Martinez (2002) introduced the idea of using bigrams and trigrams as a text mining tool. Multigrams such as bigrams and trigrams have the possibility of capturing semantic content because they capture a richer semantic structure than a simple bag-of-words approach, i.e. they capture noun-verb, verb-object, and adjective-noun pairs. While the lexicon may be relatively high dimensional, the set of bigrams is even larger. In a simple example of 503 documents, the lexicon contains somewhat over 7000 terms, but the number of bigrams is more than 91,000. Thus the structures involved are very high dimensional. In this talk I explore some aspects of the multigram approach.