Improved automatic pronunciation feedback for English learners

改进了英语学习者的自动发音反馈

• 批准号: 488458-2015
• 负责人: Mendivil, Franklin
• 金额: $ 1.78万
• 依托单位: Acadia University
• 依托单位国家: 加拿大
• 项目类别: Engage Grants Program
• 财政年份: 2015
• 资助国家: 加拿大
• 起止时间: 2015-01-01 至 2016-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=586261
• 关键词: Improved; automatic; pronunciation; feedback; English

Fundy Language Analytics, Inc. is a young Nova Scotia company which is developing a novel software product to help language students learn English. The primary goal of the product is to supplement the one-on-one learning of the student by providing speech exercises with automated feedback. This automated feedback is in the form of detecting and highlighting words which are pronounced incorrectly. The prototype software can already detect a limited class of pronunciation errors, however, the company needs to extend the functionality of the speech detection system. One critical limitation is that feedback is simply a pass/fail (i.e. either you said the word correctly, or you didn't). This diminishes the quality of feedback for language learners and also limits the software's ability to automate error correction. For example, once a problem is detected and well understood, videos of correct pronunciation could be provided that get to the core problem that the learner is having. The company doesn't have the expertise to add the needed ability to score the errors. In this new partnership, Dr. Franklin Mendivil, Department of Mathematics and Statistics will bring his expertise in applied image and signal analysis, to bear on Fundy Language Analytics' challenge to provide more functionality to the prototype. He will work closely with the company to build in the ability of the speech detection algorithm to determine how close the student is to speaking a word correctly, and then give them richer feedback based on this knowledge.

Fundy Language Analytics，Inc。是一家年轻的新斯科舍省公司，正在开发一种新颖的软件产品 帮助语言学生学习英语。该产品的主要目标是补充一对一的 通过提供自动反馈的语音练习来了解学生。这种自动反馈在 检测和突出显示的单词的形式是错误的。原型软件可以 但是，已经检测到有限类的发音错误，但是，公司需要扩展功能 语音检测系统。一个关键的限制是，反馈只是通过/失败（即您说的是 这个词正确，或者您没有）。这降低了语言学习者的反馈质量以及限制 该软件能够自动校正自动化。例如，一旦检测到问题， 可以理解，可以提供正确发音的视频，以达到学习者的核心问题 有。该公司没有专业知识来增加得分错误所需的能力。 在这种新的伙伴关系中，数学和统计系富兰克林·门迪维尔博士将带来他的 应用图像和信号分析方面的专业知识，以遵守Fundy Language Analytics的挑战，以提供更多 原型的功能。他将与公司紧密合作以建立演讲能力 检测算法以确定学生正确地说单词的距离，然后给他们 基于这些知识，更丰富的反馈。