This paper describes the multi-level annotation process of Urdu speech corpus and its quality ass... more This paper describes the multi-level annotation process of Urdu speech corpus and its quality assessment using PRAAT. The annotation of speech corpus has been done at phoneme, word, syllable and break index levels. Phoneme, word and break index level annotation has been done manually by trained linguists whereas syllable-tier annotation has been done automatically using template matching algorithm. The mean accuracy achieved at phoneme and break index label and boundary identification is 79.07% and 89.67% respectively. The quality assessment of word and syllable tiers is still under investigation.
The paper is about the acoustic effect of Urdu phonological rules on Pakistani Urdu speakers'... more The paper is about the acoustic effect of Urdu phonological rules on Pakistani Urdu speakers' English speech. The objective of the study is to discuss the phenomenon of multiple pronunciations of an English word that has the same spellings, meaning, and part of speech but different pronunciations in the English speech of Urdu speakers. Sometimes these alternative pronunciations are considered mispronunciation rather than multiple pronunciations. The primary purpose of this study is to make a boundary-line between mispronunciations and multiple pronunciations of English vocabulary. Thus, an acoustic analysis of Urdu speakers' English speech has been done by collecting speech data of 30 Urdu speakers from the Public sector universities of Pakistan. Consequently, this paper caters to language-dependent variations of Urdu. This paper only deals with three phonological rules, i.e., segment alternation, ellipsis, epenthesis, which become the cause for re-syllabification of English...
This study is a qualitative review of the test match format of Cricket; therefore, different webs... more This study is a qualitative review of the test match format of Cricket; therefore, different websites and online articles were reviewed critically. The study motivation is to find out the reason(s) for the less viewership and business for the test format. Therefore, this study provides the critical analysis of three different cricket formats: test format, one-day format, and t-twenty format. The scope of this study is; Cricket provides business, entertainment, and promotion to the stakeholders. The audience does not like the test format as compared to the t-twenty and one-day match format because the test format requires more time than the other two formats. Moreover, it will prove helpful for researchers, and beginners, who have a great interest in Cricket
This study explores the intra-rater reliability of holistic and rubric-based assessment of essay ... more This study explores the intra-rater reliability of holistic and rubric-based assessment of essay compositions of Pakistani students. Through the variance among the results of scoring, this research proves the reliability of holistic and rubric-based assessment. The data has been collected from 200 BS students of Public Sector in Faisalabad, Pakistan. The students are selected according to convenient sampling procedure. To collect the data, the students have been given a task of writing expository composition on the essay, ‘Smoking’. Afterwards, among the 200 essays, 100 essays have been selected randomly for data analysis. The data has been analyzed by one rater. The rater has scored the tests into two sessions. In the first session, the same tests have been checked holistically and with rubrics. After 10 days interval, in the second session, the same tests have been again scored holistically and with rubrics. All the results have been analyzed through SPSS software. The ANOVA analy...
Automatic Speech Recognition (ASR) is a key component in Human Computer Interaction (HCI) applica... more Automatic Speech Recognition (ASR) is a key component in Human Computer Interaction (HCI) applications. Stability of ASR systems largely depends on accent, gender, age of speakers, background noise and channel variations. In this paper, a study has been conducted to classify five different accents of Urdu language spoken in Pakistan i.e. Punjabi, Urdu, Pashto, Saraiki and Sindhi. Speech data has been collected from native speakers of these accents. The five accents have been classified using mel frequency cepstral coefficient (MFCCs) and feature formants.
This is a sequel work deals with the confirmation of phonological rules for re-syllabification an... more This is a sequel work deals with the confirmation of phonological rules for re-syllabification and multiple pronunciations of Urdu vocabulary, at larger scale, in speech of Pakistani speakers. The motivation of study is the identification of re-syllabification and multiple pronunciations in 10 hours audio-corpus of a native Pakistani Urdu speaker. This speech corpus has identified 103902 words with different annotations. The observations have identified the multiple pronunciations of same tokens (Farooq & Mumtaz, 2016), (Farooq & Mahmood, 2020). These tokens have same parts-of-speech (POS), spellings and meanings but different pronunciations which ultimately becomes the cause of re-syllabification in different contexts. CLE annotated speech corpus has been used as baseline of this research (Mumtaz, et al., 2014), (Habib, Hijab, Hussain, & Adeeba, 2014). Later, the multiple pronunciations and re-syllabification phenomenon have been cross checked in the speech of 29 native Pakistani U...
