Using Shallow Semantic Parsing and Relation Extraction for Finding Contradiction in Text
Minh Quang Pham, Le-Minh Nguyen, Akira Shimazu · Institutional Repositories DataBase (IRDB) · 2013
The problem of text representation is an important issue in textual inference tasks. Given the fact that full predicate-logic analysis is not practical in wide-coverage semantic processing, using shallow semantic representations is an intuitive and straightforward approach. Previous work on finding contradiction in text incorporate information derived from predicate-argument structures as features in supervised machine learning frameworks. In contrast to previous work, we explore the use of shallow semantic representations for contradiction detection in a rule-based framework. We address the low-coverage problem of shallow semantic representations by using a backup module which relies on binary relations extracted from sentences for contradiction detection. Evaluation experiments conducted on standard data sets indicated that using the backup module increases the coverage of contradiction phenomena for the contradiction detection system. Our system achieves better recall and F1 score for contradiction detection than most of baseline methods, and the same recall as a state of the art supervised method for the task.