Annotation Based Collective Opinion Analysis
We present a tool which analyzes annotations on a document to infer collective sentiments of annotators. Annotations may include
comments, notes, observation, explanation or question, help etc. Comments are used for evaluative purpose where as others
are used either for summarization or for expansion. Further, these comments may be on another annotation, not on the original
document and referred as meta-annotations. Collective sentiments of annotators are classified as positive, negative or neutral
based on sentiments of words found in annotations. All annotations may not get equal weightage. If an annotation has higher
number of meta-annotations on it, it is assigned higher weight. If a comment is on another annotation and negates the sentiments
of previous annotator, then the weightage of that annotation is either reduced or annotation is excluded from inference. Our
tool computes collective sentiments of annotators in two steps. In first step, it computes sentiment scores of all annotations.
In second steps, it computes weighted average of sentiment scores of annotation to obtain the collective sentiments. We demonstrate
the use of tool on research papers.