International Journal of English Linguistics, 2017
The current research is done for the verification of two different claims. According to Kachru, (... more The current research is done for the verification of two different claims. According to Kachru, (2005) that Punjabi English speakers are unable to create distinction between /Ɛ/ and /ae/ front vowels but Bilal et al. (2011) has refused this claim after verifying it in the speech of Punjabi speakers of Sargodha, Pakistan. If Bilal is right than there is a big need to study this claim in broader perspective. Therefore, in the current research, 9720 utterances (of 72 native Punjabi speakers from 12 districts of Punjab, Pakistan) are recorded and analyzed in PRAAT software. Data analysis is done in two steps i.e., (i) auditory analysis is done by listening wave files and (ii) acoustic analysis is based on the measurement of first three formant values (F1, F2, F3) and vowels’ duration. The results clarify that Pakistani Punjabi English speakers have maintained difference in short and long, stressed and unstressed articulation at word initial and medial positions. But the limited number o...
The present work deals with the phonological rules in Urdu language. All these rules have been re... more The present work deals with the phonological rules in Urdu language. All these rules have been reported by considering the multiple pronunciations of a word, which has same spellings and parts of speech (POS). For the confirmation of multiple pronunciations, firstly a word list of 13717 words has been extracted from 10 hours speech corpus of a female native Urdu speaker. Secondly, in order to confirm whether these multiple pronunciations are speaker dependent or language dependent, data from 9 more native speakers have been collected for the confirmation of multiple pronunciations. In this paper, phonological rules related to the segment alternation, segment deletion and segment insertion have been investigated. Analysis reports that (i) segment alternation occurs due to stress, (ii) unstressed articulation causes segment deletion and (iii) segment insertion emerges to break consonant cluster at coda position. Keywords—Urdu phonological rules, multiple pronunciations, segment deleti...
This study deals with the reformation for English Language Learning (ELL) by considering the need... more This study deals with the reformation for English Language Learning (ELL) by considering the need and importance of viva exam. Therefore, it has been conducted in ELL classes of Public Sector University in Pakistan. English language teaching is a crucial issue therefore different methods have been used to teach and evaluate the performance of learners. But the most important element, viva is being ignored in our Public Sector Universities. Strategically, language teaching is a quite different subject but is being taught similar to other subjects which adds more difficulty. Actually, English must be considered a practical subject where learners have to participate in different experimental tasks. It is an observational study where a checklist has prepared by the researcher for collecting objective information. So, it is designed with sixty learners, divided in 3 groups of 20 participants in each. Two groups have been taught by the same teacher, having same curriculum and methodology. Group 1 is informed about the viva in qualifying exams therefore actively participated in spoken activities. But Group 2 is kept uninformed till the end of the course so tries to avoid participation. Group 3 has been asked to give opinion about the research by answering questionnaire. Then, the data interpretation and results concluded that emergence of viva exams and exam fear fact are the important elements to achieve English language proficiency
The paper is about the acoustic effect of Urdu phonological rules on Pakistani Urdu speakers' Eng... more The paper is about the acoustic effect of Urdu phonological rules on Pakistani Urdu speakers' English speech. The objective of the study is to discuss the phenomenon of multiple pronunciations of an English word that has the same spellings, meaning, and part of speech but different pronunciations in the English speech of Urdu speakers. Sometimes these alternative pronunciations are considered mispronunciation rather than multiple pronunciations. The primary purpose of this study is to make a boundary-line between mispronunciations and multiple pronunciations of English vocabulary. Thus, an acoustic analysis of Urdu speakers' English speech has been done by collecting speech data of 30 Urdu speakers from the Public sector universities of Pakistan. Consequently, this paper caters to language-dependent variations of Urdu. This paper only deals with three phonological rules, i.e., segment alternation, ellipsis, epenthesis, which become the cause for re-syllabification of English words. These three foci of research have been selected because the data analysis has confirmed that the 'multiple pronunciation' is mainly occurred due to these three elements. These three categories cover several sub-categories that cover many instances in the data analysis. The data also confirms that phonological variations occur due to stress shifting in Urdu speakers' English speech in Pakistan.
