Annotation task description

This page provides an overview on the procedure of Engagement resource annotation. Essentially, we will provide annotations for the following layers in that order:

Step 1: Clause boundary detection

Step 2: Engagement span detection

Step 3: Engagement stateragy annotation

Step 4: Primary vs Secondary engagement classification

Step 5: Supplementary rhetorical device annotation

These steps are explained below.

Step 1 Clause boundary detection

The annotation should start with clause boundary segmengation and clause type identification. In this task, you will segment a running text into T-units and clauses, and then categorize the type of clauses (e.g., MAIN, SUBORDINATE). The purpose of this layer is help you better identify engagement strategies (Step 2 and 3) as well as detection of primary vs secondary engagement (Step 4).

The following figure shows the result of this step.

clause-boundary

Essentially, you would follow the steps described in clause boundary detection. You can also see real examples of this step in the clause here.

Step 2 Engagement span detection

Once the clause layer has been tagged, the next step is to identify any engagement strategies in the sentences. Step 2 and 3 are dedicated to this process. Although Step 2 and 3 are presented separately, you would start identifying engagement categories (Step 3) immediately without explicitly following the recommended steps (Step 2 -> 3). This is totally fine. Just make sure that you identify the span of the engagement strategy correctly, which is described in this step. Alternatively, you can repeat Step 2 and 3 for each items you identified in the running sentence(s).

engagement-span

For more details of this step see tag spans.

Step 3 Engagement strategy annotation

For each of the engagement resource spans, decide specific category from the list of engagement strategies (e.g., ENTERTAIN, ATTRIBUTE, CONCUR, etc.).

For details, see engagement category.

Once you are done with this step, the annotation should look like the following:

engagement-category

Step 4 Primary vs Secondary strategy classification

As the final step, we will classify engagement resource into primary and secondary strategies. Primary strategies are those that determines the engagement values of the whole T-unit. Secondary strategies are those that supplement the primary engagement values in the clause; this often occurs (but not always) in the subordinate/embedded clauses.

For details, see primary vs secondary strategy.

Figure needs update {. .label-red}

primary and secondary

Step 5—Supplementary rhetorical move annotation

To be updated {. .label-red}

Resource layers

Part of Speech layer

You can refer to Part-Of-Speech (POS) tags automatically generated by spaCy during the annotation. There are two layers in the POS annotation. One is coarse-grained (UPOS); the other is finer-grained (XPOS). For UPOS, see the following documentation from Universal Dependency project. For XPOS, see the following table from spaCy documentation on POS scheme.

Table of Content

The following is the table of content for the manual. The original deanonymized version of the manual has sidebars for annotators to navigate through the contents. This could not be implemented in this anonymized version for review.

Back to Home