A corpus of fine-grained entailment relations
Rodney D. Nielsen, Wayne Ward · 2007
This paper describes on-going efforts to annotate a corpus of almost 16000 answer pairs with an estimated 69000 fine-grained entailment relationships. We illustrate the need for more detailed classification than currently exists and describe our corpus and annotation scheme. We discuss early statistical analysis showing substantial inter-annotator agreement even at the fine-grained level. The corpus described here, which is the only one providing such detailed annotations, will be made available as a public resource later this year (2007). This is expected to enable application development that is currently not practical.