Acoustic Behavior of RP Diphthongs in Pakistani English (PakE), 2017
In this research, the acoustic behavior of RP diphthongs has been analyzed in Pakistani English (... more In this research, the acoustic behavior of RP diphthongs has been analyzed in Pakistani English (PakE). A diphthong is a gliding movement from one vowel to another. Received Pronunciation (RP) has eight diphthongs; comprised with long and short vowel combinations. Contrarily, this condition is not compulsory for PakE as speakers show difference in the articulation. It is understood that two diverse languages are following different approaches. If there are some similarities, those are just coincidence. Presently, it has been proposed that Urdu effects the pronunciation of RP diphthongs. Therefore, two different experimental approaches have been selected for the identification of RP diphthongs and their acoustic behavior in PakE. Firstly, auditory approach has reported vocalic segments by using syllable count technique. Secondly, the identified segments have been acoustically investigated in PRAAT software. Data analysis and results have reported that Urdu language has influenced and transformed the features of PakE. According to the results, two diphthongs have been articulated as monophthongs (/o/ and /e/). Coincidently, four RP diphthongs have been identified as diphthongs (/aɪ/, /ɔɪ/, /aʊ/, /eə/) but only at word medial position. Two diphthongs (/eə/ and /ɪə/) have behaved as merger (/eə/). Whereas, one diphthong has been articulated in three different forms; it may be uttered either as a diphthong (at word medial position as /ɔɪ/ and /ɔe/ at word final position) or as a triphthong (/ʊae/). Although, these segments are not the part of British inventory but they are autonomous phonetic segments of PakE.
Coarticulation is a fundamental feature of the connected speech because phonemes are overlapping,... more Coarticulation is a fundamental feature of the connected speech because phonemes are overlapping, interacting and consequently effecting the vocal tract configuration (Farnetani & Recasens, 2010). This research investigates coarticulation effects in Urdu speech corpus. For this purpose, one hour of Urdu speech corpus has been annotated phonologically at segment level. Various types of coarticulation effects have been observed during manual annotation such as (i) assimilation of sounds (e.g.رکھتا /rək ͪ t̪ɑ:/ as رتّا /rət̪.t̪ɑ:/), (ii) nasalization of the segment (e.g. ماہ /mɑ:h/ as ماں /mɑ̃:/), (iii) aspiration / unaspiration of a segment (e.g. ساتھ /sɑ:t̪ ͪ/ as سات /sɑ:t̪ /), (iv) addition of a segment (e.g. the insertion of schwa in the word فرض /fərz/ as /fərəz/) and (v) deletion of a segment (بہت سے /bohət̪.se:/ as بہسے /bohəse:/), (vi) voicing of adjacent segments (e.g. ان کے /un.ke:/) and devoicing of adjacent segments (e.g. باپ /bɑ:p/ as پاپ /pɑ:p/). The vocal tract of the speaker is the major reason of coarticulation which starts articulating the next phoneme before hardly finishes the earlier segment (Roach, 2004). Contextual effect of the segments is the other main reason for coarticulation and occurs in different ways e.g. assimilation of sounds is triggered by the stress on a specific segment then the other. A voiced segment becomes devoiced and devoiced segment turns into voiced because of the influence of the adjacent segments. For breaking the consonant cluster, vocalic sounds are inserted. Different reasons have been observed for the deletion of a segment e.g. the deletion of a stop is caused by the following fricative segment and unstressed articulation also causes deletion of a segment especially at coda position. Usually, these effects are not audible but can be analyzed in spectrograph. Therefore, in this research, Standard Annotation Process (SAP) has been adopted for resolving these coarticulation issues e.g. (i) voicing and aspiration/unaspiration of a segment is resolved by annotating it phonetically not phonologically (ii) deletion and addition of short vowel have been accepted by consulting multiple pronunciations of the word, (iii) deletion and addition of a long vowel or a consonant was not allowed and (iv) contextual nasalization has been accepted while contrastive nasalization has been refused for rerecording. Moreover, it is also observed that coarticulation can be controlled by providing sufficient articulatory speed and temporal space to each segment during articulation (Farnetani & Recasens, 1999). Coarticulation issues have been catered for corpus annotation as this would become the lexicon for the development of Phonetic Recognizers such as TTS and ASR systems.
