The report (EAGLES, 1996f) describes the use of a tagset (STTS) derived from, the German ELM specification in a series of tests for the evaluation of the impact of:
The emphasis is on the design of these tests, and on directions for the interpretation of their results, rather than on the concrete individual results themselves. The document shows the applicability of the ELM specification as a tagset for corpus annotation, as well as the impact of certain much discussed tagset modifications on the results (e.g. giving up the separate class of ``proper names'' and merging it with ``common nouns'').
The report first states background and objectives and briefly describes the resources and tools used for the tests in question. It goes on to make proposals for test setups for the three types of problems described above, including proposals for the kinds of statistics which need to be extracted from the results and compared (section 5). These setups and interpretation guidelines are then applied to concrete tests (German material), and the results are displayed and interpretations proposed.
The appendices contain detailed descriptions of the tested phenomena, suggestions for further tests on the impact of tagset modifications and an overview of the tagset used.