Type

Text

Type

Thesis

Advisor

Choi, Yejin | Skiena, Steven | Fodor, Paul.

Date

2014-12-01

Keywords

Computer science | Arborescence, Evolving Entities, Implicit arguments, Language Understanding, Semantic Parsing

Department

Department of Computer Science.

Language

en_US

Source

This work is sponsored by the Stony Brook University Graduate School in compliance with the requirements for completion of degree.

Identifier

http://hdl.handle.net/11401/77308

Publisher

The Graduate School, Stony Brook University: Stony Brook, NY.

Format

application/pdf

Abstract

In this work, we generate activity diagrams of the recipe text automatically, with the help of narrative event chains. Cooking recipes, that have predominantly imperative sentences, present unique challenges in the form of frequent argument drops and co-reference resolution over evolving (or merging) entities. Here, we introduce an unsupervised approach to automatically recover the recipe flow graphs. Then we illustrate the usefulness of our learned models via narrative cloze task; for the automatic evaluation of our learned models, we make use of one of the oft-used task in Discourse analysis, namely the sentence re-ordering and illustrate their utility. To report the effectiveness of our diagramming algorithm, we report the Precision-Recall metric based on the small set of gold-standard annotations that we made. To evaluate a diagramming algorithm on a large scale we need a lot of human annotations. To this end we propose a design for a Game with a purpose (GWAP) that could help us in crowd-sourcing the human annotations. We describe the game design with regards to ease of game play and the considerations that could be taken to keep the player motivated. This game could be used to build a richly annotated recipes corpus. We leave the task of building the game as a future work. | 45 pages

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.