Named Entity Recognition and Zero Pronoun Resolution for Question Answering

Hideki Isozaki

We are working on SPIQA, a Spoken Interactive Question Answering system. One of the essential components is named entity recognition that extracts answer candidates such as person names or location names from documents. However, subjects and/or objects are often omitted in Japanese, and we expect resolution of these zero pronouns will be useful. We describe our approaches to these tasks.