This paper describes the multi-level annotation process of Urdu speech corpus and its quality ass... more This paper describes the multi-level annotation process of Urdu speech corpus and its quality assessment using PRAAT. The annotation of speech corpus has been done at phoneme, word, syllable and break index levels. Phoneme, word and break index level annotation has been done manually by trained linguists whereas syllable-tier annotation has been done automatically using template matching algorithm. The mean accuracy achieved at phoneme and break index label and boundary identification is 79.07% and 89.67% respectively. The quality assessment of word and syllable tiers is still under investigation.
The present work deals with the phonological rules in Urdu language. All these rules have been re... more The present work deals with the phonological rules in Urdu language. All these rules have been reported by considering the multiple pronunciations of a word, which has same spellings and parts of speech (POS). For the confirmation of multiple pronunciations, firstly a word list of 13717 words has been extracted from 10 hours speech corpus of a female native Urdu speaker. Secondly, in order to confirm whether these multiple pronunciations are speaker dependent or language dependent, data from 9 more native speakers have been collected for the confirmation of multiple pronunciations. In this paper, phonological rules related to the segment alternation, segment deletion and segment insertion have been investigated. Analysis reports that (i) segment alternation occurs due to stress, (ii) unstressed articulation causes segment deletion and (iii) segment insertion emerges to break consonant cluster at coda position.
This paper describes the multi-level annotation process of Urdu speech corpus and its quality ass... more This paper describes the multi-level annotation process of Urdu speech corpus and its quality assessment using PRAAT. The annotation of speech corpus has been done at phoneme, word, syllable and break index levels. Phoneme, word and break index level annotation has been done manually by trained linguists whereas syllable-tier annotation has been done automatically using template matching algorithm. The mean accuracy achieved at phoneme and break index label and boundary identification is 79.07% and 89.67% respectively. The quality assessment of word and syllable tiers is still under investigation.
The paper is about the acoustic effect of Urdu phonological rules on Pakistani Urdu speakers'... more The paper is about the acoustic effect of Urdu phonological rules on Pakistani Urdu speakers' English speech. The objective of the study is to discuss the phenomenon of multiple pronunciations of an English word that has the same spellings, meaning, and part of speech but different pronunciations in the English speech of Urdu speakers. Sometimes these alternative pronunciations are considered mispronunciation rather than multiple pronunciations. The primary purpose of this study is to make a boundary-line between mispronunciations and multiple pronunciations of English vocabulary. Thus, an acoustic analysis of Urdu speakers' English speech has been done by collecting speech data of 30 Urdu speakers from the Public sector universities of Pakistan. Consequently, this paper caters to language-dependent variations of Urdu. This paper only deals with three phonological rules, i.e., segment alternation, ellipsis, epenthesis, which become the cause for re-syllabification of English...
