関連度計算補正により認識率向上を図った会話文音声理解システム

摘要

In this research, it is an aim to realize the natural conversation with a computer and the human being. Therefore, the technology of the speech recognition is indispensable. However, it is difficult to get high recognition in the current speech recognition, and it can't realize the natural conversation with a computer and the human being. Then, it is purpose to construct the Conversation Speech Understanding System to plan improvement of the speech recognition precision by revision of the language processing with Concept-Base and Degree of Association for a recognition result provided from speech recognition system, it is suggested the technique of the Conversation Speech Understanding System.%本研究では，コンピュータと人間との自然な会話の実現を目指している．そのためには音声認識の技術が欠かせない．しかしながら，現在の音声認識では高い認識率を得るのが困難であり，コンピュータと人間との自然な会話は実現できない．そこで，音声認識装置から得られた認識結果に対して，更に概念べ-スと関連度計算を用いて言語処理的な補正を行うことで音声認識精度の向上を図る会話文音声理解システムの構築を目的とし，その手法を提案する．