Gregor Kališnik (2012) Mapping between subtitle units in movie subtitles: Design of an efficient method by using search algorithms and semantic analysis. EngD thesis.
Abstract
In this bachelor thesis I have developed a solution for the mapping problem between two different subtitles for a movie or episode of a series by using algorithms for finding shortest paths. I have narrowed the search space with classifiers and semantic analysis. Maps can be used for building aligned corpora, adjusting subtitles to different movie releases and finally for clustering subtitles into groups of replaceable subtitles. Developed method is in most cases usable in practice. With properly configured settings the method achieves F score of 88.83% while it's time efficient and more useful in practice version has F score of 72.64%.
Actions (login required)