Abstract

The aim of this report is to explore, collect and classify the surface characteristics of texts which contain arguments, and in particular arguments pertaining to policy. Our interest is in the automated identification of publically presented arguments, rather than in their success or failure as arguments. We are therefore not dwelling on their validity or their soundness, but on the way they are typically expressed. Of special interest to us are the policy arguments which give reasons for and against legislation proposals. The automated identification of arguments which are to be found in the public domain is an approximate science. Nevertheless, the redeeming factor is that in building an assistant to the policy maker, the interest is in the main lines of argument that pertain to an issue, which, if significant, will not be expressed only once, in atypical logical language. The computational tool we are constructing will assist the policy maker with texts that display even minute hints that there may be some reason giving enfolded within their content.