Local time: Friday, 05-Dec-2008 12:11:29 EST
Last update at /special/conference/apwww95,Friday, 21-May-2004 09:47:27 EST

World-Wide Collaborative Writing: A Case Study

Masafumi Higuchi

NTT Software Laboratories, 250 Cambridge Ave.,
Suite205, Palo Alto, CA 94304, USA

higuchi@nttlabs.com

Kenji Takahashi

NTT Software Laboratories, 250 Cambridge Ave.,
Suite 205, Palo Alto, CA 94304, USA

kt@nttlabs.com
Abstract
A collaborative translation project between the US and Japan is reported. A structured process model, the Inquiry Cycle model, was implemented on the web and adapted to the project. This is a cyclic model for elaborating documents through expression, discussion, and commitment. This report discusses:
  • the effectiveness and limitations of current web technology in collaborative writing,
  • the effects of the model in facilitating discussions, making decisions, maintaining consistency over changes, and coordinating people in an asynchronous, distributed environment, and
  • the implications for developing and adapting methods and tools for collaborative writing via the web.
Keywords
CSCW, collaboration, distributed environment, collaborative writing, coordination

Introduction

When people collaborate in creating artifacts, such as documents and software, communication among them is very important. Communication allows discussion and agreement to be reached. Management of changes to the artifacts is also important. For example, field studies[10] in requirements analysis have shown that communication and reaching agreement occupy two thirds of all analysis efforts.

To support these types of cooperative work, we are using an ``information server'' that lets us record and manage shared information related to the work, and coordinate people and tools. We can also manage versions of the artifacts when recording and managing shared information. We record changes to the artifacts and the discussion leading to the changes so that we have a complete change history. Storing and managing the information and the artifacts related to the cooperative work in one server makes it easier to keep the consistency of the artifacts.

By monitoring access to the artifacts in a server, participants can see how the work is progressing. This also allows progress to be managed. In cooperative work, some participants may perform actions that depend on the actions of other participants. If they are working asynchronously, the relevant participants should be explicitly notified of these dependencies, otherwise it may cause delay in the collaborative work. In these cases, the information server can assist participants by sending messages to relevant participants to notify them of the dependency between their actions.

Cooperative Work Support

In asynchronous work, participants do not all need to work at the same time, which makes their schedules more flexible and increases the efficiency of their work. In addition, the Internet allows participants in different countries to collaborate, and asynchronous working is essential because of time differences. Here, we consider asynchronous collaborative work in a globally distributed environment.

Many of the existing distributed/asynchronous collaborative tools are based on e-mail[7]. However, they have various problems:

Our approach to these problems is:
  1. To develop a structured process model and work with the model. All messages related to a collaborative work are put into a structure. These structured messages are easy to follow, and their changes can be managed easily.

    We started from the Inquiry Cycle[5] model, which is a cyclic model for elaborating system requirement documents through expression, discussion, and commitment. And we expanded this model for asynchronous cooperative work.

  2. To implement an information server based on the model. We stored all the information involved in the cooperative work in the server, and shared the information among participants. This makes it easy to maintain the consistency of the information and we can directly incorporate the results of cooperative work into artifacts. This also makes it easy to monitor the progress of the work, which in turn makes it easier to manage the cooperative work.

    Pieces of the information are interconnected and shared among coworkers over a network, which is best represented in networked hypertext. The web is a popular way to access hypertext and it is available in various environments. This popularity and interoperability potentially solves the common ``critical mass'' issue with collaborative technologies.

    For these reasons, we decided to use a WWW-based information server.

Inquiry Cycle

The Inquiry Cycle[5] is a model of creating and elaborating documents. It was originally invented as a requirements analysis model, but it can also be applied to other fields. The model consists of three activities: expression, discussion, and commitment. These three activities repeat under the control of a strategy. (see Figure 1)
  1. Expression is the proposing or writing of information related to the cooperative work. This includes documents and software. This is the objective of the cooperative work.
  2. Discussion includes not only discussions about expression, but also the circulation of comments and individual annotations of expression.
  3. Commitment includes making decisions resulting from discussion, such as change requests, agreements about terminology, and commitments to seek missing information.
The Inquiry Cycle can be instantiated with concrete activities depending on application. For example, it is instantiated with natural language text and scenario descriptions for expression and with a speech act theory[12] similar to IBIS (Issue Based Information System)[13] for discussions[17]. To apply to the collaborative translation project, we instantiate the Inquiry Cycle. We regard the process of collaborative writing as a sequence of the following steps:
  1. Design the whole structure of a document according to a theme.
  2. Assign parts to each participants.
  3. Each participant compose their part.
  4. Assemble all the parts.
  5. Polish the document.
Figure 1: Inquiry Cycle The steps 1 and 3 are individual work and steps 4 and 5 are cooperative work. It may be necessary to return to step 3 and repeat until finising the document. This process (steps 3, 4, and 5) is the Inquiry Cycle itself. In this case expression corresponds to some parts of a document (e.g. paragraphs) and discussion corresponds to questions about part of a document, answers to these questions, and reason supporting the answers. Commitment corresponds to agreements reached via discussion and changes requested as a result of the agreement.

