Academia.eduAcademia.edu

Outline

Title
Abstract
Previous Work in Arabic Speech Recognition
Other Data Resources
Evaluation Method
Preliminary Study
Data
Oracle Result: N-Best Output
Future Work
Previous Work
Experimental Framework
Data-Driven Classes
Background
Background on Backoff and Smoothing
Factored Language Model Training Data File Format
Perplexity Results on Callhome Arabic
Using Msa Text Data
Analysis
References

Novel speech recognition models for Arabic

Profile image of JI SvoJI Svo

2002, Johns-Hopkins …

Abstract
sparkles

AI

This research presents innovative speech recognition models tailored for Arabic, addressing unique challenges such as diacritics and lexical ambiguity. By employing various supervised and unsupervised learning approaches, the study evaluates the accuracy and out-of-vocabulary performance of these models, revealing significant improvements in recognition accuracy. The findings suggest that morphologically aware models enhance performance, thereby paving the way for more effective speech processing technologies in Arabic.

Related papers

The Morpho-Phonology-Semantics Interface in Modern Standard Arabic
Cherif TEIMI

International Journal of Language and Literary Studies, 2021

The issue of the interface between linguistic components has got much interest in contemporary linguistics especially after the advancements that took place in the discipline of Conceptual Semantics. In essence, linguistics components have all been considered to be equal in terms of deriving meaning. In this article, the researcher scrutinized how the component of morpho-phonology interfaces with semantics. This aim was achieved through the analysis of intransitive verbs in Modern Standard Arabic (MSA, henceforth). It was found that intransitive verbs serve well in terms of demonstrating the interface between morpho-phonology and semantics. The morphological and phonological alternations of intransitive verbs result in changes in meaning. Cognition has a central role in generating causative verbs from intransitive ones.

View PDFchevron_right
Phonologically Conditioned Morphological Process in Modern Standard Arabic: An Analysis of Al-Ibdaal 'Substitution' in Ftaʕal Pattern Using Prosodic Morphology
Aseel Zibin, Abdel Rahman Mitib Altakhaineh

This paper provides an analysis of Al-ibdaal 'substitution' phenomenon found in Modern Standard Arabic (MSA) derivational processes, particularly, in the pattern ftaʕal. Al-ibdaal 'substitution' is defined in many Arabic morphological books (Al-Raagihi, 1984; Al-Galaayiini, 1991) as removing a letter and replacing it by another. For example, the /t/ in ztaha: is replaced by /d/ resulting in zdaha: 'flourished'. The paper analyzed examples of words representing the pattern ftaʕal based on prosodic morphology; an approach used to describe non-concatenative morphological processes as those found in Arabic and other Semitic languages (McCarthy, 1981). The analysis of the words revealed that there is actually no substitution of sounds and the change that takes place in that pattern is phonologically motivated. The paper recommends the reconsideration of the concept of Al-ibdaal 'substitution' in the pattern ftaʕal in Arabic morphological books.

View PDFchevron_right
Word Reading in Arabic: Influences of Diacritics and Ambiguity
Ali Idrissi

2019

In the Arabic orthographic system, consonants and long vowels are written, while short vowel diacritics are generally omitted. Although this situation is well-known, the specific role that diacritics play in the processing of written Arabic words by adult readers remains understudied. Reading studies incorporating tasks such as lexical decision are well-established in the literature regarding other languages such as English, but relatively few such studies have been carried out on Arabic word reading. The Arabic orthographic system has three diacritic symbols which can be used to represent the short vowel phonemes /u/, /i/ and /a/ (respectively: ِ ــ َ ــ ُ ــ , with the dashes indicating the consonant symbol above or below which the diacritic would appear). A vowel diacritic appearing above or below a consonant symbol indicates that as pronounced, the consonant is followed by the short vowel. Other diacritics are used in order to indicate the absence of such a short vowel (as when ...

