Question Answering Based Annotation for a Corpus of Spoken Requests.

This paper introduces a semantic annotation scheme for spoken information access requests, specifically derived from QuestionAnswering (QA) research. We argue that spoken requests annotations can be effectively enriched with useful information (e.g. the Expected Answer Type and the Topic of a request), which is broadly used in theQA framework to fully capture the content of a request and extract the sought-after information. The proposed annotation scheme has been adopted in the creation of the QALL-ME benchmark, a corpus of spoken requests related to cultural events of a town.