Uncovering the Intellectual Core of the Information Systems Discipline

In this appendix we discuss some technical details of our implementation of latent semantic analysis on a set of 1,615 research abstracts published in MIS Quarterly, The reader who is interested in a theoretical introduction to LSA and an illustration example is referred to Appendix C. In accordance with well-accepted information retrieval and text mining procedures we started the analysis by compiling a list of all terms used in the MIS abstracts (dictionary). The 1,615 MIS abstracts produced a dictionary of 9,706 MIS research terms. We then examined and eliminated the unique terms (those appearing in only one document). That reduced the dictionary size to 5,776 terms. As a second step, we removed trivial English words (stopwords) such as " and, " " the, " and so on. Our customized stoplist included a few additional words that we felt should be filtered out from our paper abstracts collection. For example, we added the terms " paper " and " author " as stopwords because they were not expected to add information useful to our analysis. This step reduced the dictionary to 5,410 nontrivial terms. As a third step, we removed term suffices, applying what is commonly known as term stemming. For example, we replaced " collaborate, " " collaborating, " " collaboration, " and " collaborative " by " collabor–. " This resulted in a dictionary of 3,172 stemmed terms. As a fourth step in term reduction we conducted an initial singular value decomposition (SVD) to