An Evaluation of a Spoken Document Retrieval Baseline System by Indexing Confusion Networks

This paper presents a investigate methods for improving the performance of morph-based spoken document retrieval in Finnish by extracting relevant index terms from confusion networks and a baseline spoken document retrieval system in Finnish. Due to its agglutinative structure, Finnish speech cannot be adequately transcribed using the standard large vocabulary continuous speech recognition approaches. The effect of pruning in recognition is analyzed, showing that when recognition speed is increased, the reduction in retrieval performance due to the increase in the 1-best error In this work we apply a recently developed unlimited vocabulary speech recognition system that allows the use of n-gram language models based on morpheme-like sub word units discovered in an unsupervised manner. In addition to word-based indexing, we also propose an indexing based on the sub-word units provided directly by our speech recognizer.