Namhee Kwon

Maximum entropy models for FrameNet classification

The development of FrameNet, a large database of semantically annotated sentences, has primed research into statistical methods for semantic tagging. We advance previous work by adopting a Maximum Entropy approach and by using previous tag information to find the highest probability tag sequence for a given sentence. Further we examine the use of sentence level syntactic pattern features to increase performance. We analyze our strategy on both human annotated and automatically identified frame elements, and compare performance to previous work on identical test data. Experiments indicate a statistically significant improvement (p<0.01) of over 6%.

Multidimensional text analysis for eRulemaking

To support rule-writers, we are developing techniques to automatically analyze large number of public comments on proposed regulations. A document is analyzed in various ways including argument structure, topics, and opinions. The individual results are integrated into a unified output. The experiments reported here were performed on comments submitted to the Environmental Protection Agency in response to their proposed rule for mercury regulation.

FrameNet-based semantic parsing using maximum entropy models

As part of its description of lexico-semantic predicate frames or conceptual structures, the FrameNet project defines a set of semantic roles specific to the core predicate of a sentence. Recently, researchers have tried to automatically produce semantic interpretations of sentences using this information. Building on prior work, we describe a new method to perform such interpretations. We define sentence segmentation first and show how Maximum Entropy re-ranking helps achieve a level of 76.2% F-score (answer among topfive candidates) or 61.5% (correct answer).

Information acquisition using multiple classifications

Given a large collection of documents, we often need to extract various aspects of information that may be integrated to form a coherent overall picture. Especially for subjective documents addressing a single topic, traditional summarization techniques are limited in differentiating and clustering similar information. We apply multiple classifications to handle diverse aspects, including subtopic identification, keyword extraction, argument structure analysis, and opinion classification, in order to provide a summarized overview of the collection, complete with distributional information. From this overall summary, system users can effectively obtain more fine-grained information. Our methods for individual modules significantly outperform the baseline and achieve human-level agreement.

A Semi-Automatic Evaluation Scheme: Automated Nuggetization for Manual Annotation

In this paper we describe automatic information nuggetization and its application to text comparison. More specifically, we take a close look at how machine-generated nuggets can be used to create evaluation material. A semiautomatic annotation scheme is designed to produce gold-standard data with exceptionally high inter-human agreement.

Integrating semantic frames from multiple sources

Semantic resources of predicate-argument structure have high potential to enable increased quality in language understanding. Several alternative frame collections exist, but they cover different sets of predicates and use different role sets. We integrate semantic frame information given a predicate verb using three available collections: FrameNet, PropBank, and the LCS database. For each word sense in WordNet, we automatically assign the corresponding FrameNet frame and align frame roles between FrameNet and PropBank frames and between FrameNet and LCS frames, and verify the results manually. The results are avilable as part of ISI’s Omega ontology.