This study is a qualitative review of the test match format of Cricket; therefore, different webs... more This study is a qualitative review of the test match format of Cricket; therefore, different websites and online articles were reviewed critically. The study motivation is to find out the reason(s) for the less viewership and business for the test format. Therefore, this study provides the critical analysis of three different cricket formats: test format, one-day format, and t-twenty format. The scope of this study is; Cricket provides business, entertainment, and promotion to the stakeholders. The audience does not like the test format as compared to the t-twenty and one-day match format because the test format requires more time than the other two formats. Moreover, it will prove helpful for researchers, and beginners, who have a great interest in Cricket
This study explores the intra-rater reliability of holistic and rubric-based assessment of essay ... more This study explores the intra-rater reliability of holistic and rubric-based assessment of essay compositions of Pakistani students. Through the variance among the results of scoring, this research proves the reliability of holistic and rubric-based assessment. The data has been collected from 200 BS students of Public Sector in Faisalabad, Pakistan. The students are selected according to convenient sampling procedure. To collect the data, the students have been given a task of writing expository composition on the essay, ‘Smoking’. Afterwards, among the 200 essays, 100 essays have been selected randomly for data analysis. The data has been analyzed by one rater. The rater has scored the tests into two sessions. In the first session, the same tests have been checked holistically and with rubrics. After 10 days interval, in the second session, the same tests have been again scored holistically and with rubrics. All the results have been analyzed through SPSS software. The ANOVA analy...
Automatic Speech Recognition (ASR) is a key component in Human Computer Interaction (HCI) applica... more Automatic Speech Recognition (ASR) is a key component in Human Computer Interaction (HCI) applications. Stability of ASR systems largely depends on accent, gender, age of speakers, background noise and channel variations. In this paper, a study has been conducted to classify five different accents of Urdu language spoken in Pakistan i.e. Punjabi, Urdu, Pashto, Saraiki and Sindhi. Speech data has been collected from native speakers of these accents. The five accents have been classified using mel frequency cepstral coefficient (MFCCs) and feature formants.
This is a sequel work deals with the confirmation of phonological rules for re-syllabification an... more This is a sequel work deals with the confirmation of phonological rules for re-syllabification and multiple pronunciations of Urdu vocabulary, at larger scale, in speech of Pakistani speakers. The motivation of study is the identification of re-syllabification and multiple pronunciations in 10 hours audio-corpus of a native Pakistani Urdu speaker. This speech corpus has identified 103902 words with different annotations. The observations have identified the multiple pronunciations of same tokens (Farooq & Mumtaz, 2016), (Farooq & Mahmood, 2020). These tokens have same parts-of-speech (POS), spellings and meanings but different pronunciations which ultimately becomes the cause of re-syllabification in different contexts. CLE annotated speech corpus has been used as baseline of this research (Mumtaz, et al., 2014), (Habib, Hijab, Hussain, & Adeeba, 2014). Later, the multiple pronunciations and re-syllabification phenomenon have been cross checked in the speech of 29 native Pakistani U...
International Journal of English Linguistics, 2017
The current research is done for the verification of two different claims. According to Kachru, (... more The current research is done for the verification of two different claims. According to Kachru, (2005) that Punjabi English speakers are unable to create distinction between /Ɛ/ and /ae/ front vowels but Bilal et al. (2011) has refused this claim after verifying it in the speech of Punjabi speakers of Sargodha, Pakistan. If Bilal is right than there is a big need to study this claim in broader perspective. Therefore, in the current research, 9720 utterances (of 72 native Punjabi speakers from 12 districts of Punjab, Pakistan) are recorded and analyzed in PRAAT software. Data analysis is done in two steps i.e., (i) auditory analysis is done by listening wave files and (ii) acoustic analysis is based on the measurement of first three formant values (F1, F2, F3) and vowels’ duration. The results clarify that Pakistani Punjabi English speakers have maintained difference in short and long, stressed and unstressed articulation at word initial and medial positions. But the limited number o...
The present work deals with the phonological rules in Urdu language. All these rules have been re... more The present work deals with the phonological rules in Urdu language. All these rules have been reported by considering the multiple pronunciations of a word, which has same spellings and parts of speech (POS). For the confirmation of multiple pronunciations, firstly a word list of 13717 words has been extracted from 10 hours speech corpus of a female native Urdu speaker. Secondly, in order to confirm whether these multiple pronunciations are speaker dependent or language dependent, data from 9 more native speakers have been collected for the confirmation of multiple pronunciations. In this paper, phonological rules related to the segment alternation, segment deletion and segment insertion have been investigated. Analysis reports that (i) segment alternation occurs due to stress, (ii) unstressed articulation causes segment deletion and (iii) segment insertion emerges to break consonant cluster at coda position. Keywords—Urdu phonological rules, multiple pronunciations, segment deleti...
