HYBRID WIKIS: EMPOWERING USERS TO COLLABORATIVELY

STRUCTURE INFORMATION

Florian Matthes, Christian Neubert and Alexander Steinhoff

Software Engineering for Business Information Systems (sebis), Technische Universit¨at M¨unchen

Boltzmannstr. 3, 85748 Garching, Germany

Keywords:

Enterprise wikis, Semantic wikis, Enterprise information management, Enterprise 2.0, Web collaboration.

Abstract:

Wikis are increasingly used for collaborative enterprise information management since they are ﬂexibly appli-

cable and encourage the contribution of knowledge. The fact that ordinary wiki pages contain pure text only

limits how the information can be processed or made accessible to users. Semantic wikis promise to solve this

problem by capturing knowledge in structured form and offering advanced querying capabilites. However, it

is not obvious for business users, how they can beneﬁt from providing semantic annotations which are not

familiar to them and often difﬁcult to enter.

In this paper, we ﬁrst introduce the concepts of hybrid wikis, namely attributes, type tags, attribute sugges-

tions, and attribute deﬁnitions with integrity constraints. Business users interact with these concepts using a

familiar user interface based on forms, spreadsheet-like tables, and auto-completion for links and values. We

then illustrate these concepts using an example scenario with projects and persons and highlight key imple-

mentation aspects of a Java-based hybrid wiki system (Tricia). The paper ends with the description of practical

experiences gained in two usage scenarios, a comparison with related work and an outlook on future work.

1 MOTIVATION AND PROBLEM

STATEMENT

To keep pace with the growing amount of digital in-

formation that has to be managed, enterprises have

to adopt new tools and methods (Edmunds and Mor-

ris, 2000). In the recent past, wikis are increas-

ingly used as lightweight shared knowledge reposi-

tories that allow to collaboratively gather and consol-

idate information that was previously scattered across

emails, ﬁles on personal computers and paper docu-

ments (Stocker and Tochtermann, 2009). Having this

information integrated in a central place, being able to

search it and to connect related pieces of information

with hyperlinks is in fact a major advance.

However, with a growing knowledge base soon

the demand arises to access information in more

structured ways that classical wikis do not support.

For example it is not possible to query a wiki for a

company’s research projects that started in the year

2010 or to export data about these project to a spread-

sheet. So even if only rudimentary structured query-

ing functionality is required, the enterprises have to

resort to separate applications, usually specialized to

manage information of particular domains (like em-

ployees, projects or customers) or they have to de-

velop customized solutions. In both cases the advan-

tages of storing information in a central repository are

lost.

Technically, semantic wikis are promising tools

to tackle this problem. They allow to combine the

textual content with structured data. Typically, users

have to provide this data in the form of semantic an-

notations to wiki pages or parts thereof. The struc-

tured part of the information in the wiki can then be

queried similar to the contents of a database. How-

ever, in practice they are rarely used as a general pur-

pose tool that dynamically adapts to new needs. In

contrast semantic wikis often are pre-conﬁgured by

experts to solve rather speciﬁc problems. Although,

from a theoretical point of view, they can be used to

structure arbitrary information, there are several bar-

riers users are facing when editing content:

• Usually, a special syntax has to be used to add

semantic annotations which makes it difﬁcult and

cumbersome to edit structured content.

• The modelling concepts are not familiar to the

users.

• It is not obvious for users how they can beneﬁt

250

Matthes F., Neubert C. and Steinhoff A..

HYBRID WIKIS: EMPOWERING USERS TO COLLABORATIVELY STRUCTURE INFORMATION.

DOI: 10.5220/0003509402500259

In Proceedings of the 6th International Conference on Software and Database Technologies (ICSOFT-2011), pages 250-259

ISBN: 978-989-8425-76-8

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

from providing semantic annotations.

This paper describes a novel approach to mitigate

these problems.

In Section 2, our approach of so-called hybrid

wikis is presented and illustrated using an example

scenario. The term ‘hybrid’ expresses that a subset of

the features of semantic wikis are integrated into clas-

sic wiki software. The main modelling concepts are

described and the limitations of the approach are dis-

cussed. Important and interesting technical aspects of

the implementation are covered in Section 3. Section

4 contains two case studies demonstrating the prac-

ticability of the approach. In section 5, we give an

overview of related work and highlight some exam-

ples of semantic wikis that use different approaches

to facilitate data entry. The paper concludes with a

short summary and an outlook on further research and

planned improvements of the prototype.

2 HYBRID WIKIS

In this section, our approach to the problem of facil-

itating the management of structured data for non-

expert users is presented. Before we describe the

structuring mechanisms that are available in the cur-

rent software implementation on a conceptual level,

we ﬁrst give an overview of our goals, assumptions

and motives that guided its development. Finally, we

demonstrate how the modelling concepts can be ap-

plied in practice using a small example scenario.

