k-means is not suitabel for clustering terms. k-means needs the objects to be in multi-dimensional numeric (continuous space) and strings are neither. What you can do is first compute distances between your terms using the distance matrix/function nodes and an appropriate distance function (e.g. Levenshtein). Then use k-medoids or hierarchical clustering. Both should work when you only have distances between objects but not a continuous space.

I believe it should be possible to use th highlighting feature. If you open the original table in a Interactive Table View node, select a cluster in the Pie Chart and highlight the selected rows via the menu then you should see the corresponding rows in the table view.