back next next
Table of Contents
Index
Glossary






























































































































Table of Contents
Index
Glossary
back next next

Next: Workload: the size of Up: The feasibility of modular Previous: The feasibility of modular

  
Difficulty: from sections to modules

Following the definitions of the modules given in section 4.2.1 and their specification in the guidelines given in appendix A, the information presented in the original, linear articles A05 and A08 has been recast in two modular articles (which consist of  microscopic modules and links) and in a separate set of mesoscopic  modules. The mesoscopic modules represent information that is used, and partly presented, in both articles. As an illustration, we have also provided two macroscopic  Theoretical methods modules. The macroscopic modules are supposed to be parts of monographs or tutorials. One of them is based on a section of O'Malley's book [O'Malley, 1971], rephrased to form a self-contained module.

At the microscopic and mesoscopic level, we have distinguished modules firstly by their conceptual function. As we saw in section 4.2.2, the definitions of such modules are based on the prototypical sections in scientific articles. As we see in table 5.1a, the original version of A05 indeed has a rather prototypical structure, which facilitates the comparison of the new modules with the original sections. The original version of A08 is less prototypical. It has more sections, as A08 is a long article of thirty pages. The theoretical component of the article is emphasised by the inclusion of the sections 3.Potential curves and 4.Calculations. However, these different sections still form an overall IMRDC structure. The collection of the sections 2.Experimental, 3.Potential curves and the first half of 4.Calculations could be grouped under the heading Methods, the collection of the second half of the section 4.Calculations and the 5.Measurements could be seen as Results, and the sections 6.Comparison of the measurements and the calculations, 7.Discussion and 8.Rotational coupling could be grouped under a higher-level heading Discussion.5.8

To distinguish constituent modules within the modules created by means of the conceptual function, we have used a rudimentary physics classification sketched in table 4.1. This simple typology takes into account the INSPEC index terms associated to the original articles. The difference with the original terms is that we do not only use simple index terms, such as `rotational coupling', but also complex ones, e.g. `differential cross section[chemi-ionisation; Na, I; 13-85 eV]'. We have used simple bibliographic data to characterise the modules thus created. The original articles are uniquely identified by the name of the journal, the volume and the page numbers of the published article. In the modularised versions of corpus articles we have characterised each module by the family names of the authors, the name of the institute, the name of the journal in which the article was originally published and the year of its original publication. We have also employed the corpus-specific unique identification code described in Appendix C.

The original versions of the articles are connected to other work by means of references, and occasionally different parts are connected by an internal reference, explicitly (e.g. ``see appendix'' in the section 3.Results of A05, p.67), or more implicitly (e.g. ``[...] as will be derived later on in this paper'' in the 1.Introduction of A05, p.62). In the modularisation process, we created characterised hyperlinks between modules and parts of modules, to express one or more different types of relations between those modules or parts of module, between information units underlying them, or between entities addressed in them. The different types of relations are defined in section 4.3, and in appendix A.3 guidelines are given for making these relations explicit in links.

  In the modularisation process, the core of the various modules could easily be established. In many cases, information of a particular type was concentrated in a single section of the original article, so that a large part of that a section could be transferred directly to the corresponding main module. Other types of information were distributed over various sections, albeit in such a way that they could nevertheless be easily identified and coherently represented in a module, either by directly copying sentences or paragraphs, or after straightforward rephrasing. In A05, the sections 1.Introduction and 3.Results in particular contain a lot of information that did not fit in the corresponding modules Positioning and Results but that could easily be recast in other, more appropriate modules. In A08, the section 4.Calculations contains various kinds of information: information on the calculation methods, which we have recast in the module Theoretical methods, as well as a presentation and a discussion of the results of the calculations, which have been recast in respectively the Results and the Interpretation.5.9 In section 5.3, we shall compare the different modules and sections in greater detail.

Mesoscopic  information concerning the situation, the central problem, the experimental methods and the theoretical methods could easily be identified. These are the types of information that, according to the domain-specific interactants profile we described in section 5.1.2, have to be represented in a form suitable for multiple usage. The contents of the mesoscopic  modules could not solely be derived from the two original articles. In order to make the modules complete, as self-contained units that can be consulted by less informed readers, we derived information from other publications in the corpus (mainly review articles) as well. We also obtained information from the senior author of the research project in private communications.

Difficulties were caused by the borderlines between most modules. Creating the modular versions, we encountered the following kinds of problems.

Firstly, some information was inherently difficult to characterise unambiguously: it bore the characteristics of more than one type of information. In some cases, this problem could be solved by taking into account the internal structure of the modules, as described in section 4.2.6. For example, at the level of the entire article, we characterised the information about the way experimental results were presented in a figure as `results'. At the lower level of the Results module, we considered this particular information to play the role of a method in a sub-problem-solving process aimed at the final treated results. In the other cases when information could, from different perspectives, be considered to play different roles, we have presented it in more than one module in order to allow the reader to locate it using different labels. So, difficulties in the characterisation of information led to overlap between modules. Deciding on the exact type of particular information will always be difficult in some cases. However, it will be less difficult for the author of a new modular article, who has a thorough understanding of the information, than for the outsider who tries to modularise an existing article.

Secondly, even when the characterisation of the different `strands' was unambiguous, some information of different types was interwoven so closely, that it was difficult to disentangle it. For example, the presentation of the results in the original version was intertwined with that of their (qualitative) interpretation. In practice, this intertwinement led to some overlap between modules: the principle figures with the results were included in both the Results module and the Interpretation module. In the conversion of existing linear articles, this problem is unavoidable, as in the linear version the information has been interwoven intentionally, to form a single narrative. Directly writing a modular article will reduce the problem, although it probably cannot be avoided entirely when different types of information are strongly interdependent.

Thirdly, in order to obtain a complete module, it was necessary in some cases to add information to the information provided in the original version. This is merely a conversion problem. The author of a new modular article can follow the  guidelines and directly include all required information.

With respect to the links between the modules, it was easy to identify and express the organisational relations. However, in many cases it was difficult to determine which scientific discourse relations have to be made explicit in links. In section 5.4 we shall discuss the different types of relations in detail.

Present-day authors are accustomed to writing experimental research articles with an `IMRDC' structure.  If an author is supposed to write an article with an entirely different structure, he will not find it feasible, as long as he has not become used to the new structure. This is one of the reasons why we have taken the prototypical section structure as a starting point in the development of the model; the basic structure of the modular article resembles the structure of the linear article. This similarity of the modular and the linear article is demonstrated by the fact that large parts of the modular articles could be copied from the original version, though the sequence of the paragraphs and sentences does not necessarily remain the same.

The modular structure is more fine-grained than the structure of the linear article. Therefore, the author of a modular article has to make more decisions concerning the correct placement of the information, which could make it more difficult.5.10 However, the modular structure is more systematic and thereby provides the author with stronger guidance. Moreover, we have specified more elaborate  guidelines for the author than is usual in traditional  journals, which can be implemented in software assisting the author. The other side of the coin is that the modular structure is more rigid, which may frustrate authors who wish to elaborate in more essayistic prose.

We have found that the modular version differs distinctly from the original versions, so that articles written in linear form cannot easily be converted automatically into a linear form. As we have also found that is was difficult to reconstruct the information underlying an article written by someone else, we conclude that writing a modular article can be done most efficiently and effectively by the author himself, provided he has adequate tools at his disposal.



Next: Workload: the size of Up: The feasibility of modular Previous: The feasibility of modular