2.1 General Ideas and Design Rationale

In the development of hybrid wikis, our primary goal

is to lower the barriers for non-expert users, that have

been described in Section 1. This means for using

the wiki neither special knowledge of wiki syntax or

modelling concepts should be required nor should the

user be forced to learn a query language to utilize the

structured part of the wiki content. We try to ﬁnd

ways to enable all users to enter structured data, in

contrast to a two-phase process where unexperienced

users enter textual content that is later enriched with

annotations by experts.

From our point of view, all attempts to translate

between the expressivity of classic, established se-

mantic web formalisms (like RDF

or OWL

) and the

user by the means of new user interfaces are unsatis-

factory. Our approach primarily focuses on the user

http://www.w3.org/RDF; visited on May 1st 2011.

http://www.w3.org/2004/OWL; visited on May 1st

2011.

and accepts that there are limitations in the complex-

ity of modelling concepts users can be expected to

understand. We try to start from lightweight struc-

turing concepts and metaphors that users are familiar

with and then, in a second step, we examine how the

data the users provide by these simple means can be

exploited by the system to offer features that usually

require a formally deﬁned data model.

Therefore, we rely mainly on simple keyword-like

annotations of wikipages, dynamically compiled and

easily extensible forms for data entry, and the pre-

sentation of data in automatically generated tabular

views. In turn, we try to avoid the notion of seman-

tic annotations being something that is optionally ap-

pended to pure text content and that is deﬁned in a

separately maintained ontology or schema. Instead

we attempt to allow the user to implicitly provide se-

mantics by ﬁlling data in particular ﬁelds of a form

or a table, by optionally creating new such ﬁelds on

demand, and by the way the data is queried and dis-

played in different contexts.

Since we do not require the users to explicitly

maintain a data model, we focus on dynamically min-

ing the wiki to ‘guess’ the model and provide users

with input options to guide them towards a consistent

data model and vocabulary. However, advanced users

can impose a schema and deﬁne certain integrity con-

straints.

For querying and browsing we provide a general

search interface that allows a faceted drill-down based

on the structured contents of the wiki pages. Further-

more, contents can be accessed in a spreadsheet-like

tabular form. Our assumption is that users feel fa-

miliar with this representation and in consequence are

less inhibited in manipulating the data.

2.2 Structuring Concepts

Hybrid wikis were implemented as an extension to the

wiki component of the commercial enterprise collab-

oration platform Tricia

. Wiki pages basically consist

of a name and some rich-text content. Pages can be

organized using text labels (tags) and page/sub-page

relationships. We added two means for structuring the

information on a wiki page: attributes and so-called

type tags. Both can be used in combination or inde-

pendently. They are described in the following.

2.2.1 Attributes

In their simplest form, attributes are key-value pairs

that can be added to wiki pages. They consist of an

attribute name – the key – and a value, being either a

http://www.infoasset.de; visited on February 1st 2011.

HYBRID WIKIS: EMPOWERING USERS TO COLLABORATIVELY STRUCTURE INFORMATION

251

short text literal or a link to another wiki page. At-

tributes do not represent metadata but constitute the

structured part of the content. For the user, the at-

tributes are presented in a box containing a list of key-

value pairs at the right border of the page (see Figure

1). The appearance is inspired by a kind of templates

widely used in Wikipedia

– and thus the MediaWiki

software

– to structure the contents of pages describ-

ing objects of the same type, like for example cities,

countries or planets. Since the box forms the struc-

tured part of the page, it can also be compared to the

so-called fact box of Semantic MediaWiki, that sum-

marizes the facts being expressed by annotations in

the text.

However, in hybrid wikis neither does the attribute

box reﬂect facts deﬁned somewhere else nor is the

selection of attributes speciﬁed by a template. This

means that new attributes can be added to individual

pages simply by adding a new entry to the list of at-

tributes. It is possible to assign multiple values for

one attribute. The values are ordered and can be a

mix of literals and links.

In order to encourage users to structure the content

they create, we strived for a convenient user interface:

When viewing a wiki page, the user is offered a se-

lection of names of attributes she might want to add.

These suggestions are shown together with the al-

ready assigned attributes at the bottom of the attribute

box so that the user only has to provide proper values.

The list of suggestions consists of the most frequent

attribute names used on similar pages (we will cover

page similarity later). Furthermore, the input ﬁelds

for attribute names and values display suggestions as

the user is typing. When an attribute value is typed,

the respective attribute name is factored in to improve

the quality of the suggestions. This makes it comfort-

able to contribute structured content and additionally

fosters consistent usage of terms.

To avoid redundancy, and thus inconsistencies,

and to foster navigation in the wiki, references from

other pages are shown in an additional ‘references’

section of the attribute box. For example if a page for

a country S references the capital M, i.e., it contains

an attribute with one value being a link to M, the page

for M will show an additional attribute entry ‘capital

of S’ having the value M. The references section of

the attribute box is similar to the incoming and outgo-

ing links as displayed in the KiWi system

http://www.wikipedia.org; visited on February 1st

2011.

http://www.mediawiki.org; visited on February 1st

2011.

http://www.kiwi-project.eu; visited on February 2st

2011.

