动学习（M-learning）是中国远程教育本阶段发展的目标方向，特点是实现“Anyone、Anytime、Anywhere、Any style”（4A）下进行自由地学习。移动学习依托目前比较成熟的无线移动网络、国际互联网以及多媒体技术，学生和教师使用移动设备（如无线上网的便携式计算机、PDA、手机等），通 模板版次：2.0 I COPYRIGHT? 2000- JITE SHANGHAI

知识管理制度通用

[ 蓝凌管理咨询报告之三] ****集团基于LKS4.0知识管理 实施推进制度 （建议稿） 目录 一、知识管理的目标和策略 二、知识管理组织架构和岗位主要职责 三、知识贡献管理 四、知识内容的推进管理 五、系统内知识的安全管理 六、知识管理的绩效考评 七、附则 附件1：知识需求贡献表（样表） 附件2：内部专家名单表(含专家背景调查表)附件3：常用资料需求表

附件4：定期检查的知识表 附件5：模块的权限分配 深圳市蓝凌管理咨询支持系统有限公司 地址：深圳高新区中区麻雀岭工业区7栋五楼A区 电话：（0755）26710066-1808、1708 传真：（0755）26710099 邮编：518057 ****集团知识管理制度（建议稿） 一、知识管理的目标和策略 1.1目标 1.1.1****集团的知识管理目标，就是要把支持企业各方面工作的信息、知识管理起来，提高工作效率、保证工作质量、降低工作成本。通过“将最恰当的知识在最恰当的时间传递给最恰

当的人”，使集团全体员工掌握好企业知识，建立和强化企业的核心利润源，谋取企业长期的、稳定的、增长的利润。 1.1.2 较具体地, 要在这次知识管理的推动中，建立起集团的知识体系架构；整理出全集团各岗位工作所产生的知识、员工贡献的知识、员工所需求的知识、常用的知识、基础知识、专家知识、员工的工作经验和总结等知识内容；确定这些知识内容的管理方式；通过工具平台、组织和制度逐步使知识管理工作走向正规化。 1.2 策略 1.2.1 根据****集团业务和工作特点，进行知识管理信息化办公平台和知识管理组织和机制的建设，在专业咨询机构的指导下，分阶段稳步推进实施。 1.2.2 知识管理信息化办公平台选用蓝凌公司的LKS4.0系统。该系统以IBM/LOTUS 的DOMINO协作平台为基础，它是基于知识管理的工作平台，可以满足****集团的知识管理需求。

完善知识管理理论构建知识共享系统_王玉晶

完善知识管理理论构建知识共享系统 王玉晶 【摘　要】知识共享是知识管理的一个重要组成部分。本文从知识管理的概念入手,指出了知识共享在知识管理中的重要性和特殊地位,分析了企业中进行知识共享存在的障碍与不利因素,从建立知识共享技术支撑体系、构建知识共享组织结构、营造知识共享文化、完善知识共享激励机制等方面深入探讨了如何提高企业知识共享程度,加强企业竞争力。 【关键词】知识管理　知识共享　体系 Abstract:Kno wledge sharing is an important component of the kno wledge management.This paper obtained from the kno wledge management concept,has pointed out the importance and the special status of knowledge sharing in the knowledge manage ment,and analyzed the barrier and th e disadvantage factor of knowledge sharing in the enterprise.This article pointed out that the knowledge sharing should be established technical support system,built knowledge sharing structure,created a kno wledge sharing culture,and improved knowledge sharing incentive mechanism. Key words:kno wledge sharing　knowledge management　system 当今社会,企业竞争力已不局限于仅仅依赖其规模、信息和技术,而更注重其创新和应变能力。知识管理就是在知识经济大背景下,以当代信息技术为依托,着力于知识的开发和利用、积累和创新,帮助人们共享信息,进而对扩大了的知识资本进行有效运营,最终通过集体智慧提高企业创新和应变能力的一种全新的信息管理理论与方法。而创新本身,无论是技术创新还是管理创新,其实质就是一种新知识的创造,它要求企业内部各个部门之间以及各员工之间,企业内部与外部之间进行广泛的知识交流与共享。所以知识共享一直是知识管理的一个重要组成部分。它是发挥知识的效用,实现知识价值的必由之路,是实现知识共享的前提,是走向知识共享世界的先导。理解知识共享的内涵,分析企业知识共享存在的障碍与不利因素并提出解决的方案,对完善组织中的知识共享机制具有重要意义。 1　知识共享的含义 关于知识共享的概念,由于标准、角度的不同,学者对知识共享的界定也有所不同。知识共享是指个体知识、组织知识通过各种交流手段为组织中其他成员所共享。同时,通过知识创新,实现组织的知识增值。对知识共享的认识应从知识共享的对象———知识 内容,知识共享的手段———知识网络、会议及团队学习,知识共享的主体———个人、团队及组织这三个层面考虑。其中,人是知识共享的主体,文化是知识共享的背景,信息技术是知识共享的重要工具。 2　知识共享障碍因素分析 2.1　技术应用障碍 知识管理作为一种全新的管理模式,如果缺乏应有的技术支撑,将难以发挥其应有的效力,最终失之于形式,流于空泛。许多企业物质技术基础薄弱,缺少有效的计算机网络和通信系统,很多员工不知道到哪里去寻找所需要的知识。存在于员工头脑中的隐性知识难以明确的被他人观察、了解,也无法用言语表达或者只能用言语进行部分表达。因此,薄弱的技术基础不能够帮助人们跨越时间、空间和知识的数量及质量的限制,从而不能为有效地实现知识共享提供强有力的技术支持。 2.2　组织结构缺陷 传统的组织结构管理层次多,信息传递速度缓慢,信息衰退或信息失真现象非常严重,员工之间没有超越命令以外的直接接触和交流,无法实现面对面的互动式交流,无法突破岗位对个人的约束。多层管理结构森严,越级的知识传递几乎不可能实现;员工之间 27 RESEARCH ES I N LIBRARY SCIENCE　 DOI:10.15941/https://www.wendangku.net/doc/918971742.html, ki.issn1001-0424.2008.05.027