Sie sind auf Seite 1von 4

Providing Mark-up and Feedba

k to Students with Online Marking

David V. Mason Denise M. Woit


dmason@sarg.ryerson.ca dwoit@scs.ryerson.ca
S hool of Computer S ien e
Ryerson Polyte hni University
350 Vi toria Street
Toronto, Ontario
Canada M5B 2K3

Abstra t the first author started teaching all 150 students in the CS1
course with a dozen teaching assistants. The original system
Online marking of assignments can lead to improved mark- was programmed as a CGI script written primarily in Bourne
ing consistency and integrate well with on-line mark report- shell script. It was functional, but extremely slow. Over the
ing. It can also be easier for the marker. Unfortunately, Christmas break it was re-programmed in Scheme as a CGI,
most such systems do not include good feedback mecha- resulting in a system that was usable but still a little slow
nisms for the students. This paper describes an environment because of the CGI start-up time. During summer’97 it was
that provides online marking with convenient, structured and re-implemented as an HTTP server in Scheme. This ver-
detailed feedback. sion had perfectly acceptable performance, but crashed oc-
casionally because of problems with the underlying Scheme
1 Introdu tion
implementation (scm) and because of a monolithic imple-
mentation. It is currently undergoing a (hopefully last!) re-
There is an increasing trend to automate as many aspects of implementation, being modularized and implemented on a
course operation as possible. This promises to be more effi- more robust version of Scheme.
cient for markers and more convenient for students because
they will be able to access information at their convenience. 2 Operation
There are many aspects of course operation that are amen-
able to some degree of computerization. Systems such as Although the program includes many parts that perform other
WebCT[1, 2] provide WWW access to course content, ad- functions such as allowing students to upload files and to ac-
minister quizzes, and provide some tools to manage grades. cess their marks, only the part which is relevant to marking
Other systems such as [5, 4] handle on-line grading of pro- assignments is described in this paper. We will describe the
gramming examinations. BOSS[3] is designed to partially current implementation and leave most comments about im-
automate marking of student assignments and labs. As well provements to the section, “Lessons Learned”.
as reducing marking time, this provides more consistent mark- The files that make up an assignment are submitted to the
ing, particularly over a range of markers. The main problem system through either an interactive “submit” command
with the BOSS system is that feedback is limited to grades on our Unix systems or via the WWW. The students do not
and out-of-context notes attached to the returned evaluations. have to modify their files in any way. All of the annota-
The system described in this paper uses the WWW to tion and recognition of parts of the assignment that are de-
provide in-context annotation of work returned to students. scribed below are automatically determined by the marking
environment itself, based on the file extension (“ .scm”,
1.1 Ba kground
“.c”, etc.).
The examples here are from marking a Scheme program,
The system described here has gone through several incar- but the annotation program also has rules for marking C and
nations over the past couple of years. It was created when simple prose. While we have not included examples for the
other assignment types, they are very similar and it is fairly
simple to extend the annotator.
Figure 1 shows the first page presented to the marker for
a particular assignment and a given student. The annota-
tion program will generate a hyperlink for each significant
feature of the assignment to provide marking and annotation
for that section. In this case, each comment block is indepen-
dently markable and the top-level define is also markable.
In addition to the markable sections, there is also a com-
ment field and assignable mark for the total assignment. The
ing information and mark for different things). There are
2 pull-down menus here: one to assign a mark (-3: : :+3),
and one to assign a mark and a comment. In this snap-
shot, we have selected one of the predefined mark/comment
choices, “-1 No course number”, and will now click
on “Submit Query”. On this page, the other predefined
comments include “-2 Missing description” and
“-10 No header comment block”, among others. As
mentioned in the next section this is probably more restric-
tive than it should be.

Figure 1: Initial marking page

name of the file is shown just below that with the date of its
last modification, and in this case there are also the results of
an automatic execution which can be reviewed by the marker
(in some cases, the execution can be set up to calculate part
of the mark automatically).

Figure 3: Marking of Code

From figure 1, if the marker had instead clicked on the


define, they would be presented with the page shown in
figure 3. On this page, the code block has been isolated, and
the type of the section has been identified as “Code”. The
same pull-down menus are provided, but a different set of
pre-defined mark/comment choices are available (including
“-2 Bad indentation”, “-3 Poor algorithm”,
“-5 Wrong interface”, and “-2 Useless code”).
In this case, the marker has typed in a comment and decided
to give a bonus mark to the student.
After the marking for an assignment has been completed
(the “Submit Mark” link on the first page has been clicked)
the page looks like the image in figure 4. This shows the
mark, the entered comment, who marked it, and when it was
marked. There are many other parts of this system, and all
of this information is provided to the student on a different
page. From here, the assignment can be remarked, or it can
be examined as the student would see it. The current system
Figure 2: Marking of Headers just keeps the latest marking of an assignment, but it should
probably be extended to keep a log of all markings of an
From the previous image, if the marker clicked on the assignment.
header comment, they would be presented with figure 2. On Figure 5 shows the student view. Here the underlining
this page, the comment has been isolated, and the type of of the hyperlinks makes it is obvious that annotations have
the section has been identified as “header documentation” been provided by the marker for the header comments and
(we treat the initial comment on assignments as identify- the first code block. The student can click on either of these
to see the particular annotation.

