Abstract: In natural language processing, scripts encode knowledge of stereotypical event sequences encountered in text and can be used to improve text comprehension and draw inferences that are naturally made by human readers. Schank introduced the idea of scripts in the 1970's as a form of symbolic knowledge; however, Chambers and Jurafsky (2008) and other recent work has attempted to learn probabilistic script models from large corpora of raw text. We have recently developed an improved approach to learning statistical scripts using Long Short Term Memory (LSTM) recurrent neural networks (which have recently demonstrated impressive performance on machine translation and many other tasks). We evaluate the approach using Chambers and Jurafsky's 'narrative cloze' task in which an event (represented as a verb and its arguments) is deleted, and the system must correctly infer it from the other events in the document, as well as with crowdsourced human judgements of the system's inferences. LSTMs demonstrate significantly improved performance compared to previous methods.

Short Bio: Raymond J. Mooney is a Professor in the Department of Computer Science at the University of Texas at Austin. He received his Ph.D. in 1988 from the University of Illinois at Urbana/Champaign. He is an author of over 150 published research papers, primarily in the areas of machine learning and natural language processing. He was the President of the International Machine Learning Society from 2008-2011, program co-chair for AAAI 2006, general chair for HLT-EMNLP 2005, and co-chair for ICML 1990. He is a Fellow of the American Association for Artificial Intelligence, the Association for Computing Machinery, and the Association for Computational Linguistics and the recipient of best paper awards from AAAI-96, KDD-04, ICML-05 and ACL-07.