Abstract
We introduce a new approach to defining and sub-categorising over-specification, which offers a better analysis than commonly used metrics such as DICE. Next, we report on a primary analysis of the Chinese MTuna corpus based on the new approach.
