WordVenture – COOPERATIVE WordNet EDITOR

Architecture for Lexical Semantic Acquisition

Julian Szyma´nski

Gda´nsk University of Technology, Narutowicza 11/12, 80-952 Gda´nsk, Poland

Keywords:

Linguistic semantic networks, WordNet, Collaborative editing, Lexical semantics, Wikipedia.

Abstract:

This article presents architecture for acquiring lexical semantics in a collaborative approach paradigm. The

system enables functionality for editing semantic networks in a wikipedia-like style. The core of the system

is a user-friendly interface based on interactive graph navigation. It has been used for semantic network

presentation, and brings simultaneously modiﬁcation functionality.

1 INTRODUCTION

WordNet (Miller et al., 1990b) is one of the largest

semantic lexicons of English. It has been developed

since 1985 by the Cognitive Science Laboratory at

Princeton University. Its authors, based on theories

of human cognition, try to reﬂect all linguistic de-

pendencies between concepts in a common lexical

database. The WordNet team has been working on

a semantic dictionary for over 22 years. Nowadays

the dictionary contains about 155287 words, orga-

nized in 117659 synsets (meaning representations),

and includes 206941 pair words – meaning. Intro-

duction of all words with their connections, as well

as examples of their usage in language, requires a lot

of human work, however the WordNet team has only

seven members. The WordNet project has been sup-

ported by plenty of grants, which brought together

3 millions dollars. Currently the third release of the

WordNet lexical database is available at the project

website

. WordNet develops as a research project in

a closed academic environment. The ﬁrst version of

the dictionary appeared in 1993, and now a third ver-

sion is available. The dictionary is publicly available,

but its modiﬁcation is restricted to internauts. Prob-

ably, the reason for that, is the fact that the lexicon

is organized as a set of text ﬁles in a speciﬁc format,

which makes it hard to apply cooperativeapproachfor

WordNet development. Lack of cooperative editing

functionality is the biggest barrier to scale-up seman-

tic database.

ver. 3.0

http://wordnet.princeton.edu

The most well known application of a coopera-

tive approach for gathering data is Wikipedia. The

project has experienced great interest from the In-

ternet community which brought many positive re-

sults. Wikipedia has been developed since 2001 by

volunteers from all over the world. Currently, the

Wikipedia initiative is supported by almost 75000

people, working on over nine million articles written

in 125 languages. The largest set of articles is avail-

able in English, and contains over 2 million articles.

Nowadays, a lot of projects has been created on

the basis of WordNet

. They use semantic dictio-

nary as a core knowledge base about language, what

enables to implement elementary linguistic compe-

tences in a machines.

Some of the implementations do the mapping

from WordNet ﬁles to other models, especially rela-

tional. This can be used to enable a cooperative edit-

ing approach.

2 DESCRIPTION OF WordVenture

SYSTEM

A WordVenture portal

has been developed at the

Gdansk University of Technology at the Faculty

of Electronics, Telecommunications and Informat-

ics. It provides mechanisms for simultaneous work

on lexical dictionaries for distributed groups of peo-

ple and enables cooperative work on aWordNet lex-

see: related projects http://wordnet.princeton.edu/links

http://wordventure.eti.pg.gda.pl

Szyma

nski J. (2009).

WordVenture – COOPERATIVE WordNet EDITOR - Architecture for Lexical Semantic Acquisition.

In Proceedings of the International Conference on Knowledge Engineering and Ontology Development, pages 28-33

DOI: 10.5220/0002295900280033

 SciTePress

ical database. The Cognitive Science Laboratory

approach to WordNet development required huge

amounts of resources e.g human, time, money (Miller

et al., 1990a). With WordVenture, lexical database

development becomes common and cheap.

With WordVenture, a user can browse a Word-

Net dictionary, and display its content on the screen

with a graphical user interface based on an interac-

tive graph. It gives a user-friendly way for visualizing

very large sets of contextual data. A user can also

query WordVenture to ﬁnd a speciﬁed word and dis-

play its senses and related concepts. Connections be-

tween nodes (words or senses) are illustrated as edges

of a given type. To keep graphs clear, a user can set

some constraints to visualize only required types of

data. There is also the possibility of interactive graph

traversing. Selecting one node all elements that are

connected with the marked one are displayed (accord-

ing to given constraints on data selection).

The advantage brought to WordNet development

by WordVenture system is a possibility of editing

semantic database by the open, Internet community,

which fasten lexical data acquisition process. To pro-

vide high quality of the acquired data, all changes in-

troduced by users are represented as change proposi-

tions, which are approved or rejected by a privileged

user – moderator.