Analysis of Collaborative Writing

A guidebook on telecommuting, ``Smart Valley Telecommuting Guide'', written in English by the Smart Valley Consortium[1] was collaboratively translated into Japanese. This translation project lasted for about three months.

We analyzed this project and looked into three points:

The document we translated was 16,000 words (77 pages) long.

Work Environment

Sixteen people participated in the project: four in California and twelve in four cities in Japan. They never met in one place at any time during the project.

The participants consisted of three administrators responsible for managing the project and providing working environments, and thirteen volunteer translators responsible for translating parts of the guidebook and reviewing other parts.

Most participants used e-mail and the web via LANs in their offices. Some obtained access from home or hotels via ISDN or via an ordinary telephone line using a modem.

Procedure

The three-month project was divided into two phases using a mailing list, HyperNews[3], an interactive web program that accepts and displays users' inputs. At the end of the project, a survey was conducted.

Phase I

In the first phase, the participants communicated to manage the project and to distribute the original and translated documents by the mailing list.

First, we announced a ``call for volunteers'' via some existing mailing lists inside the company. We made a new mailing list which included all the volunteers, for collaborative translation and discussion about this project. We discussed the schedule and the procedure for the project via this mailing list. Based on these discussions, the administrators decided the procedure and announced it to the mailing list. They divided the original document into paragraphs, and distributed them to the mailing list. The administrators also decided the assignments for the volunteers and announced it to all the volunteers via the mailing list.

Each volunteer translated his own part and returned it. The administrators put all the received parts into a web server to browse the whole document.

Phase II

In the second phase, all the information, including versions of the document and discussions about the document, was stored in the HyperNews web and the participants elaborated the translation via the web. E-mail was also used to allow participants to contact each other.

HyperNews is implemented with cgi-scripts that let one attach responses to a web document. You can also attach a response to another response. HyperNews displays the responses in a structured way --- a response is indented and displayed just below the original text being responded to.

The participants were asked to follow the Inquiry Cycle model and categorize their submissions to the web into new versions of document paragraphs, questions, answers, change requests, or reminders. The category was explicitly shown in the message title.

N (New version):
New version of a translated parts of the document.
Q (Question):
Question about another message.
A (Answer):
Answer to a question.
C (Change request):
Request for a change resulting from a discussion. All changes can be explicitly recorded.
R (Reminder):
Other types of messages. For example, a message about scheduling.
Figure 2:HyperNews

Figure 2 is a example of HyperNews running on XMosaic[18]. The upper half of the view shows a paragraph of the original document. paragraphs. The lower half in Japanese shows the responses related to this paragraph. These responses are displayed in a structured way by subject and author. The first response is the translated paragraph. The second one is a question on the translated paragraph and the third is an answer to the question. Based on this answer, a change was requested as a response to the answer. The last response is a new version of the translated paragraph.

We wrote an explanation of how to use HyperNews and how to apply the Inquiry Cycle on HyperNews, (about 2400 characters in Japanese) and distributed it to all participants via the mailing list.

At the end of the project, a questionnaire survey was conducted.

Lessons Learnt

To study the process of collaborative translation, we compared the process using a mailing list with the one using HyperNews.

By using mailing lists, the information is automatically sent to each participant, while the information stored in the web server should be accessed by each participant. As a result, the process using the web had trouble staying on schedule and the administrators often had to urge the participants to complete their own tasks on time. Administrative messages including process management ones accounted for about 40% of all the messages in the e-mail process and about 20% in the HyperNews process. Such messages, which do not contribute to the content of the document, are an administrative overhead, which is seldom found in collaborations among fewer than six people located in the same place[6].

In order to examine this administrative overhead, we classified messages into two types:

  1. messages related to the contents of the document, and
  2. messages related to project administration.
We also examined the administrative messages in detail. Some were urgent messages, such as help for participants who are experiencing trouble in accessing the server or instructions to reactivate the stuck process. These urgent messages would be better sent as e-mail rather than on the web. To support collaborative work, the information server should send messages to appropriate participants.

