Objectives

The following are the main objectives of the annotation task:

• Increasing the knowledge about semantic corpus annotation, in particular with a view to evaluate the usability of the classifications embodied in semantic descriptions of recent European projects, such as EuroWordNet and SIMPLE. Even though for German, other than for Italian, no full taxonomy is available, an attempt was made to use the same inventory of semantic types (derived from SIMPLE), for both languages, and this task has indeed proven feasible without greater obstacles.
• Analyzing to what extent syntactic and semantic information can be used for word sense disambiguation, and, in the long run, for semi-automatic alignment of equivalent candidates. Therefore, a tabular summary of the data will be generated (see database dump).
• Determining to what extent a corpus annotated with both syntactic and semantic information serves purposes of a detailed lexicographic description of the verbs analyzed. With a view to work on collocations, semantic selection restrictions, as well as constraints for use in applications of NLP (e.g. in the fields of Information Extraction and Information Retrieval), it is useful to have a small scale resource available, which can serve as a test bed for different approaches in this area.
• Testing available tools for query and retrieval of data from corpora with multi-layered annotation: the data output of the ELSNET action is a corpus with multi-layered annotation, which can be used as a test resource for query tools, such as the MATE Workbench (designed in the MATE project, which itself is the result of an ELSNET exploratory action). A version of the data compatible with the MATE guidelines for linking of elements is under construction.

The corpus is experimental in several respects, in particular because of its small size, and of the specific conditions on item selection.