3 SYSTEM ARCHITECTURE

It was decided that the WordVenture system will be

implemented in client-server architecture, with the

following assumptions:

• WordNet database and data access logic resides

on the server,

• Data visualization mechanisms reside at the client

side and provide interfaces to the lexical database

in the form of interactive graphs.

This architecture obliges a developer to imple-

ment some functionalities at the server side of the ap-

plication, but imposes some limitations, especially to

communication. The developer has to deﬁne commu-

nication protocol which will assure ﬂexibility of the

data interchange. To widen client-server architecture

some elements of the Service Oriented Architecture

(SOA) (Erl, 2005) has been introduced. One must

meet the following expectations to efﬁciently imple-

ment SOA:

• Communication Interoperability – must be as-

sured between different systems and different pro-

gramming languages. A well-known example of

such, is message passing oriented communication

Figure 1: Basic concept of the WordVenture architecture

and its elements.

(Palmer et al., 2006). Messages in a deﬁned for-

mat are sent between sender and receiver, who

performs content-based computations. Neither re-

ceiver nor sender have to have precise knowledge

about the other’s side of a platform.

• Publishing, Discovery and Service Inquiry –

these are basic concepts of SOA architecture. The

three operating sides can be distinguished: Ser-

vice Provider (creates and publishes his services

– service producer), Service Broker (gives mech-

anisms to store information about services e.g.

physical location of remote service and performs

search operations), and Service Requester (in-

vokes remote service – service consumer).

One of the most popular implementations of SOA

are web services. They’ve met above requirements,

especially communication interoperability between

many development platforms e.g. J2EE and Mi-

crosoft .NET. Every web service is described in a

well-deﬁned and common language – WSDL (Web

Service Description Language) (Christensen et al.,

2001) and it uses SOAP (Simple Object Access Pro-

tocol) as a transport protocol (Scribner et al., 2000).

In SOAP, messages are passed as XML documents.

The original implementation of a WordNet

database uses text ﬁles. Because of their structure,

modiﬁcation is available only with dedicated tools.

This type of storage doesn’t support synchronous ac-

cess for modiﬁcation, nor allows to perform efﬁ-

ciently large amount of queries. It also requires us to

create special mechanisms for editing, including syn-

chronization and ﬁle structure refactoring, after any

operation. To enable editing of a WordNet lexical

database we had to perform mappings between Word-

Net text ﬁles and a relational database.

Transformation from text ﬁles to its relational

representation was performed by the WordNet SQL

Builder tool

http://wnsqlbuilder.sourceforge.net

WordVenture - COOPERATIVE WordNet EDITOR - Architecture for Lexical Semantic Acquisition

3.1 Server-side Architecture

The server-side of the WordVenture application

makes its functionalities available through web ser-

vices. According to communication interoperability

requirement, it is possible to connect client applica-

tion that can be implemented in different technolo-

gies. Web services have been developed and deployed

with the Apache Axis framework

, which resides in

the servlets container – Apache Tomcat

. Apache

Axis framework is a set of libraries and tools which

allows a developer to create and publish web ser-

vices. Axis is just an ordinary web application that

can be deployed to any servlet container, especially

to Apache Tomcat. It listens for a request from client

application that is sent as a SOAP envelope. When

a message comes, Axis interprets it and calls a local

procedure. Subsequently, a response message is cre-

ated and sent to the client application. Apache Axis

allows programmers to deploy web services such as

Plain Old Java Objects (POJOs), which have to have

changed extension (from .java to .jws). Deployment

can be done by copying jws (Java Web Service) ﬁle to

proper Axis directory.

The second edition of WordVenture (Szyma´nski,

J. and Dusza, K. and Byczkowski, Ł., 2007) intro-

duce mechanisms which allows an moderator to con-

trol user actions. To control modiﬁcations of lexi-

cal database authentication and authorization mech-

anisms have been created. An anonymous user can

only browse the data from WordNet database, if he

wants to edit it, he must log in. Every modiﬁcation in-

troduced to database is represented as ,,change propo-

sition” and is sent to the moderator. This moderator,

as a privileged user, can commit or reject every mod-

iﬁcation proposed by an ordinary user.

