Cross-genre Document Retrieval: Matching between Conversational and\n Formal Writings

This paper challenges a cross-genre document retrieval task, where the\nqueries are in formal writing and the target documents are in conversational\nwriting. In this task, a query, is a sentence extracted from either a summary\nor a plot of an episode in a TV show, and the target document consists of\ntranscripts from the corresponding episode. To establish a strong baseline, we\nemploy the current state-of-the-art search engine to perform document retrieval\non the dataset collected for this work. We then introduce a structure reranking\napproach to improve the initial ranking by utilizing syntactic and semantic\nstructures generated by NLP tools. Our evaluation shows an improvement of more\nthan 4% when the structure reranking is applied, which is very promising.\n