Estimating Annotation Complexities of Text Using Gaze and Textual Information

The basic requirement of supervised data-driven methods for various NLP tasks like part-of-speech tagging, dependency parsing, machine translation is large-scale annotated data. Since statistical methods have taken places overrule/heuristic methods over the years, text annotation has become an essential NLP research. Annotation refers to the task of manually labeling of text, image, or other data with comments, explanation, tags or markups—for NLP, often carried out by linguists to label raw text. While the outcome of the annotation process, i.e., the labeled data is valuable, capturing user activities may help in understanding the cognitive subprocesses underlying text annotation.