We integrate automatic speech recognition (ASR) and question answering (QA)to realize a speech-driven QA system, and evaluate its performance. We adapt anN-gram language model to natural language questions, so that the input of oursystem can be recognized with a high accuracy. We target WH-questions whichconsist of the topic part and fixed phrase used to ask about something. Wefirst produce a general N-gram model intended to recognize the topic andemphasize the counts of the N-grams that correspond to the fixed phrases. Givena transcription by the ASR engine, the QA engine extracts the answer candidatesfrom target documents. We propose a passage retrieval method robust againstrecognition errors in the transcription. We use the QA test collection producedin NTCIR, which is a TREC-style evaluation workshop, and show the effectivenessof our method by means of experiments.
展开▼