Skip navigation

The benefits of contextual information for speech recognition systems

The benefits of contextual information for speech recognition systems

Kinch, Martin W. ORCID: 0000-0002-4737-7118, Melis, Wim J.C. ORCID: 0000-0003-3779-8629 and Keates, Simeon ORCID: 0000-0002-2826-672X (2018) The benefits of contextual information for speech recognition systems. In: The 10th Computer Science and Electronic Engineering Conference (CEEC), 19 - 21 September 2018, University of Essex, Colchester. (In Press)

[img]
Preview
PDF (Final Author Manuscript)
21076 KINCH_The_Benefits_of_Contextual_Information_(Final)_2018.pdf - Accepted Version

Download (125kB) | Preview
[img]
Preview
PDF (Updated Author Manuscript)
21076 KINCH_The_Benefits_of_Contextual_Information_2018.pdf - Updated Version

Download (114kB) | Preview
[img]
Preview
PDF (Author Accepted Manuscript)
21076 KINCH_The_Benefits_of Context_in_Human_Communication_2018.pdf - Accepted Version

Download (95kB) | Preview
[img]
Preview
PDF (Acceptance Email)
21076 KINCH_Acceptance_Email_2018.pdf - Additional Metadata

Download (86kB) | Preview

Abstract

This paper demonstrates the significance of using contextual information in machine learning and speech recognition. While the benefits of contextual information in human communication are widely known, their significance is rarely explored or discussed with a view to their potential for improving speech recognition accuracy. The presented research primarily focuses on an undertaken empirical study that looks at how context affects human communication and understanding. During the study, comparisons between human communication with and without context, have shown overall recognition improvements of over 30% when contextual information is provided. The study has also investigated the importance of the former/middle/latter part of a word towards recognition. These results show that the first two-thirds of a spoken word are key for humans to correctly infer a word. The conclusions from the performed study are then drawn upon to identify useful types of context that can help a machine’s understanding, and how such contextual information can be gathered in speech recognition and machine learning systems. This paper shows that context is not only highly important for human communication, but can easily provide a wealth of information to enhance computational systems.

Item Type: Conference or Conference Paper (Paper)
Uncontrolled Keywords: machine learning, contextual information, speech recognition, natural language processing, context- aware computing, artificial intelligence
Subjects: T Technology > TA Engineering (General). Civil engineering (General)
Faculty / Department / Research Group: Faculty of Engineering & Science
Faculty of Engineering & Science > Department of Engineering Science
Faculty of Engineering & Science > Future Technology and the Internet of Things
Last Modified: 31 Jan 2019 14:59
Selected for GREAT 2016: None
Selected for GREAT 2017: None
Selected for GREAT 2018: None
Selected for GREAT 2019: None
URI: http://gala.gre.ac.uk/id/eprint/21076

Actions (login required)

View Item View Item

Downloads

Downloads per month over past year

View more statistics