View PDFchevron_right
The Import of the Syllable in Phonological Analysis: Epenthesis and Deletion in Mostaganem Spoken Arabic
Arab World English Journal (AWEJ)

The study at hand explores the crucial position that the syllable holds in phonological analysis and theorizing by highlighting the analytical issues that might follow the exclusion of the syllable from phonological study. Effectively, the present study attempts to answer three main research questions, namely 1. Can rule-based phonology provide a satisfactory account of non-assimilatory processes such as epenthesis and deletion in Mostaganem Spoken Arabic? 2. Can a syllable-devoid account of epenthesis and deletion in Mostaganem Spoken Arabic be satisfactory? 3. Is the syllable indispensable in accounting for epenthesis and deletion in Mostaganem Spoken Arabic? Indeed, the present study demonstrates that a syllable-devoid analysis provides unsatisfactory accounts of certain phonological epenthesis and deletion. In order to attain the aim of this study, instances of epenthesis and deletion in the Mostaganem spoken variety of Algerian Arabic are analysed. Such instances of epenthesis and deletion were obtained from the recorded speech of a randomly selected sample of fifty native speakers of Mostaganem Spoken Arabic. The findings of this study reveal the indispensability of the syllable in describing and explaining epenthesis and deletion in MTG by comparing a syllable-devoid account and a syllable-based account of such processes. Effectively, syllable-devoid account pertaining to The Sound Pattern of English (Chomsky & Halle, 1968) tradition proved short of explanations as to why epenthesis and deletion take place in Mostaganem spoken Arabic as contrasted to a syllable-based account which provides adequate and convincing explanations of those processes.

View PDFchevron_right
UC Merced Proceedings of the Annual Meeting of the Cognitive Science Society Title The time-course of morphological, phonological and semantic processes in reading Modern Standard Arabic Publication Date The time-course of morphological, phonological and semantic processes in reading Modern Stand...
Sami Boudelaa

2001

We investigate deverbal noun and verb morphology in Modern Standard Arabic (MSA), using two masked priming experiments in which the morphological, orthographic and semantic relationship between prime and targets are varied in four SOA conditions (32 ms, 48 ms, 64 ms, and 80 ms). Results show that early in the visual processing of MSA deverbal nouns and verbs, the role played by morphological structure (word patterns and roots), is significantly different from that played by orthographic and semantic factors. Additionally, while word pattern effects are transient and overlap with root morpheme effects only in the early stages of processing, effects of the root are reliably present throughout the recognition process.

View PDFchevron_right
The time-course of morphological, phonological and semantic processes in reading Modern Standard Arabic
Sami Boudelaa

Proceedings of the Twenty- …, 2001

We investigate deverbal noun and verb morphology in Modern Standard Arabic (MSA), using two masked priming experiments in which the morphological, orthographic and semantic relationship between prime and targets are varied in four SOA conditions (32 ms ...

View PDFchevron_right
Conflicts of Vowel Governed Constraints in Arabic Derivation
Prof. Dr. Fareed H Al-Hindawi

Arabic language, literature & culture, 2021

The paradigm of constraints that is proposed by optimality theory as a substitute to the rules paradigm has influenced all the fields of linguistics. Different phonological phenomena in different many languages have been investigated in terms of optimality theory. The phonological structures of Arabic suffer from some changes like deletion, substitution and other processes that underlie the systematic derivation (Al-ishtiqaaq) of vowels. In terms of Optimality Theory, this paper investigates two issues: First, the conflicted faithfulness and markedness constraints that govern vowels through derivation, and second, how these constraints are ranked. Finding out such constraints can provide the storage of the universal constraints by new ones and discovering how Arabic ranks these universal constraints. The current study adopted the mechanism that is proposed by the optimality theory. At the same time, the paper suggested some constraints that has been tested and formulated according to the theory in question. The ranking of these constraints also have been investigated. The study led to demonstrating the existence of the suggested constraints that work during derivation. It also represented how these constraints are ranked. Some of the discovered constraints are already found in the universal storage of constraints and others are peculiar to Arabic. It is found that the SAK constraint (the absence of the Arabic short vowels) plays a fundamental role in vowels derivation.

