Guidelines for Annotating the LUNA Corpus with Frame Information

This document defines the annotation workflow aimed at adding frame information to the LUNA corpus of conversational speech. In particular, it details both the corpus pre-processing steps and the proper annotation process, giving hints about how to choose the frame and the frame element labels. Besides, the description of 20 new domain-specific and language-specific frames is reported. To our knowledge, this is the first attempt to adapt the frame paradigm to dialogs and at the same time to define new frames and frame elements for the specific domain of software/hardware assistance. The technical report is structured as follows: in Section 2 an overview of the FrameNet project is given, while Section 3 introduces the LUNA project and the annotation framework involving the Italian dialogs. Section 4 details the annotation workflow, including the format preparation of the dialog files and the annotation strategy. In Section 5 we discuss the main issues of the annotation of frame information in dialogs and we describe how the standard annotation procedure was changed in order to face such issues. Then, the 20 newly introduced frames are reported in Section 6.