Every server functionality allows a user to per-

form three different groups of actions depending on

the role that user has:

• Functionalities for browsing WordNet Lexical

Database – are available to every user (anony-

mous and logged-in). After invoking an action

on the client-side of application, a proper remote

procedure is called on server. The server queries

database and sends data to client application. Be-

cause of efﬁciency reasons, all data that is sent be-

tween sender and receiver is serialized and com-

pressed, so transmission through the Internet is

much more faster.

• Functionalities that allows a User to edit Word-

Net Lexical Database – are only available to

http://ws.apache.org/axis

http://tomcat.apache.org

Figure 2: Main view of server-side architecture.

registered users. After invoking an edit action

on the client-side of an application, the proper

change proposition is created. Subsequently, this

proposition is sent to the server to be added to

database. A privileged user (moderator) can view

all change propositions and select commit, other

cancel. After committing, a proposition is perma-

nently added to database and can be seen by other

users.

• Administrative Functionalities which are Con-

nected With user Management – are available

only for privileged users – administrators. They

can perform user deletion or user rights editing

in WordVenture system. Every administrator can

give administrative rights to another user.

It was decided to use an object-relational mapping

mechanism to make our system reusable. Almost all

of the object-relational mapping engines use DAO ob-

jects (Data Access Object). This project pattern al-

lows the developer to separate data access logic from

logic operating on those data. Object-relational map-

ping mechanisms enable developer to translate data,

from relational structure to object structure, what

keep proper relations. Each row in a table is translated

to a proper object. We’ve use Hibernate as O/R map-

ping engine

what is a highly-developed and effective

solution. One of its main features is the ”lazy loading”

mechanism (Arendt et al., 1998). It prevents from re-

trieving at one time all data from database (which can

be very inefﬁcient). Lazily loaded mechanisms get

data from database only when a end-user wants to see

it.

http://www.hibernate.org

KEOD 2009 - International Conference on Knowledge Engineering and Ontology Development

Figure 3: WordNet entities supported by the tool. Grayed

out entities have support for both visualization and editing,

white entities have only visualization support. Arrows rep-

resent relationships between entities.

The Figure 2 presents a detailed diagram of

server side architecture, which includes all the above-

mentioned technological solutions. It shows how the

server handles SOAP messages sent by client applica-

tion. Web service invocation starts when a SOAP en-

velope comes to the server. Apache Axis framework,

resides on Apache Tomcat servlet container, and is

responsible for handling SOAP messages. Creating a

new web service in Java, from the developers’ point

of view, requires programming public class with pub-

lic methods and deploying it to Apache Axis. In a

WordVenture system those public classes are used to

exploit the Hibernate O/R mapping engine to access

database and perform all required queries.

The elements of the WordNet like a word position

or morphological deﬁnitions are not as much neces-

sary as lemmas and synsets. To simplify the editing

process, it was decided to allow only for modiﬁcation

of the semantic net structure. The database structure

for handling data provided by WordVenture is pre-

sented in the Figure 3, where editable and dictionary

tables of the system are shown.

3.2 Client-side Architecture

WordVenture has been developed in rich-client archi-

tecture (Boudreau et al., 2007). Because of that, some

logic connected with data visualization, can be exe-

cuted on the client-side of application. Due to ease-

of-use requirement, it was decided that client applica-

tion will be developed as a J2SE 5 applet. The client

is a modiﬁed TouchGraph component

for interactive

graph visualization, where graph elements represent

WordNet entities. The applet allows a user to:

• Browse WordNet Lexical Database – this mech-

anism is based on a modiﬁed TouchGraph engine.

It enables the user to navigate over the WordNet

semantic network in a user-friendly way. Words

http://www.touchgraph.com

and synsets are visualized as graph nodes, connec-

tions between them are visualized as graph edges.

Additionally, the user can ﬁlter graph nodes and

edges to obtain required content (according to a

selected type), what makes user interface clean

and readable.

• Perform Modiﬁcations on WordNet Lexical

Database – the tool enables a user to change

graph content by adding, editing, or deleting its el-

ements: nodes and edges. Modiﬁcation of above-

mentioned elements of WordNet lexicon (see Fig-

ure 3) does not cover all components of WordNet.

It only covers the four most desired, from the user

point of view, elements of the semantic network:

words, synsets, senses and relations, presented in

the Figure 3.

Modiﬁcation of WordNet lexicon is based on

well-known rules from other cooperative projects like

Wikipedia (Viegas et al., 2007):

• Changes Patrolling – every modiﬁcation of

WordNet lexicon is represented as a change

proposition that is sent to a privileged user – mod-

erator, who can commit or reject the proposi-

tion. This approach is used to trace every ac-

tivity performed by the cooperative community.

