"Summarization and Question Answering Using Text-to-Text Generation"

Kathleen McKeown, Department of Computer Science, Columbia University

This talk will present research on summarization and question-answering over a variety of sources, featuring Columbia's Newsblaster system for online news summarization and question answering over a mixed corpus of news, broadcast news, talks shows and blogs. I will show how text-to-text generation can be used to highlight salient information for summarization and to remove irrelevant information when responding to a question. We are also investigating how text-to-text generation can exploit redundancy in multilingual summarization and question-answering scenarios to improve the fluency of translated sentences.