European language equality : D2.14 technology deep dive – speech technologies

Abstract

D2.14 provides an overview and describes the state of the art and developments within the field of Speech Technologies (ST). This field is interpreted to comprise technologies aimed at the processing and production of the human voice, both, from a linguistic as well as paralinguistic angle. It provides an in-depth account of current research trends and applications in various ST sub-fields, details technical, scientific, commercial and societal aspects, relates ST to the wider fields of NLP and AI and provides an outlook of ST towards 2030. Chapters 3 and 4, presenting the main ST components and the state-of-the-are are divided according to the different sub-fields covered: Automatic Speech Recognition (ASR), Speaker Identification (SID), Language Identification (LID), technologies targeting paralinguistic phenomena and Text to Speech (TTS). Chapter 5 discusses the main gaps in speech technologies related to issues such as data requirements, ST performance, explainability of the critical methods, regulations influencing the pace of development in the field or specific requirements for less-resourced languages. The following chapter presents aspects of the wider impact of ST on society and describes the contributions of speech technologies to Digital Language Equality. Chapters 7-9 outline some breakthroughs needed, the main technology visions and present how ST may fit into and contribute to a wider vision of what may be termed Deep Natural Language Understanding. The deliverable integrates the views of companies and institutions involved in research, commercial exploitation and application of speech technologies.