iTAG: Automatically Annotating Textual Resources with Human Intentions

Abstract

Annotations represent an increasingly popular means for organizing, categorizing and finding resources on the “social” web. Yet, only a small portion of the total resources available on the web are annotated. Work on automatic tag generation algorithms aims to tackle this problem by developing algorithms that attempt to approximate and support human tagging behavior. While existing algorithms largely focus on automatically describing the general topics covered by a resource (such as “career”, “education”), we suggest focusing on a different tagging dimension: i.e. automatically annotating resources with human intentions. Intent annotations aim to describe which goals are referenced in given textual resources (such as “find a job”, “get a degree”), thereby offering a new, interesting perspective on textual resources on the web. We describe a prototype – iTAG – for automatically annotating textual resources with human intent, and investigate the extent to which the automatic analysis of human intentions in textual resources is feasible. For evaluation purposes, we present results from an exploratory study that focused on annotating intent in transcripts of political speeches given by US presidential candidates in 2008.