語音文件摘要方法之研究

This three-year research project develops a novel language modeling (LM) framework that is especially suited to spoken document summarization. Under this framework, we explore several spoken sentence modeling and estimation techniques, as well as different ways to incorporate a wide array of acoustic and prosodic features for use in spoken document summarization. In addition, we propose a novel clarity measure for important sentence selection, which can help quantify the thematic specificity of each individual sentence that is deemed to be a crucial indicator orthogonal to the relevance measure provided by the LM-based methods. Finally, an extensive set of empirical evaluations seem to demonstrate the effectiveness of our methods when compared to several existing state-of-the-art supervised or unsupervised summarization methods. Our research results on spoken document summarization and related applications had also been published at several major prestigious international journals and conferences, such as IEEE Transactions on Audio, Speech and Language Processing, Information Processing & Management, Multimedia Tools and Applications, ICASSP, ASRU, Interspeech, EMNLP and ICME, among others. Especially, our papers also won the best paper awards of ROCLING 2012 and 2013. From here on down, we will focus exclusively on our development of a novel LM-based summarization framework and discuss in detail how to leverage the sentence clarity for important sentence selection, which are mainly based on the material published at IEEE Transactions on Audio, Speech and Language Processing.