Hu and Kibler 1996 and Pagallo and Haussler 1990).  In this work, we
investigate interactions between {\it collocational} properties and
methods for organizing them into features.  {\it Collocations} are
words that tend to appear together, for example, {\it rate} tends to
appear with the financial sense of {\it interest}.  More generally,
collocations can be defined as words that tend to co-occur with any
meaning distinction (for example, with particular word senses, topics,
or dialog acts).  From this perspective, collocations are key sources
of information in a wide range of NLP applications, from word-sense
disambiguation to discourse processing.  In this work, we
systematically investigate relationships among type of property,
method of organization, and machine learning algorithm, to identify
beneficial ways to include collocations in machine learning
experiments.