摘要
随着智能化产品的不断推广,众多新成果、新技术应用于各领域。其中智能语音交互技术被广泛应用于各产品之中,因此自动语音识别(简称ASR)技术近年来得到了飞速的发展,各标准协会及研究机构也相应发布了相关ASR的测评方法及标准。文章探讨字错率(简称WER)及句错率(简称SER)这两项指标在ASR测评中的应用,浅谈两项指标的关系及区别,通过实验分析提出指导意见,确保测评的可靠性和客观性。
With the increasing promotion of intelligent products,more and more new achievements and technologies are used in various fields.As intelligent voice interaction technology is widely used in a variety of products,the technology of automatic speech recognition(short for ASR)has boomed in recent years.Many standards associations and research establishments published evaluation methods and standards of ASR of their own.In this paper,we discussed the applications of word error rate(short for WER)and sentence error rate(short for SER)in ASR evaluation,analysis on the relationship and difference between the two indicators.Experiments and analyses were carried out to conclude guidance and ensure the reliability and objectivity of ASR evaluation.
出处
《现代传输》
2020年第1期61-64,共4页
Modern Transmission