The process using e-mail

In the process using e-mail, the administrative messages included two types of messages, ones for process management and ones for other miscellaneous jobs. Both types had almost the same number of messages. (see Figure 3)
Total number of messages: 54
Administrative messages: 21 (38.9%)
Process management: 9 (42.8%)
Miscellaneous jobs: 12 (57.2%)
Figure 3: Administration messages by e-mail

The process using HyperNews

In phase II, messages for miscellaneous tasks greatly exceeded those for process management. Almost all of these miscellaneous messages concerned trouble accessing the web server, such as a participant being trouble to view Japanese document correctly. We should be able to avoid this type of administrative message (see Figure 4) by preparing the environment for accessing the information server before starting the collaborative work.
Total number of messages: 497 (by e-mail: 92, by HyperNews: 405)
Administrative messages: 88 (17.7%; by e-mail: 85, by HyperNews: 4)
by e-mail
Process management: 16 (18.8%)
Miscellaneous jobs: 69 (81.2%)
by HyperNews
Process management: 3 (75.0%)
Miscellaneous jobs: 1 (25.0%)
Figure 4: Administration messages by HyperNews Most of the messages (94%) were classified correctly according to the Inquiry Cycle. This percentage is very high considering that we did not spend much effort on teaching the model to the participants. We only sent a mail message explaining the model and how to use HyperNews, and did not receive any questions about using them. This result shows that the Inquiry Cycle is simple and intuitive so the participants can easily follow the model. In the Inquiry Cycle model, however, the meaning of a message depends largely on the context, and links between messages are very important. If some links are incorrect, a set of the linked messages could be misinterpreted. Users should be sufficiently trained, and tools should be designed to facilitate correct classifications including a user interface that intuitively shows the linkage between responses.

In this project, there were many problems accessing the server. Since the participants used various different setups (hardware, software, links), some trouble could not be fixed. Also, current web technology cannot handle Japanese language properly. Some client software could not display Japanese documents correctly and some could not handle Japanese input.

Questionnaire

At the end of the project, a questionnaire was distributed.

The results show that the participants thought that e-mail-based collaboration was better in terms of accessibility and speed of propagation, and that HyperNews-based collaboration was better in terms of system usability and process traceability.

Learning how to use the HyperNews-based system was easy except for the handling of Japanese. And participants thought it was useful for the project because they could easily follow discussions displayed in a structured way.

When using a web-based system, the participants must take the initiative to access the server. This sometimes caused delays in the collaboration. Also participants had trouble recognizing the changes made to the document by other participants. To make the changes easier to see, we made a ``what's new'' page that recorded all responses in sequence by time. However, participants still did not know about the change until they accessed the information server. To facilitate the work flow, we need a mechanism to proactively notify relevant coworkers. One possible solution is to send notification messages to designated participants when a specific event occurs.

Conclusion

The results of our case study show that applying the Inquiry Cycle to collaborative writing on the web is useful. However, we encountered several problems:
Notification of change
Using the web, participants must take the initiative to access the web server in order to get information. Participants might be unaware of changes that affect his/her tasks, which could delay of the whole project.
Agreement
In asynchronous collaboration, it was difficult to get the agreement of every participant. We could not achieve unanimous agreement on the wording of the document.
Training
The participants thought the Inquiry Cycle was suitable for collaborative writing and they could easily use the system, but they occasionally made mistakes using it.
Accessibility
Accessing the web requires an IP link. Because most web clients were written in the USA, they could not handle Japanese correctly. Some participants had problems viewing or inputing Japanese language text.

Future Work

From this analysis, the information server that supports asynchronous collaborative working should have the following functionalities.
Coordination
When a participant puts a new message on the server or makes a change to the document, the system should notify relevant participants. Also the system should automatically send a message to relevant participants according to some prescribed schedule.
Agreement
The system should use e-mail or web forms to receive and count votes, and then display the results.
Version Control
As the collaborative working progress, various versions are made. Management of variants is critical in concurrent editing of the some document.
Interoperability
The system should have interfaces to various clients so participants can use their favorite clients to access the system. For this reason, the system should be able to handle various protocols such as HTTP[16] and SMTP[14].
We will utilize these functionalities in EColabor, a hypermedia system we are developing for collaborative requirements analysis. The main focuses for EColabor are: (1) extending HTTP to enable sophisticated version control necessary for authoring and managing large and complex documents, such as system requirements specifications, and (2) full utilization of the WWW and MBone[15] as powerful communication and sharing technologies.

The extension of HTTP includes management of traceability between different versions and support for concurrent authoring (e.g., checking in/out and management of variants). The communication support includes synchronous and asynchronous audio-video communication and application/data sharing.

Acknowledgements

The research described in this paper largely depneds on the volunteers who participated in the collaborative translation project. The authors acknowledge their efforts and assistance.