View PDFchevron_right
A lexical phononology study of Modern Standard Arabic (MSA)
Mohamed Akidah

2012

This is a descriptive study of the phonology and morphology of Modern Standard Arabic (MSA) spoken in Cairo, Egypt. The study, conducted within the framework of the Lexical Phonology (LP) theory, aims at examining and describing in detail the nature of the relationship between the phonology and morphology of the MSA lexicon with a view to determine what role morphological and phonological rules play in the word formation processes and organization of the MSA lexicon. In view of the arguments presented by proponents of the LP theory with re pect to the supposedly universal phonological and morphological principles of languag , this study sets out to investigate and determine, the extent to which the e uni ersal principles are applicable in MSA. The study is guided by our research problem, which is to investigate the nature of the relationship between phonology and morphology in the organization of the MSA lexicon. It seeks to achieve se eral objecti es· key among them being to conduc...

View PDFchevron_right
The Routledge Handbook of Arabic Linguistics
Zainab Hermes

2017

View PDFchevron_right
A logical approach to Arabic phonology
Patrick Blackburn

1991

Abstract Logical approaches to linguistic description, particularly those which employ feature structures, have generally treated phonology as though it was the same as orthography. This approach breaks down for languages where the phonological shape of a morpheme can be heavily dependent on the phonological shape of another, as is the case in Arabic.

View PDFchevron_right

References (56)