Figure 6: Annotation seen by student


Figure 4: Remarking page
Assuming the student has clicked on the “define” block,
they would be presented with figure 6. This shows the type
of the block, the mark and the comment from the marker, as
well as the source text of the block and some miscellaneous
identifying information.

3 Lessons Learned

To evaluate the system, there are 2 relevant aspects: how


well it works for students, and how well it works for markers.

3.1 Students

No formal poll has been done, but the student perception of


the system is positive. The active links are very visible to
the student so they can see where there were problems in
the program. By following the links, they can easily see the
comments applicable to the particular section of the assign-
ment.
The other improvement from the perspective of the stu-
dent is that because all of the markers use the same set of
marking rules consistency is improved. Unlike in [3], the
markers knew the students they were marking, but we never
received a complaint that markers were picking on, or being
easy on, any students.

3.2 Markers

The markers comments were also mostly positive. They


Figure 5: Student view found the system (when it was working well) to be easy to
use and fairly efficient. The complaints from the markers
were of two types: speed and customization.
Speed When a marker wants to make an annotation about creation of sophisticated web-based learning environ-
part of the assignment, they are only willing to wait a second ments. In Proceedings of NAUWeb ’97 - Current Prac-
or two before being able to enter the annotation. Any longer tices in Web-Based Course Development (Flagstaff, Ari-
delay led to the markers became frustrated. The CGI ver- zona, June 1997).
sions of the program took about 5 seconds, which became a
major irritation. [3] J OY, M., AND L UCK , M. Effective electronic marking
for on-line assessment. In Proceedings ITiCSE’98 (3rd
Annual Conference on Integrating Technology into Com-
Customization and Convenien e The CGI version of puter Science Education) (Dublin, Ireland, Aug. 1998),
the program enforced security by only accepting connec- ACM, pp. 134–138.
tions from particular machines. This meant that the mark-
ers could only mark from an on-campus lab, which meant [4] M ASON , D., AND W OIT, D. Integrating technol-
that the markers sometimes had difficulty finding the time ogy into computer science examinations. In Proc. of
to mark expeditiously. In addition to speed, this was one 1998 ACM Conference on Computer Science Education
of the main reasons to switch to the server implementation. (SIGCSE’98) (Atlanta, Georgia, Feb. 1998).
The current system can be accessed by students and mark-
ers from anywhere on the Internet with security enforced by [5] W OIT, D., AND M ASON , D. Lessons from on-line pro-
passwords. gramming examinations. In Proc. of 3rd Annual Confer-
Markers also found that, for a particular assignment, they ence on Integrating Technology into Computer Science
were often repeating the same comments for several stu- Education (ITiCSE’98) (Dublin, Ireland, Aug. 1998).
dents. If the instructor had thought of that annotation, then
it would be on a pull-down menu, but otherwise the mark-
ers would have to repeatedly type in the same comment.
This was inconvenient and also diminished the consistency
of marking. The new version will pool the annotations so
that an annotation used by one marker for a particular as-
signment will be automatically added to the pull-down list
provided to all markers. The list will also be sorted by usage
frequency so that the most common comments will appear
at the top of the pull-down.

4 Con lusions and Future Work

This paper has presented one of the possible design points


in an on-line marking system to provide detailed feedback
to students. One of the most significant lessons learned was
that finding a good interface for the markers was a fairly
difficult, iterative process. It was only after each problem
was resolved that the next set of problems was recognized,
despite all of the work being done by experienced teachers
and computer scientists.
As mentioned earlier, we are currently working on the
fourth version of the system. As we feel that the marking en-
vironment is now fairly stable we will be adding other course
management features to this system, as well as convenience
features for instructors (the current system requires a lot of
knowledge and editing of control files to set up properly).
Once this is complete, we will be making this system avail-
able for use by others.

Referen es

[1] G OLDBERG , M. W. Calos: An experiment with


computer-aided learning for operating systems. In Pro-
ceedings of the ACM’s 27th SIGCSE Technical Sympo-
sium on Computer Science Education (1996).
[2] G OLDBERG , M. W., AND S ALARI , S. An update
on webct (world-wide-web course tools) a tool for the

Das könnte Ihnen auch gefallen