Techniques for Summarization of Spoken Document: A Survey

Abstract

In recent times, the availability of data had increased tremendously and so extracting the required information on a particular topic became a tedious task. When the required information is produced in the form of summary, it would be beneficial to number of end users. Information can be in the form of text, audio (speech, music), video and the summarization process can be either extractive or abstractive. Speech summarization helps in building an efficient speech repository thereby improving the efficiency of information retrieval from speech signals. Typical summarization process of spoken document system includes speech recognition, segmentation of spoken sentences, extraction of sentences and sentence compaction. This paper presents a survey of recent techniques for extractive speech summarization that pull out a vital piece of information and discard the unrelated data from a spoken document. Also, the comparison of various approaches that are used to characterize the spoken document and their inherent sentences are made and inferences are discussed.