Such a mechanism can be used to detect unde-

sirable users’ activities: vandalism, violation of

copyrights and others.

• ”Free” Character of Wikipedia – every inter-

ested user can join the WordVenture community

and cooperate with its creation.

In the WordVenture system a user is able to use

the context menu which is available under right click

of mouse. Selecting a word or synset makes the sys-

tem show options available to choose. Functionality

of WordNet lexicon editing in cooperative paradigm

(Yang et al., 2000) is available only for a logged-in

user. In previous release of the system (Szyma´nski,

J. and Dusza, K. and Byczkowski, Ł., 2007) syn-

chronous work of many users caused saving only of

the last modiﬁcation. From now on, every modiﬁca-

tion is saved as a change proposition, and is sent to

an moderator. He can choose whether a proposition is

permanently saved, or deleted.

Graph-based visualization in a WordVenture sys-

tem allows a user to work efﬁciently, and keep

clean and readable a large amount of lexical data.

In every moment a user can enable or disable re-

quired elements of the visualization, which makes his

workspace personalized. Additionally, it is possible

to zoom in or zoom out view of graph, so a user is

able to keep a lot of graph nodes on his workspace.

WordVenture - COOPERATIVE WordNet EDITOR - Architecture for Lexical Semantic Acquisition

4 COOPERATIVE APPROACH

TO BUILDING LEXICAL NETS

Lack of tools for cooperative editing of semantic dic-

tionary databases is the main barrier for rapid Word-

Net development. Our mission is to deliver a tool en-

abling a cooperative editing approach for many users

placed in distributed Internet environment. Coopera-

tive editing is connected with publishing the WordNet

database and making it open to the Internet commu-

nity. This brings advantages for faster WordNet de-

velopment, however some problems may arise:

• Vandalism – may cause loss of all important data,

kept in current release of lexical database. It also

can affect the data structure e.g. creating pointless

connections between words and senses. Because

of that, it is important to deliver tools which will

reduce the risk of the above-mentioned.

• Simultaneous work on the same part of database,

by many users, may reveal some conﬂicts result-

ing from concurrent work of many users at the

same time. In the worst case, one user can add

connection to an element of the WordNet dictio-

nary that was deleted by another.

The best solution of these above-mentioned prob-

lems is to introduce the role of privileged user – mod-

erator. He or she is able to see every change that is

proposed to the lexical database dictionary. Every

user, after logging in, can edit the lexical database

in a restricted way. All introduced modiﬁcations are

represented as change propositions that are sent to an

moderator, who can browse them, and decide whether

propositions can be added to database or deleted. All

administrative actions result in permanent semantic

network update. This approach will also allow us to

save history of the database modiﬁcations and to de-

tect users – vandals, whose rights can be permanently

taken back. According to the basic rule of effective

team work (,,communication, coordination and coop-

eration”) (Kling, 1991), the users were delivered the

possibility of continuous communication via a web-

based forum. While using it, users can deﬁne their

own strategies for WordNet lexical database develop-

ment, reach their own conclusions and also feel all

advantages of synergic effect.

The current release of WordVenture sys-

tem includes all above-mentioned function-

alities and is available on project web site:

http://wordventure.eti.pg.gda.pl. At present, we

are evaluating future proposals for the system,

gathering more feedback from users via our web-

based forum system, prioritizing future goals, and

evaluating the applied solution as a base for a generic

approach to semantic data editing tasks.

5 CONCLUSIONS

Our project has been developed and successfully de-

ployed. Currently the WordVenture system has been

extended to introduce mechanisms avoiding problems

connected with cooperative editing approach:

• Authentication, authorization, and logging users

activity – while switching to editing mode, a user

is asked to ﬁll-in an authentication form. After

logging-in, a user is able to create change propo-

sitions that are sent to an moderator. He can trace

all the changes and decide whether to approve or

reject them.

• Tracing users’ change proposals – a privileged

user has rights to manage change propositions in-

troduced by other users. Because of that, it is pos-

sible to avoid all unwanted changes, and also to

review all proposals by a qualiﬁed person.

The WordVenture system starts from the newest

version of WordNet lexical database (3.0). The sys-

tem architecture allows us to perform trouble-free ac-

tualization of dictionary version with assumption that

data structure will not change.