2.2.2 Type Tags

Type tags allow the user to make a statement about the

type of the object being described on the page. They

are shown at the top of the attribute box as shown in

Figure 1. Like for the normal tags used in the wiki

(i.e., arbitrary text labels assigned to pages for cate-

gorization), users may choose an unlimited number

of terms they consider appropriate.

One of the most important functions of type tags

is to provide a reliable indicator of page similarity for

the automatically generated attribute suggestions de-

scribed above. They determine which additional at-

tributes the user is recommended to ﬁll in.

In the simplest case, exactly one type tag is as-

signed to a page. The system then determines the set

of attributes used in combination with this type tag

and displays the most frequent attribute names as sug-

gestions.

In the case of multiple type tags, ﬁrst the attributes

of pages having at least one of them assigned are de-

termined. Then, to select the most relevant attribute

names for the current page, not only the frequency

of occurrence is considered, but additionally, attribute

names are favoured when they occur together with

many or all of the given type tags.

In effect this makes it very convenient to create a

new page of a type previously used somewhere else

in the system. Once a type is assigned, providing the

attributes is hardly more demanding for the user than

ﬁlling out a form. From this point of view, type tags

can be considered a more ﬂexible alternative to tem-

plates.

It is important to note that on the one hand it was

consciously avoided to force users to explicitly create

relationships between type tags and attribute names

as it would be done when deﬁning a template. On the

other hand, for experienced users it is still possible to

assign a list of attribute names to a type tag. These

attribute names are then highlighted to the user when-

ever the respective type tag is used. For each attribute

name associated with a type tag it is further possible

to specify integrity constraints like the type of value,

the number of values or allowed value ranges.

Type tags are also used to generate lists of pages

describing objects of the same type. While this is

possible with standard tags as well, for example by

searching for all pages tagged ‘university’, the results

are more precise when type tags are used: users might

use the same tag ‘university’ to categorize pages de-

scribing for example people working at a university,

software products targeted at an academic audience or

the concept of a university. Thus, type tags allow the

user to make an important distinction when tagging

ICSOFT 2011 - 6th International Conference on Software and Data Technologies

252

Figure 1: Type tags and attributes in the context of a wiki page.

wiki pages by stating that the tagged page describes

an individual instance of the respective type – in con-

trast to vaguely relating it to a broad topic.

Retrieving a list of pages for a speciﬁc type tag

can be achieved by simply clicking on the tag. If the

pages contain attributes, they are displayed in a tab-

ular view (as shown in Figure 2). The table can be

sorted by all columns. Using the faceted search func-

tionality of the Tricia platform it is further possible to

combine type tag and attribute ﬁlters. In this way it

is possible to deﬁne very speciﬁc subsets of the wiki

contents. The results can optionally be presented in

tabular form embedded in any wiki page.

2.3 Limitations

Since the expressivity of established semantic web

technologies is sacriﬁced in favour of a better user

experience, modelling capabilities of hybrid wikis are

constrained. In particular it is not possible to explic-

itly deﬁne type-subtype relationships between type

tags. If a type tag is the generalisation of another type

tag in the system, it has to be manually ensured that

each page with the more speciﬁc type tag also has the

more general type tag assigned. However, with very

little effort it is possible to conﬁgure a wiki page that

can be used as a dashboard to monitor violations of

such integrity constraints.

For attributes it is not possible to specify any se-

mantic relation. Since no reasoning capabilities are

provided by the system it would have no effect to

add properties like symmetry or transitivity. How-

ever, as mentioned above, attributes containing links

are also visible on the target pages as a reference. This

means that relationships between two pages are al-

ways owned by one of them and it is often not triv-

ial to determine which one should be the owning side

(does capital city link to country or country to cap-

ital?). This depends on the multiplicity of the rela-

tion and in particular cases on the access rights of the

pages. However, in order to lower the barriers for en-

tering structured data, the fact that this decision possi-

bly has important implications is not explicitly com-

municated to the user. In contrast, it is relied on the

ability to invert these links later when it becomes nec-

essary.

Finally, querying capabilities are limited in so far

that results can only be ﬁltered according to the very

attributes of pages. It is for example not possible to

express a query targeting all pages having the attribute

‘Owned by’ set to a page with the type tag ‘Com-

pany’. Requiring that the attribute points to a speciﬁc

page is of course possible. From a technical point of

view, attribute (and type tag) ﬁlters could be ﬂexibly

combined using boolean operators. However, to keep

the user interface simple the current implementation

supports only conjunction of ﬁlters. From our experi-

ence this is sufﬁcient for browsing the wiki contents.

