This paper addresses the problem of the definition of a semantic rela-tion set which can be hopefully applied to more than one language. Here I propose a very simple framework with three general semantic relation classes: association, taxonomy and space. In order to train a classifier for those classes in italian and english I run an experiment with Part-Of-Speech, End-of-Sentence and Named Entity features from TextPro, a text tool suite for italian and english, and I compared three machine learning algorithms (decision lists, decision trees and support vectors) to retrieve if-then rules that worked best on general-purpose datasets (BNC for english and CORIS for italian). From the rules I developed a software, called RA (Relation Annotator), that is compatible with TextPro itself, it obtained an average F1-measure of 0.789 for english and 0.781 for italian with decision lists rules. 1 Introduction and Related work

The experiment reported in this paper is a feature engeneering op-eration that explores the syntactic structures of three different semantic relation types with tree kernels. This is done in order to find whether syntax can be a good feature to improve semantic relation separability in classification tasks. The average accuracy obtained is 63.09 %. 1 Introduction And related work In a previous experiment [1] it was found that three semantic relation classes (”role”, ”location”, ”social”) yielded better results in a classification task with respect to the seven semantic relation classes in ACE 2004 (see [3]), since those three classes are easily separable. The three semantic relations are: Role (entity