Sentence similarity measures for essay coherence
Jill C. Burstein · 2007
This paper describes the use of different methods for semantic similarity calculation for predicting a specific type of textual coherence. We show that Random Indexing can be used to locate documents in a semantic space as well as terms, but not by straightforwardly summing term vectors. Using a mathematical translation of the semantic space, we are able to use Random Indexing to assess textual coherence as well as LSA, but with considerably lower computational overhead.