Abstract This paper presents a critical analysis and explication of the approach to the problem of word sense disambiguation by the field of natural language processing. The paper is motivated by the observation that, despite the approach being interesting and promising, the actual research has some problems regarding its conceptual frameworks and methods. The methodology behind the analysis and explication is philosophical, focusing on the identification of conceptual unclarities and methodological incoherencies as well as their (practical) implications. The methodology helps to pinpoint some structural confusion in the research with respect to a central concept and an important set of methods. First, the characterization of the problem of word sense disambiguation as a computational task allows for different interpretations. When a rich interpretation is adopted, the methods developed for performing the task have certain problematic limitations. Secondly, the applicability of methods for the inductive discrimination of word senses is overestimated. The acknowledgement of this confusion makes it possible to understand the sometimes erratic course of the research and underlines the need for one, commonly shared, more refined and integrated conceptual framework and methodology. This paper may be interpreted as a humble onset of some of the necessary groundwork for developing these.
Julie Mennes, Stephan van der Waart van Gulik