By: Graham Neumann (923872)
For the first grid, I chose a topic who's domain I am fairly comfortable with. The reason for doing so was to make it easier to decide how usefull and accurate the Webgrid software really is. It is for this reason that I chose to create a grid relating to music within the context of song writing. As I used the same grid in the comparasion section of part II, I will delay a full presentation of the grid results until then. Thus, this section will focus mostly on the elements and constructs chosen.
Choosing elements was fairly straight-forward. Since I only needed to determine what I felt were the primary considerations needed to be made when writing a song, it was not difficult to simply list eleven elements. Choosing constructs, however, required the use of several of the Webgrid elicitation tools. I probably spend about equal amounts of time using the element distinguishing and triad facilities. These proved to be quite effective and were straight-forward to use, although at times, I found it difficult to come up with contrasting constructs when using the triad method. This was especially true when the only relevant construct choice for a particular triad (in my mind) was not applicable to the other elements. If this was the case, I simply cancelled the suggested triad, reselecting the option until I became suitably inspired. Finally, I should point out that I ignored the "distinguish constructs" feature since I felt additional elements would have been redundant.
In general, I found Webgrid easy to use, and potentially usefull. Although, I feel that much of
its usefullness is dependant on the topic - that what may be a fairly
powerfull tool in research areas, would have little impact on the musical community. A final note,
a certain amount of frustration did occur when I would select the FOCUS or prinComm buttons and
receive the wrong image file. It wasn't until several tries that I realized that Netscape was
repeatedly access the file from my local disk cache as opposed to sending new data and retrieving
the appropriate file. It would have been helpfull if a note regarding this was clearly visible.
Click here to go to a saved
version of this grid.
Click here to see snapshot of
the elements and constructs used.
Click here to see snapshot of
the PrinCom analysis.
Click here to see snapshot of
the FOCUS analysis.
Part II: Elicitation of constructs on Song Writing
Since I have already described how I came up with the elements and constructs used in the "song writing" grid, I will concentrate on describing the grid generated and then compare it with the grid generated by Dave Dattner using the same elements and constructs.
The PrinCom analysis is interesting in that it clearly helps to establish the relationships between the various elements and constructs. In general, it seems to depict the way I understand this domain with a relatively high degree of accuracy. I especially liked the way it depicted what I that was neccessary for a song to be good: namely emotional intensity, personal expression, dynamics, edge, etc. I was almost surprising how everything seemed to trully fit my mental model.
I found the FOCUS analysis to be significantly more interesting since the software seemed to make explicit what it thought were related items. For example, it was suprising that the grid I had created saw "Personal expression -- Consider audience" and "Emotional intensity -- Rythmic intensity" to be related. However, it was no surprise that "Impersonal -- Personal" and "Technical -- Musical" were strongly related. I particularily enjoyed the fact that it was ambigous whether any of the constructs were closely related with what was essential to making a good song. As far as relationships between elements goes, everything seemed to be as I would envision it, although some of the relationships such as "dynamics" and "edge" were weaker than I would have thought. Perhaps if more time was spend, a more precise model might have been developed.
Much of the reason I chose the topic of song writing had to do with the fact that the person I
was to compare my grids with is himself a song writer. It was interesting
see see that we did infact agree on many of the relationships. However,
some important differences did exist. For instance, I found a much stronger
relationship with popularity and the other elements then he did. Also,
it was interesting to see that I placed a stronger emphases on the
relationship between all of the constructs and that concerning if something
was nececessary or not for a song to be good.
In general, there was a definite degree of correspondance between our models
of this domain. We seemed to group many elements similarily, although
not exactly. This would indicate that we use the terminology
in slightly different ways, even though the construct relationships are
fairly well matched.
Click here to see snapshot of
the elements and constructs we used.
My stuff:
Constructing the grid for my presentation topic was much more difficult. The reason for this was two fold: I have less knowledge of the domain then I did for the first grid, and many of the elements I chose were very closely related. This aside, I was able to come up with a reasonable construct grid.
As in the first grid, choosing the elements was fairly straight forward. I simply listed all of the major ideas I discussed during my section of the presentation on intelligent agents. This resulted in several clusters of items, each of which was fairly distinct.
Coming up with the constructs was even more difficult than before. This was due to the inherant clustering of the elements I had chosen. Many times when I chose the triad function, I was faced with a set of three very distinct items, making the processes of singling one out almost impossible.
The final grid displayed several interesting relationships. The PrinCom analysis contained four major clusters. The most interesting were those centered around current research trends. I found it a little hard to believe that elements such as facilitators, ACL, and declarative communication were not clustered together around "The Ideal Agent." However, elements such as ACL, KQML, and KIT were appriatly all grouped together. Once again, I would have to conclude that further use of Webgrid would have likely resulted in a more accurate grid.
The FOCUS analysis also revealed some interesting relationships. For instance,
the elements relating to agent communication were almost perfectly ordered in terms of
abstractness. I also noticed that some of the apparent inconsistancies found in the PrinCom
analysis were not present in that of the FOCUS. Perhaps this is because the emphases of the
latter is not on the relationships between constructs and elements.
Click here to go to a saved
version of the grid.
Click here to see snapshot of
the elements and constructs used.
Click here to see snapshot of
the PrinCom analysis.
Click here to see snapshot of
the FOCUS analysis.
