|Table of Contents|
|Table of Contents|
Some of the modules in the modular versions inevitably overlap. By `overlap' we mean that the same information is represented in different modules. The overlap is necessary to achieve the efficiency required by readers who consult modules separately, which implies that the modules must be self-contained and complete. Further overlap results from the requirement of clarity, which demands the coherence of constituent modules to be expressed in module summaries. As a result of that, the overlap could reduce the efficiency for readers who consult the entire article or a large part of it linearly. However, we compensate by laying out an essay-route that avoids overlap and by hiding overlap from the reader's view in the same way we mentioned in the previous section for hiding potentially redundant details.
|Figure: An indication of the overlap in the modules in A05: the percentage of the modules listed horizontally that overlaps with the modules listed vertically. Overlap between 0% and 9% is left white, overlap of 10%-49% is indicated in light grey, 50%-89% is dark grey, and 90%-100% is black. For example, 20% of the content of the module summary of the complex module Methods A05-m3 can also be found in the module Findings A05-m6a, and the modules Treated results A05-m4bi1 and A05-m4bi2 overlap at least 90%|
To get an indication of the overlap in the modular articles, we have examined the relative overlap between modules, i.e. the percentage of information contained in a module that is repeated elsewhere, versus the percentage of `new' information that is contained only in the module at hand. Such an indication is given for article A05 in figure 126.96.36.199
There are different types of overlap:
|Figure 5.2: a) Hierarchical overlap between the module summaries in complex modules and the text in their constituent modules; b) Parallel overlap between similar modules; c) Embedding overlap between modules that all mention the key concepts of the article; d) Serial overlap between subsequent modules|
The module summaries of complex modules summarise their constituent modules for the sake of clarity and completeness. They are not supposed to contain `new' information that does not appear in the summarised document. Therefore, complex modules automatically overlap with their component modules. The module summaries of complex modules may also overlap somewhat with other modules, as a result of the coherence and completeness requirements.
The module summaries make up about a fifth of our example-articles. Hierarchical overlap is not problematic, because the reader is aware of the `redundant' status of the module summaries and can skip them. The essay-type route by-passes almost all modules summaries.
Parallel overlap is prompted solely by completeness considerations, catering for those who wish to consult only one module This overlap can be quite massive and literal. For example, the two elementary Treated results modules with the experimental results in A05 contain almost the same text accompanying the different figures. Therefore, the overlapping part should be hidden from view in order to accommodate readers of more than one of the overlapping modules.
The modules can have an internal sub-problem-solution pattern. In order to make that pattern complete, the situation of the module has to be indicated. Therefore, we include in each module an indication of the overall problem-solution pattern represented in the entire article, which embeds the module in its context.
Embedding overlap is prompted by efficiency considerations for readers who wish to consult only one module, as well as clarity considerations for those who may be interested in a larger part of the coherent reasoning presented in the article as a whole. To increase the readability for those who read more than one module, embedding overlap should not be too literal. The author can be instructed to avoid literal overlap of this type.
In the modular versions of A05 and A08, the `key concepts' of the articles are repeated - in some form, in more or less detail and with different emphasis - in almost every module. An example of such a key concept is the fact that the chemi-ionisation of a certain kind of process is studied using a particular experimental technique and some particular theoretical model (see the table 5.3). This reference to the key concepts is kept very brief; a few words may be sufficient. Thus, embedding overlap forms a small part of the modularised articles. Massive embedding overlap would imply that the modules cannot be separated.
The situation in which the sub-problem-solving of a module takes place involves the entire problem-solving process of the article, but more particularly the previous step in the process. We specifically call the resulting overlap `serial overlap'.
For example, the method is indicated in the module that describes the results obtained using that method. A special case of such overlap concerns the figures of results that are included both in a Results module, and in the Interpretation module about the interpretation of these results.
Massive serial overlap between particular modules would be a reason for merging the overlapping modules. In A05 and A08 serial overlap is quite small.
We conclude that, in general, it is possible to write modular articles that satisfy the reader's and author's requirements. The main drawbacks of a modular structure are that it entails a higher workload for the author, which can be alleviated by appropriate software, and that the modules tend to overlap. The most important assets of a modular structure in an electronic environment are its flexibility and explicitness. The flexibility allows readers to efficiently obtain information of interest, by allowing them to choose which module to consult, and in which sequence, and whether or not to ignore the details and overlap that is hidden from view. It also allows the author to repeatedly use published modules. Since the modular structure is more explicit than the structure of a linear article, modular publications can be clearer than linear publications.