goodFor/badFor Annotation Scheme Details

A full version of annotation scheme can be downloaded here. Here we give a full list of annotation types, but we do not cover all the annotation insturctions and constraints.

  • Annotation Sets and Types

    Annotations in a document are organized by GATE into sets. The goodFor/badFor set contains the annotation types in the goodFor/badFor schema. This is the annotation set you will be working with. There also may be some Default annotations, if you have used GATE to tokenize or otherwise process a document. Similarly, if the document was an html or xml document to begin with, there will be a set of Original markups.

    Below, the goodFor/badFor annotation types are listed with brief descriptions of their possible features.

    goodFor/badFor

    • id - A unique identifier automatically assigned by GATE. The annotator does not need to sepcify this value.

    • span - The text span expressing the goodFor/badFor event.

    • polarity - The polarity of a goodFor/badFor event, whether it benefits the object (goodFor) or hurts the object (badFor).

      Possible values: goodFor, badFor

    • agentId - Specify the id of the agent of this goodFor/badFor event. The agent is the NP conducting the event. If the agent is implicit, (In the sentence, "The bill was denied". The agent of the badFor event denied is implicit.), annotate the agentId as implicit. If the goodFor/badFor event has an influencer, and the agent of the influencer and the goodFor/badFor event are the same, the annotator could leave the agentId of the goodFor/badFor event as BLANK.

      Possible values: id, implicit, BLANK

    • objectId - Specify the id of the object (affected entity) of this goodFor/badFor event. An goodFor/badFor event always has an NP object.

    influencer

    • id - A unique identifier automatically assigned by GATE. The annotator does not need to sepcify this value.

    • span - The text span expressing the influencer event.

    • effect - The effect of an influencer event, whether it retains the polarity of the goodFor/badFor (retain) event or reverses the polarity of the goodFor/badFor event (reverse).

      Possible values: retain, reverse

    • agentId - Specify the id of the agent of this goodFor/badFor event. The agent is the NP conducting the event. If the agent is implicit, annotate the agentId as implicit. If the influencer event has an influencer, and the agent of two influencer events are the same, the annotator could leave the agentId of the latter influencer event as BLANK.

      Possible values: id, implicit, BLANK

    • objectId - Specify the id of the object (affected entity) of this influencer event. The object of an influencer event is either an influencer or a goodFor/badFor event.

    agent

    • id - A unique identifier automatically assigned by GATE. The annotator does not need to sepcify this value.

    • span - If the agent of the goodFor/badFor event or the influencer event is not implicit, mark the span of the agent expression. It should be a noun phrase and the semantic and syntactic argument of the event.

    • writerAttitude - The writer's sentiment toward the agent expressed within the sentence. There does not need to be any sentiment word describing the agent.

      Possible values: positive, negative, none

    object

    • id - A unique identifier automatically assigned by GATE. The annotator does not need to sepcify this value.

    • span - The span of the object (affected entity) expression. It should be a noun phrase and the semantic and syntactic argument of the event. The affected entity in the sentence must be the syntax object, not the syntax agent. Don’t mark the object which is the agent of the verb. For example, we’ll not mark the case of "The price has increases". Though increasing is goodFor price, but first it’s a description of the price itself, second the gfbf’s object here is the syntax agent, so don’t mark the case.

    • writerAttitude - The writer's sentiment toward the object expressed within the sentence. There does not need to be any sentiment word describing the agent.

      Possible values: positive, negative, none


annotation instructions by L. Deng and J. Wiebe