HYBRID WIKIS: EMPOWERING USERS TO COLLABORATIVELY STRUCTURE INFORMATION

253

Figure 2: Tabular view of wiki page attributes.

2.4 Example Scenario

In the following, we illustrate the modelling ca-

pabilities of hybrid wikis taking the example of a

small company’s intranet wiki. Among other things,

the wiki is used for gathering the knowledge about

projects and people relevant for the company. We as-

sume that while there may be many pages holding in-

formation about a person or project, there is one dedi-

cated page for each such entity that can be considered

the primary page which holds the basic information

about it and optionally links to pages with more spe-

ciﬁc information. We further assume that in the be-

ginning, there is only little content in the wiki and no

type tags and attributes are used.

As the number of projects increases over time,

there is a growing demand for a more structured view

on project related wiki content. Attributes for project

start and end dates are thus added to the respective

pages and the type tag ‘Project’ is assigned. Hav-

ing marked all project pages with the type tag, an

overview table of all projects is instantly available

showing sortable columns for the date attributes. At-

tribute values can be changed directly in this overview

table. By this means, consolidating the project data,

i.e., adding missing information or standardising the

representation of attribute values, is facilitated.

Let’s assume many of the employees have created

proﬁle pages for themselves in the wiki to provide so-

me information about their speciﬁc skills and experi-

ences. Some of them independently start to add at-

tributes to the project pages to express their relation-

ship with these projects, for example that they were

members of the project team or project managers. In

the beginning, people use different terms to describe

their roles. As these inconsistencies become visible

in the overview table, they are quickly harmonized by

the wiki users. It is not necessary to navigate to the

respective pages but the table cells can be edited di-

rectly. As a result, if somebody now creates a new

wiki page and assigns the type tag ‘Project’, she is of-

fered to provide values for the attributes ‘Start date’,

‘End date’, ‘Project manager’ and ‘Team’. If she now

adds a link to her proﬁle page to the ‘Project manager’

attribute, this reference will be automatically visible

on her page in the ‘References’ section of the attribute

table. A new entry ‘Project manager of’ with a single

value being the link to the project is displayed. If she

is already manager of other projects, the list will sim-

ply be extended by one entry.

Starting from this basic schema, the company can

further adapt it to new needs: Besides adding more

attributes, the existing ones can be reﬁned. The at-

tribute ‘Project manager’ can be made mandatory for

the type tag ‘Project’ so users are additionally re-

minded to provide this attribute. If it is omitted, the

page is ﬂagged as invalid. It can be further speciﬁed

that only a single link to a page having the type tag

ICSOFT 2011 - 6th International Conference on Software and Data Technologies

254

‘Person’ is accepted as a value.

New type tags can be added to distinguish differ-

ent types of projects like ‘Research project’ or ‘Inter-

nal project’. On the one hand, this allows the gen-

eration of lists and tables containing only the respec-

tive subset of the projects, on the other hand, the sys-

tem is supported in offering the user more relevant at-

tribute suggestions when editing structured data. For

example for a research project the attribute ‘Field of

research’ could be suggested whereas attributes only

relevant for internal projects are not shown. It is also

possible to add a temporal dimension to the project

types by adding type tags like ‘Project in preparation’,

‘Current project’ or ‘Completed project’. Using type

tags instead of a status attribute has the advantage that

again attribute constraints can be related to the types.

For example each page of a completed project can be

required to contain the actual end date of the project.

3 IMPLEMENTING HYBRID

WIKI

The concept of hybrid wikis is developed by mem-

bers of our chair since 2009. The system is built on

the experiences made with classic and semantic Wiki

technology as well as integrated Enterprise 2.0 plat-

forms and it is currently realized based on the mod-

elling frameworks provided by Tricia enabling the

model-driven development of web cooperation sys-

tems (B¨uchner et al., 2010). Hybrid wikis extend

the Tricia wiki functionality by a few mechanisms

for classiﬁcation, linking, consistency checking and

visualization which can be ﬂexibly combined, as de-

scribed in Section 2.

The data model underlying the implementation of

hybrid wikis is shown in Figure 3. The illustrated

concepts are currently mapped to a SQL database

means of Tricia’s object relational mapping mecha-

nisms. We will explain the data model of the hybrid

wiki elements and some impacts on the system be-

haviour in the following.

For each wiki page multiple attributes can be as-

signed. Each of the attributes has a non-empty list of

ordered values being either string- or link-values. A

wiki page can have multiple tags assigned. A tag is a

simple character sequence and either a normal tag or a

type tag.

WikiPage

TypeTag

Attribute

and

Value

constitute the set of concepts which are provided for

structuring data in hybrid wikis.

TypeTagDefinition

AttributeDefinition

SQL is the default storage implementation, but other

implementations are also supported, e.g., NoSQL storages

and

Validator

are additional concepts to enable

users to specify integrity constraints. A type tag def-

inition contains the set of attributes that the users are

urged to provide in combination with the respective

type tag. These attribute deﬁnitions can be further

used to restrict the range of attribute values with val-

idators.

