Interspeech 2020 2020
DOI: 10.21437/interspeech.2020-2827
|View full text |Cite
|
Sign up to set email alerts
|

Development of Multilingual ASR Using GlobalPhone for Less-Resourced Languages: The Case of Ethiopian Languages

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
2
1

Citation Types

0
7
0

Year Published

2021
2021
2024
2024

Publication Types

Select...
5
2

Relationship

1
6

Authors

Journals

citations
Cited by 12 publications
(7 citation statements)
references
References 15 publications
0
7
0
Order By: Relevance
“…Adaptation subsequent to multilingual training was attempted in several ways, but in all cases led to deteriorated performance. Similar observations have for example been made for Ethiopian languages [23]. For computational reasons, we did not explore all the language combinations exhaustively.…”
Section: Acoustic Modellingmentioning
confidence: 86%
“…Adaptation subsequent to multilingual training was attempted in several ways, but in all cases led to deteriorated performance. Similar observations have for example been made for Ethiopian languages [23]. For computational reasons, we did not explore all the language combinations exhaustively.…”
Section: Acoustic Modellingmentioning
confidence: 86%
“…Since the writing systems used by the languages are different, we have converted the training transcriptions and LM training texts of the languages to phone-based representation so as to use data from all the languages. To enable the development of ML speech processing, the phone names are made consistent across languages using the ML phone representation we developed in [15]. Table 3 presents the PER and WER of ML4 E2E ASR systems.…”
Section: Multilingual E2e Asr Systemsmentioning
confidence: 99%
“…For this purpose, we have used GlobalPhone, which provides speech and text data for 22 languages as described in Section 3, together with the speech data of the four Ethiopian languages. As we did in ML4 E2E experiments, we have converted the transcriptions of the GlobalPhone languages to phone-based representation taking advantage of the ML phone representations we have developed in [15] and the available pronunciation dictionar for all the languages. However, the conversion to phone transcriptions could not be successful for Arabic, Thai and Japanese due to the mixed encoding used in their transcriptions.…”
Section: Multilingual E2e Asr Systemsmentioning
confidence: 99%
See 1 more Smart Citation
“…Multilingual automatic speech recognition (ASR) systems have attained significant attention over the past decade. Motivation for multilingual speech recognition includes; (i) having a single unified model capable of recognising speech of diverse languages [1,2] and (ii) using shared language representations to improve ASR performance in the low resource settings [3,4,5,6,7,8,9,10].…”
Section: Introductionmentioning
confidence: 99%