Two-stream indexing for spoken web search

• DOI: 10.1145/1963192.1963364
• Publication Date: 2011-03-29T12:02:18Z
• Authors: Jitendra Ajmera; Anupam Joshi; Sougata Mukherjea; Nitendra Rajput; Shrey Sahay; Mayank Shrivastava; Kundan Srivastava
• Keywords: Audio mining

This paper presents two-stream processing of audio to index the content for Spoken Web search. The first stream indexes meta-data associated with a particular document. is usually very sparse, but accurate. therefore results in high-precision, low-recall index. second uses novel language-independent speech recognition generate text be indexed. Owing multiple languages and noise user generated on Web, accuracy such systems not high, thus they result low-precision, high-recall attempts use these two complementary streams combined increase precision-recall performance