A type tag deﬁnition is loosely coupled to a (type)

tag by name, the same applies to attribute and attribute

deﬁnition. The fact that an attribute deﬁnition x is

bound to a type tag deﬁnition y means that on pages

with a type tag y an attribute suggestion x is shown,

which takes precedence over the default attribute sug-

gestion mechanism, i.e. it is not required that a page

having both, tag x and attribute y, assigned exists in

the system. Thus, new attributes can be seeded top-

down over typed pages by means of suggestions.

Furthermore, for particular attribute deﬁnitions,

integrity constraints can be speciﬁed using validators.

The system currently supports three kinds of valida-

tors: the

EnumerationValidator

checks if the val-

ues of an attribute conform to a given set of string-

and link-values. The

HyperlinkValidator

checks

if the values of an attribute are hyperlinks, addi-

tionally it can be deﬁned that these hyperlinks have

to be links to a wiki page with speciﬁc type tags.

MultiplicityValidator

allows to specify how

many values an attribute should have, e.g., at-least-

one, at-most-one, exactly-one.

Validators apply when a wiki page is displayed.

That means, if a wiki page has attribute x, a type tag

y, and a type tag deﬁnition for y exists having an at-

tribute deﬁnition with name x and some validators,

these validators will be matched with the current val-

ues of attribute x. In case of validation failure the

validationMessage

s of the validators are shown in

the context of attribute x. The situation that two type

tags specifying contradicting validarors are assigned

to a wiki page is currently not prevented by the sys-

tem and users have to solve this conﬂict manually.

Beside showing validation messages for wiki page

attributes, the integrity constraints are also used to im-

prove the input support for values when an attribute is

edited. For instance, if an enumeration validator ap-

plies to an attribute in the value auto-completion con-

trol only the elements speciﬁed in this validator (e.g.,

{‘open’, ‘closed’}) are offered.

Since hybrid wiki validators never prevent users

from entering values that contradict the speciﬁed val-

idation rules, we call them soft validators. The sys-

tem assists users in systematically searching for con-

tradictions. As a side-effect, it is possible to import

information from arbitrary data sources to Tricia, e.g.,

from a Microsoft Excel data sheet, without being re-

HYBRID WIKIS: EMPOWERING USERS TO COLLABORATIVELY STRUCTURE INFORMATION

255

WikiPage

...

Attribute

name:String

{unique for wikipage}

LinkValue

link:String

Value

StringValue

text:String

attributes

wikipage

1 values

1..* {ordered}

Wiki

...

pages

TypeTag

name:String

tags

TypeTagDefinition

name:String {unique for wiki}

AttributeDefinition

name:String

{unique for typeTagDefinition}

Validator

validationMessage:String

...ValidatorMultiplicityValidator

validators

* 0..1

Tag.name

conforms

TypeTagDefinition.name

Attribute.name

conforms

AttributeDefinition.name

validates

wiki

definitions

typeTagDefinition

attributes

Figure 3: The Data model of Hybrid Wikis.

strained by hard integrity constraints. Any conﬂicts

with the schema can be ﬁxed after the import.

Since the main purpose of Tricia is to enable

users to ﬁnd relevant information within enterprises

quickly, the content of all elements (e.g ﬁles, wikis,

blogs) is indexed by means of the Lucene information

retrieval software library

. Beside simple text queries,

querying for metadata like tags or date of last modi-

ﬁcation is supported in this way. For the implemen-

tation of hybrid wikis Lucene plays a critical role in

the dynamic generation of attribute suggestions and

overview tables. Lucene supports both features by

providing ﬂexible ﬁltering capabilites together with

a very fast access to particular ﬁelds of the indexed

entities. For example to display a table of all pages

having a speciﬁc type tag, Lucene can efﬁciently de-

termine the respective set of pages (ﬁltered by the ac-

cess rights of the user) and provide the set of attribute

names used on each page. From the analysis of the

frequency of attribute names the selection and order

of the table columns is then determined.

It is possible to import and export hybrid wiki data

by using the standardized Eclipse Model Framework

(EMF) exchange format. Models can be imported and

exported as ecore-ﬁles, model instances as xmi-ﬁles.

Furthermore, all hybrid wiki contents can be imported

from and exported to spreadsheets (e.g., Microsoft

Excel). Additionally, it is possible to visualize hybrid

wiki content by means of an external service (System

http://lucene.apache.org; visited on February 1st 2011.

Cartography Service

) that is accessed via RESTful

API calls.

