Titel: Better accuracy of automatic lecture transcriptions by using context information from slide contents
Vortragender: Jonathan Werner
Abstract: Scannability is crucial for academic research: you have to be able to quickly evaluate the usefulness of a given resource by skimming the content and looking for the parts that are specifically relevant to the task at hand. Spoken content (audio or video lectures, conference talks) make this difficult, leaving you "stabbing in the dark". Automatic Speech Recognition could be a solution, but recognizing technical terms is problematic. This thesis explores an approach using lecture material such as slides to create an "interpolated language model" which uses the textual information from these materials to increase recognition accuracy.