Automatic Text Skimming for Question Answering within Single Documents for Blind and Dyslexic Individuals

D. Yarrington and K.F. McCoy (USA)


Open-Domain Question Answering, Word clustering, text skimming, Natural Language Processing, Assistive Technology for Learning Disabilites


This paper describes a system to be used by blind and dyslexic individuals for automatically skimming a document. The system will take a question and a single document and return information similar to that which would be gleaned by visually skimming through the document. While the system will take advantage of techniques currently used in open-domain question answering systems for question analysis, the system differs from open-domain QA systems in that answers must be located in single documents and in that it returns a general topology of the document’s content. Thus this system must include novel approaches. We will incorporate results of word clustering analyses into the system. It is hoped that word clusters will allow the system to more closely capture the information garnered from visual skimming. To ensure that the appropriate word clusters are adopted and to learn more about what information individuals gather from skimming through a document, the system will also use information gathered from tracking eye gaze during text skimming experiments. Ultimately, the system will return a series of links that will both serve to guide users to text that is related to the question’s answer and will provide a general roadmap to the document.

Important Links:

Go Back