4 PRACTICAL EXPERIENCES

Hybrid wikis are currently evaluated in several in-

dustrial and research projects. In the following, we

brieﬂy describe the usage scenarios of the hybrid wiki

technology in two selected cases. The ﬁrst is taken

from one of our research projects Wiki4EAM

where

hybrid wikis are applied for the collaborative docu-

mentation of system landscapes in enterprises in or-

der to evolve models representing theses landscapes

bottom-up, the second from the application of a hy-

brid wiki as a simple issue tracker in a German soft-

ware development company, the InfoAsset AG

. For

each of these cases we sketch the models being de-

veloped bottom-up via attribute and type suggestions

as well as the (soft-) constraints which were deﬁned

afterwards.

http://wwwmatthes.in.tum.de/wikis/sebis/syca; visited

on February 1st 2011.

http://wwwmatthes.in.tum.de/wikis/sebis/wiki4eam;

visited on February 1st 2011.

http://www.infoasset.de; visited on February 1st 2011.

ICSOFT 2011 - 6th International Conference on Software and Data Technologies

256

4.1 Wiki4EAM Community

In the Wiki4EAM community, founded in December

2010 at TUM, members share their experiences re-

garding the use of a hybrid wiki in the context of

enterprise architecture management (EAM). A pre-

requisite for adequate management of the enterprise

architecture (EA) is to capture its current state in a

model. Since the knowledge of all different elements

to be considered in the model (e.g., business pro-

cesses, applications, organizational units) is spread

over all the different stakeholders in the company, the

documentation of the current state of the EA remains

a challenging task. Hybrid wikis enables these stake-

holder to document the particular parts of the EA they

are responsible for by means of (hybrid) wiki pages.

Thereby the data model emerges bottom-up by cre-

ating, editing, linking and structuring these particu-

lar pages. A preliminary data analysis in two Ger-

man companies from December 2010 to January 2011

showed the following model evolution:

• Company A: 12 concepts with 63 attributes (tex-

tual and links), 100 wiki pages, 2 (soft-) con-

straints were created within two weeks by four

participating stakeholders (editors)

• Company B: 18 concepts with 60 attributes (tex-

tual and links), 120 wiki pages, 28 (soft-) con-

straints were created within one month by ﬁve

participating stakeholders (editors)

Although these numbers are not a founded empir-

ical evaluation, they allow to assume that in a rela-

tively short period models can emerge bottom-up by

the collaborative documentation of the particular ele-

ments by using lightly structured wiki pages. Indeed,

feedback from Wiki4EAM community members in-

dicates that a hybrid wiki is well suited in cases when

the target model is not completely known. They also

conﬁrmed that attribute suggestions facilitate the evo-

lution of the information model.

4.2 InfoAsset Bugtracker

The InfoAsset AG providing the Tricia platform for

commercial use, uses hybrid wikis among others

things for its issue management. For each issue a hy-

brid wiki page with tag ‘unprocessed’ and type tag

‘issue’ is created. All unprocessed issues are shown

in a list on the Tricia developer dashboard, that is built

only using standard features of hybrid wikis. This

list is dynamically generated by using an embedded

query which ﬁlters all wiki pages having the tag ‘un-

processed’ and type tag ‘issue’ assigned. When a new

entry appears in this list, i.e. the underlying embed-

ded query yields a new search hit, the developers are

informed by means of a new entry in an RSS-Feed.

A responsible person processes these new entries and

categorizes them by assigning additional type tags.

Currently two kinds of additional type tags are mainly

used: ‘Bug’ and ‘Change request’. Furthermore, a

person is assigned responsibility for processing this

issue by setting the attribute ‘assigned to’ to the re-

spective name and the tag ‘unprocessed’ is deleted

in order to remove this issue from the dashboard list.

The responsible person is in turn informed by an RSS-

Feed. The complete data model as used for the change

management is shown as an UML class diagram in

Figure 3.

The schema as shown emerged bottom-up by

management of the particular wiki pages. Only the

following constraints were deﬁned afterwards top-

down by means of type tag deﬁnitions:

• enumeration types (Status, Priority, Effort), in the

beginning these values were plain strings

• ‘exactly-one’ cardinality for the relationship ‘as-

signed to’ between ‘Issue’ and ‘Person’ instances

Note, that the cardinality in the UML diagram of

Figure 3 is given with 0..1. This is due to the fact that

new reported issues do not have a responsible person

assigned. For theses instances a warning message is

shown indicating that there should be at least one per-

son deﬁned.

5 RELATED WORK

Wiki templates enable authors to reuse content struc-

tures among wiki pages. In (Haake et al., 2005), the

authors discuss the need for structure in wikis based

on templates, in (Di Iorio et al., 2008) different wiki

templating approaches are compared to each other.

Although hybrid wikis explicitly do not support tem-

plates, attribute and type (-tag) suggestions based on

a statistical analysis of typetag and attribute combi-

nations enables authors to reuse well established type

structures, similar to templates.

The Semantic MediaWiki project is the most

prominent example in the category of semantic wikis.

