Development of an Artificial Intelligence/Machine Learning Application for Automatic Speech/VoiceAuthentication and Transcription System for improved Performance and Accuracy

开发用于自动语音/语音认证和转录系统的人工智能/机器学习应用程序，以提高性能和准确性

• 批准号: 568491-2021
• 负责人: Tyagi, Purnima
• 金额: $ 1.82万
• 依托单位: Centennial College
• 依托单位国家: 加拿大
• 项目类别: Applied Research and Development Grants - Level 1
• 财政年份: 2021
• 资助国家: 加拿大
• 起止时间: 2021-01-01 至 2022-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=724665
• 关键词: Development; Artificial; Intelligence; Machine; Learning

Avondx has recently developed a biometric voice authentication, voice filtering and transcription systemaddressing the growing need for biometric voice authentication and automatic voice/speech recognitionsystems to provide fast, accurate and secure call center services. It aims to help customer service centers toauthenticate their customer service agents and improve their performance. Their currently used automaticspeech/voice transcription system has been embedded with a Phrase Exchanger to deal with accuracylimitations in transcriber service. The accuracy of speech to text transcription is still one of their majorconcerns, like other speech recognition systems available in the market [6], and it necessitates the developmentand integration of a Natural Language Processing (NLP) based Artificial Intelligence/Machine Learning(AI/ML) solution to further improve semantic and phonetic accuracy along with sentence replacement ratherthan word replacement. To achieve this milestone, Avondx will collaborate with Wearable, Interactive andMobile Technology Access Centre in Health (WIMTACH), Centennial College, to develop an NLP applicationthat scaffold the Phrase Exchanger to recognize and learn from repeated phrases and sentences; and to enhancethe performance of Phonetic Engine, Filter Feature and Language Support Libraries. The proposed project willhelp Avondx to offer highly efficient transcription service to their clients and thereby increase their customerbase and revenue.

Avondx最近开发了一种生物识别语音认证、语音过滤和转录系统，以满足对生物识别语音认证和自动语音/语音识别系统日益增长的需求，从而提供快速、准确和安全的呼叫中心服务。它的目的是帮助客户服务中心认证他们的客户服务代理和提高他们的表现。他们目前使用的自动语音/语音转录系统已经嵌入了一个短语交换器，以处理转录服务的准确性限制。语音到文本转录的准确性仍然是他们的主要关注点之一，就像市场上的其他语音识别系统一样[6]，它需要开发和集成基于自然语言处理（NLP）的人工智能/机器学习（AI/ML）解决方案，以进一步提高语义和语音准确性，沿着句子替换而不是单词替换。为了实现这一里程碑，Avondx将与Centennial College的Wearable，Interactive and Mobile Technology Access Centre in Health（WIMTACH）合作，开发一个NLP应用程序，该应用程序将短语交换器支架化，以识别和学习重复的短语和句子;并增强语音引擎，过滤功能和语言支持库的性能。该项目将帮助Avondx为客户提供高效的转录服务，从而增加客户群和收入。