Also we acknowlege valuable advice and assistance of many other people. In particular: Mr. Ichiro Nakagawa, Mr. Toshiaki Takada, and Dr. Atsuhiro Goto. Additionally we thank Smart Valley Inc. for their generous offer of the ``Smart Valley Telecommuting Guide''.

References

[1]
http://www.svi.org
[2]
http://www.svi.org/PROJECTS/TCOMMUTE/index.html
[3]
http://union.ncsa.uiuc.edu/HyperNews/get/hypernews.html
[4]
C. Potts and K. Takahashi, ``An Active Hypertext Model for System Requirements'', Proc. 7th Int. Workshop on Software Specification and Design, IEEE Comp. Soc. Press, 1993, pp.62--68
[5]
C. Potts, K. Takahashi, and A. Anton, ``Inquiry-Based Requirements Analysis'', IEEE Software, Vol.11, No.2, pp. 21--22, 1994.
[6]
C. Potts, K. Takahashi, J. Smith, and K. Ota, ``An Evaluation of Inquiry-Based Requirements Analysis for an Internet Service'', Proc. RE95, pp.172--180, 1995.
[7]
H. Ishii, ``CSCW and Groupware'', Ohm, 1994
[8]
T. W. Malone, K. R. Grant, K. Lai, R. Rao, and D. Rosenblitt, ``Semistructured Messages Are Surprisingly Useful for Computer-Supported Coordination'', ACM Trans. Off. Inf. Syst., Vol.5, no.2, Apr. 1987, pp115--131
[9]
T. W. Malone, K. R. Grant, F. A. Turbak, S. A. Brobst, and M. D. Cohen, ``Intelligent Information-Sharing Systems'', Commu. ACM, Vol. 30, no.5, May 1987, pp.390--402
[10]
B. Curtis, H. Krasner, and N. Iscoe, ``A Field Study of the Software Design Process for Large Systems'', Commu. ACM, Vol. 31, no.11, Nov. 1988, pp.1268--1287
[11]
M. Lubars, C. Potts, and C. Richter, ``A Review of the State of the Practice in Requirements Modeling'', Proc. RE '93: IEEE Int. Symp. Requirements Eng., Jan 1993, IEEE Computer Society Press, pp.2--14
[12]
J. R. Searle, ``Speech Acts'', Cambridge University Press, 1969
[13]
J. Conklin and M. L. Begeman, ``gIBIS: A Hypertext Tool for Exploratory Policy Discussion'', CSCW '88, Sep. 1988, pp.140--152
[14]
J. B. Postel, ``Simple Mail Transfer Protocol'', RFC821, Aug. 1982
[15]
S. Deering, ``Host Extensions for IP Multicasting'', RFC1112, Aug. 1989
[16]
T. Berners-Lee, R. Fielding, H. Nielsen, ``Hypertext Transfer Protocol -- HTTP/1.0'', INTERNET-DRAFT, Aug. 1995, http://www.ics.uci.edu/pub/ietf/http/draft-ietf-http-v10-spec-01.txt
[17]
C. Potts, K. Takahasihi, and A. Anton, ``Inquiry-Based Requirements Analsis'', IEEE Software, Mar. 1994, pp.21--32
[18]
http://www.ncsa.uiuc.edu/SDG/Software/Mosaic/NCSAMosaicHome.html

Appendix: Result of questionnarire

Total number of answers12
Problems
  • I have trouble in displaying Japanese text.
  • I cannot input Japanese text.
  • Browsing documents is too slow.
  • Because my office is IP-unreachable, I cannot access the server.
Time for mastering usage of the system
Max 1 day
Min 5 min.
Average about 120 min.
Easy to use or not
Easy 4 persons
Not easy 1 person (saying that there are too many hierarchy.)
Response of the system
Slow 4 persons
Fast 1 person
Functionality needed
  • When a response is made, the system should automatically notify relevant participants by e-mail or in some other way.
  • The system should indicate visually the realtionship between the parts I am reading and the whole document.
Is a paragraph suitable for handling?
Yes 9 persons
  • But I need a way of reading previous paragraph or next paragraph easily.
  • Because paragraph is a smallest element of documents.
Too small 1 person
  • The most suitable element size is about the size where I can see original and translated documents in one display.
Is the model suitable for collaborative editing?
Yes 4 persons
No 1 person
Evaluation for sevelal points (e-mail is better:1 $\sim$ HyperNews is better:5) 
--------------------------------------
                       Max   Min  Avg
--------------------------------------
Accessibility        |  4    1    1.8
Usability            |  5    1    3.4
Traceability         |  5    1    3.5
Speed of propagation |  4    1    1.6
--------------------------------------
[Return to Table of Contents]
COPYRIGHT © 1995 by AUUG95 and APWWW95 Charles Sturt University. ALL RIGHTS RESERVED.