  1. W-xalAS:M-noun+masc-sg:S-xalAS:R-xlS:P-CaCAC W-lAzim:M-adj+masc-sg:S-lAzim:R-Azm:P-CVCiC W-nitkallim:M-verb+subj-1st-plural:S-itkallim:R-klm:P-iCCaCCiC W-carabi:M-adj+masc-sg:S-carabi:R-crb:P-CaCaCi W-cala$An:M-prep:S-cala$An:R-cl$:P-CaCaCAn W-humma:M-pro+nom-3rd-plural:S-humma:R-hm:P-CuCma W-cayzIn:M-pple-act+plural:S-cAyiz:R-cwz:P-CVyiC W-%ah:M-%ah:S-%ah:R-%ah:P-%ah W-il+mukalmaB:M-noun+fem-sg+article:S-mukalmaB:R-klm:P-CuCaCCaB W-tibqa:M-verb+subj-2nd-masc-sg:S-baqa:R-bqq:P-CaCa W-bi+il+*FOR:M-bi+il+*FOR:S-bi+il+*FOR:R-bi+il+*FOR:P-bi+il+*FOR W-*FOR:M-*FOR:S-*FOR:R-*FOR:P-*FOR
  2. W-cala$An:M-prep:S-cala$An:R-cl$:P-CaCaCAn W-humma:M-pro+nom-3rd-plural:S-humma:R-hm:P-CuCma W-biysaggilu:M-verb+pres-3rd-plural:S-saggil:R-NULL:P-NULL
  3. E.T. Abdel-Massih. An Introduction to Egyptian Arabic. The University of Michigan Press, Ann Arbor, 1975.
  4. Yaser Al-Onaizan and Kevin Knight. Translating named entities using monolingual and bilingual resources. In Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics (ACL), pages 400-408, 2002.
  5. International Phonetic Alphabet. http://www.arts.gla.ac.uk/IPA/ipachart.html.
  6. S. Badawi and M. Hinds. Dictionary of Egyptian Arabic. Librairie du Liban, 1987.
  7. A. L. Berger, S.A. Della Pietra, and V.J. Della Pietra. A maximum entropy approach to natural language processing. Computational Linguistics, 22(1):39-71, 1996.
  8. P. Beyerlein. Discriminative model combination. In Proc. ICASSP, pages 481-484, 1998.
  9. P. Beyerlein, W. Byrne, J.M. Huerta, S. Khudanpur, B. Marthi, J. Morgan, N. Peterek, J. Picone, D. Ver- gyri, and W. Wang. Towards language independent acoustic modeling. In Proc. ICASSP, 2000.
  10. J. Bilmes and G. Zweig. The Graphical Models Toolkit: An open source software system for speech and time-series processing. Proc. IEEE Intl. Conf. on Acoustics, Speech, and Signal Processing, 2002.
  11. J. A. Bilmes. Graphical models and automatic speech recognition. In R. Rosenfeld, M. Ostendorf, S. Khudanpur, and M. Johnson, editors, Mathematical Foundations of Speech and Language Process- ing. Springer-Verlag, New York, 2003.
  12. M. Brent, S.K. Murthy, and A. Lundberg. Discovering morphemic suffixes: a case study in MDL induction. In Proceedings of 5th Internationl Worksh on Artificial Intelligence and Statistics, 1995.
  13. P.F. Brown, V.J. Delle Pietra, P.V. deSouza, J.C. Lai, and R.L. Mercer. Class-based n-gram models of natural language. Computational Linguistics, 18(4):467-479, 1992.
  14. T. Buckwalter. http://www.xrce.xerox.com/competencies/content-analysis/arabic/info/translit- chart.html.
  15. B. Byrne, J. Hajic, P. Ircing, F. Jelinek, S. Khudanpur, P. Krbec, and J. Psutka. On large vocabulary continuous speech recognition of highly inflectional language -Czech. In Proceedings of Eurospeech, 2001.
  16. S. F. Chen and J. Goodman. An empirical study of smoothing techniques for language modeling. In Arivind Joshi and Martha Palmer, editors, Proceedings of the Thirty-Fourth Annual Meeting of the Association for Computational Linguistics, pages 310-318, San Francisco, 1996. Association for Computational Linguistics, Morgan Kaufmann Publishers.
  17. S. F. Chen and J. Goodman. An empirical study of smoothing techniques for language modeling. Tech- nical Report Tr-10-98, Center for Research in Computing Technology, Harvard University, Cambridge, Massachusetts, August 1998.
  18. CP1256. http://www.microsoft.com/globaldev/reference/sbcs/1256.htm.
  19. T. Schultz D. Kiecza and A. Waibel. Data-driven determination of appropriate dictionary units for korean LVCSR. In Proceedings of ICASSP, 1999.
  20. K. Darwish. Building a shallow arabic morphological analyser in one day. In Proceedings of the ACL Workshop on Computational Approaches to Semitic Languages, Philadelphia, PA, 2002.
  21. F. Debili, H. Achour, and E Souissi. De l'étiquetage grammatical à la voyellation automatique de l'arabe. Correspondances de l'Institut de Recherche sur le Maghreb Contemporain, 17, 2002.
  22. S. Deerwestr, S.T. Dumais, G.W. Furnas, T.K. Landauer, and R. Harshman. Indexing by latent semantic analysis. Journal of the American Society of Information Science, 41:391-407, 1990.
  23. H. DeJean. Morphemes as necessary concepts for structures: discovery from untagged corpora. In Workshop on paradigms and grounding in natural language learning, pages 295-299, Adelaide, Aus- tralia, 1998.
  24. Y. A. El-Imam. Synthesis of arabic from short sound clusters. Computer, Speech, and Language, 15:355-380, 2001.
