Team CMU-LTI at TREC 2023 Tip-of-the-Tongue Track
Published in Text Retrieval Conference (TREC), participant paper, 2023
Luís Borges, Jamie Callan, and Bruno Martins
This TREC participant paper describes a movie search pipeline for tip-of-the-tongue queries: descriptions that can be long, uncertain, or partially inaccurate. The system combines dense passage retrieval with GPT-4 re-ranking over candidate movie titles, comparing re-ranking strategies and ways to match query sentences to Wikipedia content.