On Construction of the ASR-oriented Indian English Pronunciation Dictionary
14029Authors
nan
Abstract
AbstractAs a World English, a New English and a regional variety of English, Indian English (IE) has developed its own distinctive characteristics, especially phonologically, from other varieties of English. An Automatic Speech Recognition (ASR) system simply trained on British English (BE) /American English (AE) speech data and using the BE/AE pronunciation dictionary performs much worse when applied to IE. An applicable IEASR system needs spontaneous IE speech as training materials and a comprehensive, linguistically-guided IE pronunciation dictionary (IEPD) so as to achieve the effective mapping between the acoustic model and language model. This research builds a small IE spontaneous speech corpus, analyzes and summarizes the phonological variation features of IE, comes up with an IE phoneme set and complies the IEPD (including a common-English-word list, an Indian-word list, an acronym list and an affix list). Finally, two ASR systems are trained with 120 hours IE spontaneous speech data, using the IEPD we construct in this study and CMUdict separately. The two systems are tested with 50 audio clips of IE spontaneous speech. The result shows the system trained with IEPD performs better than the one trained with CMUdict with WER being 15.63% lower on the test data.
Keywords
Full Content
Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 6593–6598 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 6593 On Construction of the ASR-oriented Indian English Pronunciation Dictionary Xian Huang, Xin Jin, Qike Li, Keliang Zhang Information Engineering University Luoyang City, Henan Province, China a77huang852yi@126.com, 630530474@qq.com, 1037038328@qq.com, kliang99@sina.com Abstract As a World English, a New English and a regional variety of English, Indian English (IE) has developed its own distinctive characteristics, especially phonologically, from other varieties of English. An Automatic Speech Recognition (ASR) system simply trained on British English (BE) /American English (AE) speech data and using the BE/AE pronunciation dictionary performs much worse when applied to IE. An applicable IEASR system needs spontaneous IE speech as training materials and a comprehensive, linguistically-guided IE pronunciation dictionary (IEPD) so as to achieve the effective mapping between the acoustic model and language model. This research builds a small IE spontaneous speech corpus, analyzes and summarizes the phonological variation features of IE, comes up with an IE phoneme set and complies the IEPD (including a common-English-word list, an Indian-word list, an acronym list and an affix list). Finally, two ASR systems are trained with 120 hours IE spontaneous speech data, using the IEPD we construct in this study and CMUdict separately. The two systems are tested with 50 audio clips of IE spontaneous speech. The result shows the system trained with IEPD performs better than the one trained with CMUdict with WER being 15.63% lower on the test data. Keywords: Indian English, phonological variation features, pronunciation dictionary 1. Introduction Indian English (IE) refers to the English language used in the Republic of India, a South Asian nation and member of the commonwealth (McArthur, 2001:290). IE is a “World English” (Kachru, 1985), a “New English” (Platt et al., 1984:2-3), used in the outer circle of the three concentric circles of English (Kachru, 1985:12-15), and is a regional variety of English. In India, English is widely used as a second language — it is the associate language of India, the language of science and technology, one in the “three language formula” of the Indian education system, and also the language of administration and the armed forces, and is used as the link language among people who do not share a common Indian language. The English used in India has developed its own linguistic patterns as a result of the influences of indigenous Indian languages and the socio-cultural background in the country. The deviation from native English are even greater with regard to phonological and phonetic patterns, though differences in lexis and grammar also exist. Table 1 is summarized by Kachru (1983:84) of Bansal (1969) on the intelligibility of IE. Participants Highest (%) Lowest (%) Average (%) 1 IE & Received Pronunciation (RP) speakers (group) 73 67 70 2 IE & RP speakers (cline of intelligibility) 95 53 3 IE &AE speakers 81 72 74 4 IE &German speakers 67 40 57 5 IE speakers & Nigerians …
Content truncated to 500 words
Labeled Text
AbstractAs a World English, a New English and a regional variety of English, Indian English (IE) has developed its own distinctive characteristics, especially phonologically, from other varieties of English. An Automatic Speech Recognition (ASR) system simply trained on British English (BE) /American English (AE) speech data and using the BE/AE pronunciation dictionary performs much worse when applied to IE. An applicable IEASR system needs spontaneous IE speech as training materials and a comprehensive, linguistically-guided IE pronunciation dictionary (IEPD) so as to achieve the effective mapping between the acoustic model and language model. This research builds a small IE spontaneous speech corpus, analyzes and summarizes the phonological variation features of IE, comes up with an IE phoneme set and complies the IEPD (including a common-English-word list, an Indian-word list, an acronym list and an affix list). Finally, two ASR systems are trained with 120 hours IE spontaneous speech data, using the IEPD we construct in this study and CMUdict separately. The two systems are tested with 50 audio clips of IE spontaneous speech. The result shows the system trained with IEPD performs better than the one trained with CMUdict with WER being 15.63% lower on the test data.
Publication Details
- Paper ID:
- 14029
- Year:
- 2020
- Publication:
- TIPSTER
- Publication Type:
- conf
- DOI:
- nan