Clustering involves using some distance or similarity metric.
What is the best way to score the similarity of these small sets of words? Criteria: These are technical terms which are extracted from survey answers about a technology system; the sets will contain between 4 and 10 words; none of the words repeat within a given set; the words are technical terms which are directly related to the different type of hardware and software used for different parts of the system. The goal is to cluster the systemss (one at each company) by using the "distance" between the sets of words. This is intended to be exploratory in nature, rather than exact.
As the systems are hierarchical in nature, I plan to use agglomerative hierarchical clustering, but I will also try k-means.
I am considering using Jaccard Similarity (size of intersection / size of union).
Any suggestions for the distance metric for this situation?