This study deals with the reformation for English Language Learning (ELL) by considering the need... more This study deals with the reformation for English Language Learning (ELL) by considering the need and importance of viva exam. Therefore, it has been conducted in ELL classes of Public Sector University in Pakistan. English language teaching is a crucial issue therefore different methods have been used to teach and evaluate the performance of learners. But the most important element, viva is being ignored in our Public Sector Universities. Strategically, language teaching is a quite different subject but is being taught similar to other subjects which adds more difficulty. Actually, English must be considered a practical subject where learners have to participate in different experimental tasks. It is an observational study where a checklist has prepared by the researcher for collecting objective information. So, it is designed with sixty learners, divided in 3 groups of 20 participants in each. Two groups have been taught by the same teacher, having same curriculum and methodology. Group 1 is informed about the viva in qualifying exams therefore actively participated in spoken activities. But Group 2 is kept uninformed till the end of the course so tries to avoid participation. Group 3 has been asked to give opinion about the research by answering questionnaire. Then, the data interpretation and results concluded that emergence of viva exams and exam fear fact are the important elements to achieve English language proficiency
The paper is about the acoustic effect of Urdu phonological rules on Pakistani Urdu speakers' Eng... more The paper is about the acoustic effect of Urdu phonological rules on Pakistani Urdu speakers' English speech. The objective of the study is to discuss the phenomenon of multiple pronunciations of an English word that has the same spellings, meaning, and part of speech but different pronunciations in the English speech of Urdu speakers. Sometimes these alternative pronunciations are considered mispronunciation rather than multiple pronunciations. The primary purpose of this study is to make a boundary-line between mispronunciations and multiple pronunciations of English vocabulary. Thus, an acoustic analysis of Urdu speakers' English speech has been done by collecting speech data of 30 Urdu speakers from the Public sector universities of Pakistan. Consequently, this paper caters to language-dependent variations of Urdu. This paper only deals with three phonological rules, i.e., segment alternation, ellipsis, epenthesis, which become the cause for re-syllabification of English words. These three foci of research have been selected because the data analysis has confirmed that the 'multiple pronunciation' is mainly occurred due to these three elements. These three categories cover several sub-categories that cover many instances in the data analysis. The data also confirms that phonological variations occur due to stress shifting in Urdu speakers' English speech in Pakistan.
Acoustic Behavior of RP Diphthongs in Pakistani English (PakE), 2017
In this research, the acoustic behavior of RP diphthongs has been analyzed in Pakistani English (... more In this research, the acoustic behavior of RP diphthongs has been analyzed in Pakistani English (PakE). A diphthong is a gliding movement from one vowel to another. Received Pronunciation (RP) has eight diphthongs; comprised with long and short vowel combinations. Contrarily, this condition is not compulsory for PakE as speakers show difference in the articulation. It is understood that two diverse languages are following different approaches. If there are some similarities, those are just coincidence. Presently, it has been proposed that Urdu effects the pronunciation of RP diphthongs. Therefore, two different experimental approaches have been selected for the identification of RP diphthongs and their acoustic behavior in PakE. Firstly, auditory approach has reported vocalic segments by using syllable count technique. Secondly, the identified segments have been acoustically investigated in PRAAT software. Data analysis and results have reported that Urdu language has influenced and transformed the features of PakE. According to the results, two diphthongs have been articulated as monophthongs (/o/ and /e/). Coincidently, four RP diphthongs have been identified as diphthongs (/aɪ/, /ɔɪ/, /aʊ/, /eə/) but only at word medial position. Two diphthongs (/eə/ and /ɪə/) have behaved as merger (/eə/). Whereas, one diphthong has been articulated in three different forms; it may be uttered either as a diphthong (at word medial position as /ɔɪ/ and /ɔe/ at word final position) or as a triphthong (/ʊae/). Although, these segments are not the part of British inventory but they are autonomous phonetic segments of PakE.
