The Text Analysis Conference (TAC, formerly DUC) is a series of evaluation workshops aimed at encouraging research in natural language applications. This talk will describe the follwing three UMD submissions to TAC: (1) the update summarization task, (2) the opinion summarization task, and (3) the textual entailment task. For the update task, we used Trimmer -- a sentence-compression tool that extends the scope of an extractive summarization system by generating multiple alternative sentence compressions (MASC) of the most important sentences in the target documents (Zajic et al. 07). For the textual entailment task (a joint submission by UMD and Stanford University), we used a novel technique based on an automatically generated list of antonym word pairs, instead of using manually-created lists. For opinion summarization, we used word-pair antonymy as an additional feature in Trimmer. I will discuss how our TAC submissions provide new directions for technical survey creation as well as recent work in combining text summarization techniques with citation information.
Saif Mohammad is a Research Associate in the Institute of Advanced Computer Studies at the University of Maryland. In 2008, under the supervision of Dr. Graeme Hirst, he got his Ph.D. in Computer Science from the University of Toronto. Saif's interests are in Natural Language Processing, especially Lexical Semantics. His work focuses on developing monolingual and cross-lingual computational models of semantic distance and lexical-semantic relations such as antonymy for the benefit of various natural language applications, including multidocument summarization. He is also interested in combining the state-of-the-art summarization techniques with citation information and visualization techniques to generate readily-consumable technical surveys.
This talk is part of the CLIP Colloquium Series. For the complete schedule, please visit http://www.umiacs.umd.edu/research/CLIP/colloq/.