Dynamic Segmentation for Efficient Retrieval of Podcasts: The Repping Algorithm

In the following article, we present a method that makes it possible to find specific segments in a podcast from a large collection using a query (keywords or question). What differentiates our method is that there is no segmentation process at the beginning, but rather the segmentation is done dynamically according to the query entered. The core of our method is that for each term a position-based index is spanned over each individual document. These indices are laid over the individual documents like small threads of information. This multitude of threads maps the inner semantic structure of each individual document in the collection. The corresponding response segments are then individually determined according to the query at runtime using this index. Our initial tests have shown that this method significantly outperforms all current podcast-retrieval methods.