Towards better keyword search performance on Malay broadcast news data
Haihua Xu, Pham Van Tung, Eng Siong Chng, Haizhou Li · 2014
In this paper we describe approaches to building our recent Malay broadcast news audio retrieval system. This system contains speech-to-text and keyword search subsystems. The speech-to-text system is built aiming at two folds: hybrid vocabulary recognition to tackle out-of-vocabulary keyword search issue and diversified acoustic modeling for effective system combination in keyword searching afterwards. For keyword search part we focus on the strategies such as keyword expansion to alleviate the out-of-vocabulary search issue and system combination to attain better search performance. Last we also introduce our data driven based keyword selection recipe to evaluate our system performance using actual term weighted value (ATWV) metric.