1 |
The phonetics and phonology of Hong Kong English: a study of fricatives
|
|
|
|
BASE
|
|
Show details
|
|
3 |
Overview of GermEval Task 2, 2019 shared task on the identification of offensive language
|
|
|
|
BASE
|
|
Show details
|
|
4 |
Discriminative feature modeling for statistical speech recognition ...
|
|
|
|
BASE
|
|
Show details
|
|
6 |
Automatisches Spracherkennungssystem zur Quantifizierung der Nase und Nasennebenhöhlen auf die Sprachverständlichkeit ...
|
|
|
|
BASE
|
|
Show details
|
|
7 |
Sociophonetic factors in automatic speech recognition: a study on American English
|
|
|
|
Abstract:
The thesis addresses the issue of automatic speech recognition and claims that today it is almost on par with the accuracy of human speech recognition. The first part of the work discusses human speech recognition. The main theory that explains its efficiency and the effect of a number of sociophonetic factors is the exemplar theory. The second part of the thesis discusses the most advanced automatic speech recognition currently available and the technology it is based on. Due to a lack of academic research on the performance of ASR systems and on how modern systems cope with language variation the third part of the thesis is an experiment that aims to fill the gap in research and test the ASR systems developed by Google and Apple. The results of the experiment demonstrate that overall these systems are sensitive to language variation and some of the sociophonetic factors that this variation is associated with. There is a bias towards specific dialect region, namely, a positive bias towards the Western states of the USA, and also recording quality and gender biases, with a positive bias towards female speakers in the Google ASR (the latter is statistically insignificant). The age did not appear to have any effect when it was in the numeric continuous format, but there was a slight bias against the older (60+ y.o.) participants. The binary factor of knowledge of other languages did not demonstrate any effect. These findings support the initial hypothesis that ASR systems are currently sensitive to language variation with its sociophonetic factors and have room for further improvement.
|
|
Keyword:
Automatische Spracherkennung
|
|
URL: https://nbn-resolving.org/urn:nbn:de:bsz:25-freidok-1493573 https://www.freidok.uni-freiburg.de/dnb/download/149357 https://doi.org/10.6094/UNIFR/149357 https://freidok.uni-freiburg.de/data/149357
|
|
BASE
|
|
Hide details
|
|
9 |
Lexical and language modeling for Russian large vocabulary continuous speech recognition ...
|
|
|
|
BASE
|
|
Show details
|
|
10 |
Blind speech separation in distant speech recognition front-end processing
|
|
|
|
BASE
|
|
Show details
|
|
11 |
Dynamic language models for hybrid speech recognition ; Dynamische Sprachmodelle für Hybride Spracherkennung
|
|
|
|
BASE
|
|
Show details
|
|
13 |
Automatische Verarbeitung der Morphologie des Albanischen ; Automatic processing of Albanian morphology
|
|
|
|
BASE
|
|
Show details
|
|
20 |
Evaluation of Natural Language and Speech Tool for Italian : International Workshop, EVALITA 2011, Rome, January 24-25, 2012, Revised Selected Papers
|
|
|
|
UB Frankfurt Linguistik
|
|
Show details
|
|
|
|