ACE (Automatic Content Extraction) English Annotation Guidelines for Relations (Version 6.2)
(2008)Abstract
Introduction
The goal of the Relation task is to detect and characterize Relations of the targeted Types between entities. Relations are ordered pairs of entities. This means that the order of the arguments is important in the identification of Relations. To capture this idea, two different Argument slots (arg1 and arg2) are provided for each Relation. For example, in the sentence
George Bush traveled to France on Thursday for a summit.
There is a Physical.Located Relation between George Bush and France. In Physical.Located Relations, the Person that is located somewhere will always be assigned to arg1 and the place that the Person is located will always be assigned to arg2.
Types and Subtypes will be assigned to every Relation. For each Type, there is a set of possible Subtypes. Types and Subtypes are intended to categorize the Relations on the basis of their meaning. In the example above, the Type of the Relation is Physical and the Subtype is Located. For a complete description of the types and subtypes we will identify, please see Section 3 below.
We will tag the Syntactic Extent for every Relation identified and characterize the Relation by assigning one of the eight Syntactic Class types. The Syntactic Extent of the example above is the entire sentence. The Syntactic Class is Verbal. For a complete discussion of the rules for identifying Syntactic Classes and Relation Extents please see Section 2.3 below.
We will assign a Modality and Tense attribute to each Relation identified. For a complete discussion of the rules for identifying Modality and Tense, please see Section 2.2 below.