Abstract
The CLIN30 Shared Task is defined as a computational approach to classify have-doubling, which is a syntactic phenomenon combining a past participle construction with an additional participle, usually from the verb have, e.g., he has had lived there. This paper describes perfect doubling, a particular subform of the have-doubling construction, in detail, and introduces a dataset to study the phenomenon in a computational way. Different classification approaches from the Shared Task participants are discussed, and an error analysis of classification results is provided. The models reach up to nearly 80% accuracy, which is a viable starting point for further research.