Some time ago I did a content analysis of publications from communication research in Finland. This was a small project ordered by the University Network for Communication Sciences in Finland. I studied the possibilities to compare studies within the field of communication and information research. The goal was to look at what 15 different departments in the network might have in common and also to develop some methods to do this.
I collected references from the 2964 publications that I found from the last ten years from 13 departments. I collected the data partly from university libraries and Nordicom’s database and partly by contacting the departments and researchers and asking them to send me lists of their publications. The publications were then indexed with Nordicom’s thesaurus and converted to a format that BibExcel could use. BibExcel was then used to do a co-word analysis of the material.
There were some challenges and problems during the data collection. There were great differences in the amount of publications between different departments. Some departments are more practicly oriented and they do not have that many publications. It was also difficult to find all available publications and some publications may have been left outside the analysis. Another challenge was indexing. I indexed all the publications alone and among the publications there were some to me unfamiliar topics. In these cases I was forced to use titles for indexing and in some cases were the titles were not describing enough, I just had to exclude the publication from the analysis. This may have affected the proportion of general and specific keywords. Another problem was that Nordicom have used three different thesauruses and libraries also use different thesauruses. So indexing almost 3000 publications was somewhat a creative task. It is also unclear how the included masters theses might have influenced the results. It can be argued that masters theses would represent the research in the departments, but there might be some differences. The last problem was time. I had a month to do this.
The data was then analyzed with the bibliometric tool-box BibExcel. The first graph was the graph of the whole network, then I used the most frequent words from each department to draw the frequencies on the graph based on the whole network. All the departments gave slightly different graphs, or frequencies for different words, but the underlying graph was stable and didn’t change with the department. This way I could compare the patterns each department gave on the whole university networks graph. Similar pattern indicated similar research, because the same word had been used to describe the research field(s) in the departments.
Below are the frequencies of appearance of co-words in the publications from six departments. The patterns are quite similar, indicating similar publications, or topics of publications, hence, similar research. The combining factor for these departments is communication.
There are three departments of information studies in Finland. These can be seen from the image below. These also have a quite similar pattern, indicating similar research.
And finally four more departments which could be called outliers, as they do not clearly share patterns with any other departments.
As a small final exercise I asked the professors from each department to send me key terms describing the research at their department. This exercise was not in any way scientific, because the words were chosen by just one person, but the comparison between what the professors thought of the research at their department would be interesting to compare with the actual research at their department. I had two list of words; the list given by the professors and a list with the most frequent key words from the publications. Then I compared these lists, as shown in the figure below.
This exercise was more of a test of the method and the results. Did I get the right key words? Had I made any major errors while indexing the publications? This exercise showed that the most of the words chosen by the professors were among the most frequent words, confirming the reliability of the results.
And because I had the data, I used it to draw the networks based on the co-words for each departments. Below is the pattern for my department. For comparing departments, these do not have as much value as the earlier graphs which were based on a single graph of the whole network, but these single graphs were of more interest for the departments themselves.
The results showed that the departments in the university network did indeed have something in common, but that there were some outliers as well. The most interesting part of this study was the method development. The method used here could be useful in studies where one might for instance want to compare how smaller networks or clusters within a larger network relate or compare to each other. The data doesn’t have to be publications and the connections doesn’t have to be co-words. The data could be social ties or information flow. My last advice for anyone considering this kind of study, do not index 3000 publications by yourself. Get some student to do it or do some other study.