Identifying Output Interactions Among Is Projects - A Text Mining Approach

In the literature, there is anecdotal as well as empirical evidence for the existence and the business impact of output interactions among information systems projects. While a lot of sophisticated optimization models have been suggested which already provide for the consideration of output interactions when selecting information systems project portfolios, the necessary data required for their application in business practice are usually not available to the planner. There is a lack of techniques in the literature on how to identify output interactions already at the time, a portfolio is planned. We attribute this lack to the rather semantical nature of output interactions. We contribute to filling the identified gap by conferring semantic clustering - a technique originating in the text mining literature - to the field of information systems project portfolio selection. A prototypical decision support system is developed that uses latent semantic analysis and hierarchical clustering to identify potential output interactions among information systems project proposals based on semantic similarities within their goal descriptions. This paper focuses on the design of the developed prototype and argues that latent semantic analysis represents a very promising technique for the identification of output interactions among information systems projects.