Question Answering with LCC's CHAUCER-2 at TREC 2007

In TREC 2007, Language Computer Corporation explored how a new, semantically-rich framework for information retrieval could be used to boost the overall performance of the answer extraction and answer selection components featured in its CHAUCER-2 automatic question-answering (Q/A) system. By replacing the traditional keyword-based retrieval system used in (?) with a new indexing and retrieval engine capable of retrieving documents or passages based on the distribution of named entities or semantic dependencies, we were able to dramatically enhance CHAUCER-2’s overall accuracy, while significantly reducing the number of of candidate answers that were considered by its Answer Ranking and Answer Validation modules.

• Type: Conference
• Year: 2007
• Where: TREC
• Authors: Andrew Hickl, Kirk Roberts, Bryan Rink, Jeremy Bensley, Tobias Jungen, Ying Shi, John Williams
• Keywords: Answer Extraction, Answer Selection Components, Traditional Keyword-based Retrieval, TREC 2007, TREC 2008