This project adds database-like structuring and query-

ing capabilities on top of an existing wiki, with-

out requiring users to develop or adhere to a rigid

database schema when authoring content (Kr¨otzsch

et al., 2006). This project also includes various fea-

tures for browsing, searching, and aggregating the

wiki’s content as well as embedding queries in the

wiki pages. Additionally, Semantic Media Wiki

provides a model export to the standardized format

HYBRID WIKIS: EMPOWERING USERS TO COLLABORATIVELY STRUCTURE INFORMATION

257

Person

Feature Request

due date : Date

Issue

status : Status

priority : Priority

effort: Effort

related to

0..1

Bug

known since: Date

duplicate of

0..1

assigned to

0..1

reported by

«enum»

Status

open, in progress, on hold,

to be discussed, solved,

postponed, rejected

«enum»

Priority

low, medium, high

«enum»

Effort

low, medium, high

requested by

Refactoring

related to

0..1

Figure 4: Emerged Concepts of the InfoAsset Bugtracker.

OWL/RDF. Unlike in hybrid wikis, in Semantic Me-

diaWiki meta information can only be added to the

pages by directly editing the markup in wiki syntax.

This issue is addressed by Semantic Enterprise

Wiki (SWM+)

, a set of open-source extensions to

the Semantic MediaWiki. In this approach, meta data

(properties) can be deﬁned by means of a graphical

user interface, called ‘semantic toolbar’, so users are

not forced to write semantic annotations manually.

For the classiﬁcation of pages in Semantic Enterprise

Wiki the concept of categories is used. Categories and

properties are shown as a list in the semantic toolbar.

The approach differs from hybrid wikis as follows:

• A separate annotation mode (semantic toolbar) is

needed to enter structured information.

• Users have to create and edit both, the content

of the wiki page and the semantic annotations.

Users also have to be aware of content and anno-

tations being synchronized, which is an laborious

and error-prone task.

• No attribute (properties) suggestions based on the

types (categories) are provided.

AceWiki is a semantic wiki using controlled natu-

ral language for ontology management (Kuhn, 2008).

In (Paschke et al., 2009) a lightweight approach for

editing ontologies is introduced. Like hybrid wikis,

both approaches try to facilitate structured data entry.

However, in AceWiki this is achieved by natural lan-

guage processing, in the second case by introducing a

lightweight ontology editor.

The open source project TWiki

and its fork Fos-

wiki

try to combine the advantages of wikis and

database systems by allowing the user to attach data

records to wiki pages. Although the schema can be

changed at runtime, the effort is considerably higher

http://wiki.ontoprise.de; visited on April 27th 2011.

http://twiki.org; visited on April 27th 2011.

http://foswiki.org; visited on April 27th 2011.

than for hybrid wikis and the usage of a special wiki

syntax is required.

The research prototype SnoopyDB (Gassler et al.,

2010) focuses particularly on recommending at-

tributes to users when they are entering data. Types

cannot be speciﬁed at all but all attributes are sug-

gested based on attributes of other pages. This makes

it very difﬁcult to generate tabular overviews of pages

covering certain kinds of things.

MokiWiki (Ghidini et al., 2009) is a plain mod-

elling wiki, used to model the constituent parts of an

enterprise collaboratively. With hybrid wikis we fo-

cus on facilitating collaborative data and information

management within enterprises. Information models

emerge bottom-up as a by-product.

In sebis Enterprise 2.0 Tool Survey (B¨uchner

et al., 2009) the functional capabilities of integrated

E2.0 platforms are compared resulting in a multi-

dimensional classiﬁcation and evaluation framework.

Especially the survey 2010 includes functional as-

pects regarding the ‘Structuring of Content’ as well

as ‘Templates for Structured Content’. The fact that

leading social software vendors include capabilities

for structuring contents indicates an increasing de-

mand in enterprises.

6 SUMMARY & OUTLOOK

In this article, we introduced hybrid wikis, a

lightweight approach for data and information man-

agement within enterprises facilitating structuring of

contents for business users. The main purpose of

hybrid wikis is to avoid requiring users to learn

complex semantic languages for adding structure to

While hybrid wikis are not mainly built for modelling

purposes, the models emerging through data management

in hybrid wikis can also be used to represent the particular

enterprise constituents (c.f. (Buckl et al., 2009) and Section

4).

ICSOFT 2011 - 6th International Conference on Software and Data Technologies

258

wiki pages. Instead we provide an easy way for

structured data entry with a small set of elements,

namely attributes (key-value-pairs) and type tags.

Furthermore, attribute suggestions encourage users

to provide these structuring elements. Suggestions

and auto-completion foster a common vocabulary, so

users immediately beneﬁt from the structure through

better search capabilities. By means of these mecha-

nisms complex structures emerge bottom-up by the

management of particular data sets. To deﬁne in-

tegrity constraints we introduced the concepts of val-

idators, attribute deﬁnitions and type tag deﬁnitions.