Coarticulation is a fundamental feature of the connected speech because phonemes are overlapping,... more Coarticulation is a fundamental feature of the connected speech because phonemes are overlapping, interacting and consequently effecting the vocal tract configuration (Farnetani & Recasens, 2010). This research investigates coarticulation effects in Urdu speech corpus. For this purpose, one hour of Urdu speech corpus has been annotated phonologically at segment level. Various types of coarticulation effects have been observed during manual annotation such as (i) assimilation of sounds (e.g.رکھتا /rək ͪ t̪ɑ:/ as رتّا /rət̪.t̪ɑ:/), (ii) nasalization of the segment (e.g. ماہ /mɑ:h/ as ماں /mɑ̃:/), (iii) aspiration / unaspiration of a segment (e.g. ساتھ /sɑ:t̪ ͪ/ as سات /sɑ:t̪ /), (iv) addition of a segment (e.g. the insertion of schwa in the word فرض /fərz/ as /fərəz/) and (v) deletion of a segment (بہت سے /bohət̪.se:/ as بہسے /bohəse:/), (vi) voicing of adjacent segments (e.g. ان کے /un.ke:/) and devoicing of adjacent segments (e.g. باپ /bɑ:p/ as پاپ /pɑ:p/). The vocal tract of the speaker is the major reason of coarticulation which starts articulating the next phoneme before hardly finishes the earlier segment (Roach, 2004). Contextual effect of the segments is the other main reason for coarticulation and occurs in different ways e.g. assimilation of sounds is triggered by the stress on a specific segment then the other. A voiced segment becomes devoiced and devoiced segment turns into voiced because of the influence of the adjacent segments. For breaking the consonant cluster, vocalic sounds are inserted. Different reasons have been observed for the deletion of a segment e.g. the deletion of a stop is caused by the following fricative segment and unstressed articulation also causes deletion of a segment especially at coda position. Usually, these effects are not audible but can be analyzed in spectrograph. Therefore, in this research, Standard Annotation Process (SAP) has been adopted for resolving these coarticulation issues e.g. (i) voicing and aspiration/unaspiration of a segment is resolved by annotating it phonetically not phonologically (ii) deletion and addition of short vowel have been accepted by consulting multiple pronunciations of the word, (iii) deletion and addition of a long vowel or a consonant was not allowed and (iv) contextual nasalization has been accepted while contrastive nasalization has been refused for rerecording. Moreover, it is also observed that coarticulation can be controlled by providing sufficient articulatory speed and temporal space to each segment during articulation (Farnetani & Recasens, 1999). Coarticulation issues have been catered for corpus annotation as this would become the lexicon for the development of Phonetic Recognizers such as TTS and ASR systems.
This paper describes the multi-level annotation process of Urdu speech corpus and its quality ass... more This paper describes the multi-level annotation process of Urdu speech corpus and its quality assessment using PRAAT. The annotation of speech corpus has been done at phoneme, word, syllable and break index levels. Phoneme, word and break index level annotation has been done manually by trained linguists whereas syllable-tier annotation has been done automatically using template matching algorithm. The mean accuracy achieved at phoneme and break index label and boundary identification is 79.07% and 89.67% respectively. The quality assessment of word and syllable tiers is still under investigation.
The present work deals with the phonological rules in Urdu language. All these rules have been re... more The present work deals with the phonological rules in Urdu language. All these rules have been reported by considering the multiple pronunciations of a word, which has same spellings and parts of speech (POS). For the confirmation of multiple pronunciations, firstly a word list of 13717 words has been extracted from 10 hours speech corpus of a female native Urdu speaker. Secondly, in order to confirm whether these multiple pronunciations are speaker dependent or language dependent, data from 9 more native speakers have been collected for the confirmation of multiple pronunciations. In this paper, phonological rules related to the segment alternation, segment deletion and segment insertion have been investigated. Analysis reports that (i) segment alternation occurs due to stress, (ii) unstressed articulation causes segment deletion and (iii) segment insertion emerges to break consonant cluster at coda position.
Uploads