Offering the cooperative editing of the dictionary

for the Internet community, seems to be a very attrac-

tive way for gathering lexical semantics. It creates

the opportunity for fast semantic dictionary develop-

ment with the cooperation of people from all over the

world. It takes down all the duties put on the team

while creating the next versions of WordNet dictio-

nary as well. However, we should remember about

potential threats which can arise while opening the

dictionary for the wide Internet community. The sys-

tem have been developed based on the experience of

Wikipedia. In the current version of the system the

risk of vandalism or the unintentional destruction of

content has been eliminated, which makes a coopera-

tive approach more reliable.

6 FUTURE MISSION

The WordVenture system has reached the end of its

second iteration. In this section we propose changes

that will be applied in next versions of WordVenture.

Next iteration can include improvements as follows:

1. Internationalization of client-side application – all

inscriptions should be organized as supply and

should be translated.

KEOD 2009 - International Conference on Knowledge Engineering and Ontology Development

2. Integration other lexical networks to WordVenture

to make linguistic database richer. By now we

consider two projects: Microsoft MindNet (Van-

derwende et al., 2005), and ConceptNet (Liu and

Singh, 2004)

3. Extension of user activities tracking and edit func-

tionalities to other elements of WordNet database.

4. Introducing improvements to the user interface,

especially to administrative part. Currently, an

moderator has to manually merge some changes

entered by user.

5. Extending the search engine: search by keywords

in synset descriptions, etc.

6. Adaptation of user interface to a new

displaying engine, which performs

more efﬁcient graph visualization, see

http://visualwiki.eti.pg.gda.pl/wikiparser.

7. Integration of Wikipedia and WordVenture se-

mantic network.

Future development of WordVenture depends also

on users’ opinions received via our web based fo-

rum. We are waiting for any suggestions and com-

ments about WordVenture – development ideas are

welcome. We want to invite everyone to use our sys-

tem and give us feedback.

ACKNOWLEDGEMENTS

This work was supported by the Polish Ministry of

Science and Higher Education under research project

N516 035 31/3499.

REFERENCES

Arendt, J., Giangarra, P., Manikundalam, R., Padgett, D.,

and Phelan, J. (1998). System and method for lazy

loading of shared libraries. US Patent 5,708,811.

Boudreau, T., Tulach, J., and Wielenga, G. (2007). Rich

client programming: plugging into the netbeans plat-

form.

Christensen, E., Curbera, F., Meredith, G., and Weer-

awarana, S. (2001). Web services description lan-

guage (WSDL). W3C Web Site.

Erl, T. (2005). Service-oriented architecture: concepts,

technology, and design. Prentice Hall PTR Upper Sad-

dle River, NJ, USA.

Kling, R. (1991). Cooperation, coordination and control

in computer-supported work. Communications of the

ACM, 34(12):83–88.

Liu, H. and Singh, P. (2004). ConceptNet?a practical com-

monsense reasoning tool-kit. BT Technology Journal,

22(4):211–226.

Miller, G., Beckwith, R., Fellbaum, C., Gross, D., and

Miller, K. (1990a). Introduction to wordnet: An on-

line lexical database*. International Journal of lexi-

cography, 3(4):235–244.

Miller, G., Beckwith, R., Fellbaum, C., Gross, D., and

Miller, K. (1990b). WordNet: An on-line lexical

database. International Journal of Lexicography,

3(4):235–312.

Palmer, R., Gopalakrishnan, G., and Kirby, R. (2006). The

communication semantics of the message passing in-

terface. Technical Report UUCS-06-012, The Univer-

sity of Utah.

Scribner, K., Scribner, K., and Stiver, M. (2000). Under-

standing Soap: Simple Object Access Protocol. Sams

Indianapolis, IN, USA.

Szyma´nski, J. and Dusza, K. and Byczkowski, Ł. (2007).

Cooperative Editing Approach for Building Wordnet

Database. Proceedings of the XVI International con-

ference on system science, pages 448–457.

Vanderwende, L., Kacmarcik, G., Suzuki, H., and Menezes,

A. (2005). MindNet: an automatically-created lexical

resource. HLT/EMNLP. The Association for Compu-

tational Linguistics.

Viegas, F., Wattenberg, M., Kriss, J., and Van Ham, F.

(2007). Talk before you type: Coordination in

Wikipedia. In Hawaii International Conference on

System Sciences, volume 40, page 1298. IEEE.

Yang, Y., Sun, C., Zhang, Y., and Jia, X. (2000). Real

time cooperative editing on the Internet. IEEE Inter-

net Computing, 4(3):18–25.

WordVenture - COOPERATIVE WordNet EDITOR - Architecture for Lexical Semantic Acquisition