An active learning approach for statistical spoken language understanding

摘要: In general, large amount of segmented and labeled data is needed to estimate statistical language understanding systems. recent years, different approaches have been proposed reduce the segmentation labeling effort by means unsupervised o semi-supervised learning techniques. We propose an active approach estimation models that involves transcription, a small data, along with use raw data. this learn component Spoken Dialog System. Some experiments show appropriateness our are also presented.