A Grapheme to Phoneme Based Text to Speech Conversion Technique in Unicode Language

Author:

Chandamita Nath,Sarma Bhairab

Abstract

Text-to-speech conversion can be done with two approaches: dictionary-based (database) approach and grapheme-to-phoneme (G2P) mapping. One of the drawbacks of this approach is its performance depends on the size of the dictionary or database. In the case of domain specific conversion, a simple rule -based technique is used to play pre-recorded audio for each equivalent token. It is easy to design but its limitation is mapping with the sound database and availability of the audio file in the database. In general, grapheme to phoneme conversion can be used in any domain. Advantages are the limited size of the database required, ease of mapping and compliance with domain. However, G2P suffers from pronounce ambiguity (formation of audio output). This paper will discuss about the grapheme-to -phoneme mapping and its application in text to speech conversion system. In this work, Assamese (an Indian scheduled Unicode language) is used as the experimental language and its performance is analysis with another Unicode language (Hindi). English (ASCII) language will be used as a benchmark to compare with the target language

Publisher

Salud, Ciencia y Tecnologia

Reference21 articles.

1. 1. Arora A, Gessler Luke, Schneider N, (2020), Supervised Grapheme-to-Phoneme Conversion of Orthographic Schwas in Hindi and Punjabi, in Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 7791–7795.

2. 2. Nath C, Sarma B. (2023), Analysis of Inflectional Behavior in Indian Languages using Features Extraction Techniques, 2023 International Conference on Advancement in Computation & Computer Technique, IEEE, 8 June 2023, DOI: 1109/InCACCT57535.2023.10141783

3. 3. Alok Parlikar, Sunayana Sitaram, Andrew Wilkinson and Alan W Black (2016), The Festvox Indic Frontend for Grapheme-to-Phoneme Conversion, Carnegie Mellon University Pittsburgh, USA, https://www.cs.cmu.edu/~awb/papers/LREC16_parlikar.pdf, WILDRE3, W3RD WORKSHOP ON Indian language data: resources and evaluation.

4. 4. Kumar C.S.,Govind.D.Menon, Nijil Chalil, Sethunath R. and Narwaria M (2006), Grapheme to phone conversion for Hindi, Conference on Oroiiental COCOSDA, 2006,Amrita Vishwa Vidyapeetham, Ettimadai, Coimbatore, Tamil Nadu, INDIA.

5. 5. Srikanth Ronanki, Siva Reddy, BajibabuBollepalli (2016), DNN-based Speech Synthesis for Indian Languages from ASCII text , 9th ISCA Speech Synthesis Workshop, September 2016, DOI: 10.21437/SSW.2016-12

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3