Sasha Blair-Goldensohn

MEAD - A Platform for Multidocument Multilingual Text Summarization

Abstract This paper describes the functionality of MEAD, a comprehensive, public domain, open source, multidocument multilingual summarization environment that has been thus far downloaded by more than 500 organizations. MEAD has been used in a variety of summarization applications ranging from summarization for mobile devices to Web page summarization within a search engine and to novelty detection.

NewsInEssence: summarizing online news topics

A news delivery and summarization system, acting as a users agent, gathers and recaps news items based on specifications and interests.

Sentiment Summarization: Evaluating and Learning User Preferences

We present the results of a large-scale, end-to-end human evaluation of various sentiment summarization models. The evaluation shows that users have a strong preference for summarizers that model sentiment over non-sentiment baselines, but have no broad overall preference between any of the sentiment-based models. However, an analysis of the human judgments suggests that there are identifiable situations where one summarizer is generally preferred over the others. We exploit this fact to build a new summarizer by training a ranking SVM model over the set of human preference judgments that were collected during the evaluation, which results in a 30% relative reduction in error over the previous best summarizer.

NewsInEssence: a system for domain-independent, real-time news clustering and multi-document summarization

NEWSINESSENCE is a system for finding, visualizing and summarizing a topic-based cluster of news stories. In the generic scenario for NEWSINESSENCE, a user selects a single news story from a news Web site. Our system then searches other live sources of news for other stories related to the same event and produces summaries of a subset of the stories that it finds, according to parameters specified by the user.

Columbia University at DUC 2004

We describe our participation in tasks 2, 4 and 5 of the DUC 2004 evaluation. For each task, we present the system(s) used, focusing on novel and newly developed aspects. We also analyze the results of the human and automatic evaluations.

Building and Refining Rhetorical-Semantic Relation Models

We report results of experiments which build and refine models of rhetoricalsemantic relations such as Cause and Contrast. We adopt the approach of Marcu and Echihabi (2002), using a small set of patterns to build relation models, and extend their work by refining the training and classification process using parameter optimization, topic segmentation and syntactic parsing. Using human-annotated and automatically-extracted test sets, we find that each of these techniques results in improved relation classification accuracy.

Interactive, Domain-Independent Identification and Summarization of Topically Related News Articles

In this paper we present NewsInEssence, a fully deployed digital news system. A user selects a current news story of interest which is useda s a seed article by NewsInEssence to find in real time other related stories from a large number of news sources. The output is a single document summary presenting the most salient information gleaned from the different sources. We discuss the algorithm used by NewsInEssence, module interoperability, and conclude the paper with a number of empirical analyses.

Integrating Rhetorical-Semantic Relation Models for Query-Focused Summarization

We present our recent work on query-focused summarization, focusing on our efforts in building and applying models of rhetorical-semantic relations (RSRs) such as contrast and causality. We overview ongoing work in extracting and evaluating RSR models. We describe our system for query-focused summarization, focusing on an enhanced, feature-based framework. We present results of experiments to measure the impact of both RSR and other features on selection and ordering of summary content. We conclude with an overview of results from the official DUC06 evaluation.

DefScriber: a hybrid system for definitional QA

Much of the effort in Question Answering (QA) has gone into building short answer QA systems, which answer questions for which the correct answer is a single word or short phrase. However, there are many questions which are better answered with a longer description or explanation. Definitional QA is a developing research area [1] concerned with a subclass of these questions, namely questions of the form “What is X?” DefScriber is a fully implemented system that generates multi-sentence definitions to answer such questions from Internet documents, using an innovative combination of goal-driven and data-driven techniques.

A Hybrid Approach for Answering Definitional Questions

We present DefScriber, a fully implemented system that combines knowledgebased and statistical methods in forming multi-sentence answers to open-ended definitional questions of the form, “What is X?” We show how a set of definitional predicates proposed as the knowledge-based side of our approach can be used to guide the selection of definitional sentences. Finally, we present results of an evaluation of definitions generated by DefScriber from Internet documents.