Speech understanding system

Japanese: 音声理解システム - おんせいりかいシステム（英語表記）speech understanding system

A system that processes input speech information and aims to accurately extract the meaning it carries. It is one of the major research areas of artificial intelligence technology in the sense that it aims to elucidate the mechanism of speech information processing that humans have. Research on speech recognition aims to accurately extract the characteristics of input speech at the sound level and distinguish and distinguish it from others, but in speech understanding, it is not necessarily necessary to accurately recognize each sound, but the aim is to accurately grasp and understand the content that is intended to be conveyed. It consists of phases such as acoustic processing, word processing, syntactic processing, and semantic processing, but the latter half overlaps almost entirely with the field of natural language processing (→ natural language understanding), and recognition of the dialogue situation with the speaker is an important factor in content understanding. For this reason, information such as grammatical knowledge that defines the speaker's sentence structure and conceptual knowledge that defines the conceptual structure between words, and knowledge of how to use it, are also essential for accurate speech understanding. Research began in the United States in the early 1970s, and systems such as the HARPY system and HEARSAY-II system at Carnegie Mellon University were developed. Furthermore, knowledge acquisition technology has made rapid advances thanks to machine learning, which improves inference capabilities from large amounts of data, and by the 2010s it had been installed in Internet browsers, smartphones, and other devices, becoming available for everyday use by many general users.

Source: Encyclopaedia Britannica Concise Encyclopedia About Encyclopaedia Britannica Concise Encyclopedia Information

Japanese:

入力音声情報を処理し，それが担う意味内容を正確に抽出することを目的としたシステム。人間がもつ音声情報処理のメカニズムの解明を目指すという意味で，人工知能技術の大きな研究領域の一つである。音声認識の研究は，入力音声の特徴を音レベルで正確に抽出し，ほかと識別・判別することを目的とするが，音声理解では，必ずしも一音一音を正確に認識できなくともよく，伝達が意図された内容を的確に把握・理解することを目的とする。音響処理，単語処理，構文処理，意味処理などのフェーズからなるが，後半はほぼ自然言語処理（→自然言語理解）の領域と重なっており，発話者との対話状況の認識が内容理解の重要な要因である。このため，発話者の文構造を規定する文法知識，単語間の概念構造を規定する概念知識などの情報とその利用知識も，正確な音声理解には欠かせない。1970年代初頭にアメリカ合衆国で研究が開始され，カーネギーメロン大学の HARPYシステムや HEARSAY-IIシステムなどが開発された。また，大量のデータから推論能力を高める機械学習により知識獲得技術が急速に進歩し，2010年代にはインターネットのブラウザやスマートフォンなどにも搭載され，多くの一般ユーザーが日常的に利用できるようになった。

出典　ブリタニカ国際大百科事典小項目事典ブリタニカ国際大百科事典小項目事典について　情報

<<: Voice word processor - Onseiwapuro

>>: Speech understanding