They additionally can be used to deﬁne advanced

rules and parts of the data model in a top-down man-

ner.

Furthermore, we described the practical expe-

riences made with hybrid wikis in two concrete

projects, one from the industry and one as part of

our research endeavours. For the latter we sketched

the models emerging bottom-up by means of attribute

suggestions as well as the (soft-) constraints addition-

ally deﬁned top-down afterwards.

We see potential to improve and validate our ap-

proach in the following ways:

• Compare hybrid wikis to similar wiki-based ap-

proaches (c.f. Section 5). For instance, we

will analyze the revision history in order to com-

pare the evolution of structure compared to hybrid

wikis.

• Provide more validators (e.g., boolean, date),

which are the basis for advanced controls facili-

tating the input of structured data.

• Support advanced visualizations of structured

data, which also can be used for navigating the

wiki contents.

• Improve the way relationships between pages are

handled (c.f. Section 2.3).

• Provide type (tag) suggestions beside attribute

suggestions.

REFERENCES

B¨uchner, T., Matthes, F., and Neubert, C. (2009). A concept

and service based analysis of commercial and open

source enterprise 2.0 tools. In KMIS, pages 37–45.

B¨uchner, T., Matthes, F., and Neubert, C. (2010). Data

model driven implementation of web cooperation sys-

tems with Tricia. In Dearle, A. and Zicari, R., editors,

Objects and Databases, Lecture Notes in Computer

Science, pages 70–84. Springer, Berlin, Heidelberg,

Germany.

Buckl, S., Matthes, F., Neubert, C., and Schweda, C. M.

(2009). A wiki-based approach to enterprise archi-

tecture documentation and analysis. In The 17

Eu-

ropean Conference on Information Systems (ECIS) –

Information Systems in a Globalizing World: Chal-

lenges, Ethics and Practices, 8. – 10. June 2009,

Verona, Italy, pages 2192–2204, Verona, Italy.

Di Iorio, A., Vitali, F., and Zacchiroli, S. (2008). Wiki con-

tent templating. In WWW ’08: Proceeding of the 17th

international conference on World Wide Web, pages

615–624, New York, NY, USA. ACM.

Edmunds, A. and Morris, A. (2000). The problem of infor-

mation overload in business organisations: a review

of the literature. International Journal of Information

Management, 20(1):17 – 28.

Gassler, W., Zangerle, E., Tschuggnall, M., and Specht, G.

(2010). Snoopydb: narrowing the gap between struc-

tured and unstructured information using recommen-

dations. In Proceedings of the 21st ACM conference

on Hypertext and hypermedia, HT ’10, pages 271–

272, New York, NY, USA. ACM.

Ghidini, C., Kump, B., Lindstaedt, S., Mahbub, N., Pam-

mer, V., Rospocher, M., and Seraﬁni, L. (2009).

MoKi: The Enterprise Modelling Wiki. In Aroyo,

L., Traverso, P., Ciravegna, F., Cimiano, P., Heath, T.,

Hyv¨onen, E., Mizoguchi, R., Oren, E., Sabou, M., and

Simperl, E., editors, The Semantic Web: Research and

Applications, volume 5554, chapter 65, pages 831–

835. Springer Berlin Heidelberg, Berlin, Heidelberg.

Haake, A., Lukosch, S., and Sch¨ummer, T. (2005). Wiki-

templates: adding structure support to wikis on de-

mand. In WikiSym ’05: Proceedings of the 2005 in-

ternational symposium on Wikis, pages 41–51, New

York, NY, USA. ACM Press.

Kr¨otzsch, M., Vrandecic, D., and V¨olkel, M. (2006). Se-

mantic mediawiki. In Cruz, I., Decker, S., Allemang,

D., Preist, C., Schwabe, D., Mika, P., Uschold, M.,

and Aroyo, L., editors, The Semantic Web - ISWC

2006, volume 4273 of Lecture Notes in Computer Sci-

ence, pages 935–942. Springer Berlin / Heidelberg.

Kuhn, T. (2008). Acewiki: A natural and expressive se-

mantic wiki. Technical Report arXiv:0807.4618, De-

partment of Informatics University of Zurich. Com-

ments: To be published as: Proceedings of Seman-

tic Web User Interaction at CHI 2008: Exploring HCI

Challenges, CEUR Workshop Proceedings.

Paschke, A., Coskun, G., Luczak-R¨osch, M., Oldakowski,

R., Harasic, M., Heese, R., Sch¨afermeier, R., and

Streibel, O. (2009). Realizing the corporate semantic

web: Concept paper. Technical report, Freie Univer-

sit¨at Berlin, Berlin.

Stocker, A. and Tochtermann, K. (2009). Exploring the

value of enterprise wikis: A multiple-case study. In

KMIS.

HYBRID WIKIS: EMPOWERING USERS TO COLLABORATIVELY STRUCTURE INFORMATION

259