Our survey suggests that there is a consensus amongst users of language corpora with respect to the following minimal set of recommendations.
For corpora which include morphosyntactic or syntactic analysis, the picture is less clear. Where such corpora are used, respondents seem to rate all features equally highly; for many it appears to be a case of `all or nothing'
We conclude that, while individual corpora may vary in terms of the range and depth of features presented, those features which are encoded should conform to EAGLES guidelines, preferably making use of the EAGLES Intermediate Representation for morphosyntactic features for this purpose. This will facilitate automated and semi-automated validation of such mark-up against the control set of EAGLES features. This does not, however, preclude the use of different schemes, particularly where corpora are likely to be processed mainly by humans rather than by machines.