  25. G. Zavagliakos et al. The BBN Byblos 1997 large vocabulary conversational speech recognition sys- tem. In Proceedings of ICASSP, 1998.
  26. J. Billa et al. Multilingual speech recognition: The 1996 byblos callhome system. In Proceedings of Eurospeech, pages 363-366, 1997.
  27. J. Billa et al. Arabic speech and text in Tides Ontap. In Proceedings of HLT, 2002.
  28. J. Billa et al. Audio indexing of broadcast news. In Proceedings of ICASSP, 2002.
  29. N. Friedman and D. Koller. Learning Bayesian networks from data. In NIPS 2001 Tutorial Notes. Neural Information Processing Systems, Vancouver, B.C. Canada, 2001.
  30. Ya'akov Gal. An HMM approach to vowel restoration in Arabic and Hebrew. In Proceedings of the Workshop on Computational Approaches to Semitic Languages, pages 27-33, Philadelphia, July 2002. Association for Computational Linguistics.
  31. E. Gaussier. Unsupervised learning of derivational morphology from inflectional lexicons. In Proceed- ings of the ACL Workshop on Unsupervised Learning in Natural Language Processing, University of Maryland, 1999.
  32. P. Geutner. Using morphology towards better large-vocabulary speech recognition systems. In Pro- ceedings of ICASSP, pages 445-448, 1995.
  33. H. Glotin, D. Vergyri, C. Neti, G. Potamianos, and J. Luettin. Weighting schemes for audio-visual fusion in speech recognition. In Proc. ICASSP, 2001.
  34. J. Goldsmith. Unsupervised learning of the morphology of a natural language. Computational Lin- guistics, 2001.
  35. M. Huckvale and A.C. Fang. Using phonologically-constrained morphological analysis in continuous speech recognition. Computer, Speech and Language, 16:165-181, 2002.
  36. ISO. http://www.iso.ch/cate/35040.html.
  37. R. Iyer. Improving and predicting performance of statistical language models in sparse domains. PhD thesis, Boston University, 1998.
  38. F. Jelinek. Statistical Methods for Speech Recognition. MIT Press, 1997.
  39. P. Geutner K. C,arki and T. Schultz. Turkish LVCSR: towards better speech recognition for agglutina- tive languages. In Proceedings of ICASSP, 2000.
  40. O. Karboul-Zouari. Die hocharabische Sprache und Romanisierung ihrer Schrift. Technical report, University of Karlsruhe, 1999. Senior Project Report.
  41. Kevin Knight and Jonathan Graehl. Machine transliteration. In Philip R. Cohen and Wolfgang Wahlster, editors, Proceedings of the Thirty-Fifth Annual Meeting of the Association for Computational Linguistics and Eighth Conference of the European Chapter of the Association for Computational Lin- guistics, pages 128-135, Somerset, New Jersey, 1997. Association for Computational Linguistics.
  42. S.L. Lauritzen. Graphical Models. Oxford Science Publications, 1996.
  43. S. Martin, J. Liermann, and H. Ney. Automatic bigram and trigram clustering for word classes. Speech Communication, pages 19-37, 1998.
  44. J.A. Nelder and R. Mead. A simplex method for function minimization. Computing Journal, 7(4):308- 313, 1965.
  45. M.F. Porter. An algorithm for suffix stripping. Program, 14(3):130-137, 1980.
  46. Qalam. http://eserver.org/langs/qalam.txt.
  47. CAT scheme. http://almashriq.hiof.no/general/400/410/418/classical_arabic_transliteratio.
  48. P. Schone. Knowledge-free induction of inflectional morphologies. In Proceedings of NAACL, Pitts- burgh, PA, 2001.
  49. P. Schone. Towards Knowledge-Free Induction of Machine-Readable Dictionaries. PhD thesis, Uni- versity of Colorado at Boulder, 2001.
  50. P. Schone and D. Jurafsky. Knowledge-free induction of morphology using latent semantic analysis. In Proceedings of CoNLL, Lisbon, Portugal, 2000.
  51. Bonnie Glover Stalls and Kevin Knight. Translating names and technical terms in arabic text. In Pro- ceedings of the 1998 COLING/ACL Workshop on Computational Approaches to Semitic Languages, 1998.
  52. A. Stolcke. SRILM-an extensible language modeling toolkit. In Proc. Int. Conf. on Spoken Language Processing, Denver, Colorado, September 2002.
  53. Unicode. http://www.unicode.org/.
  54. D. Vergyri, S. Tsakalidis, and W. Byrne. Minimum risk acoustic clustering for multilingual acoustic model combination. In Proc. ICSLP, 2000.
  55. E.W.D. Whittaker and P.C. Woodland. Particle-based language modeling. In Proc. Int. Conf. on Spoken Language Processing, Beijing, China, 2000.
  56. D. Yarowski and R. Wicentowski. Minimally supervised morphological analysis by multimodal align- ment. In Proceedings of the ACL, Hong Kong, 2000.

Related papers

Phonologically conditioned morphological processes in Modern Standard Arabic
Aseel Zibin, Abdel Rahman Mitib Altakhaineh

This paper provides an analysis of Al-ibdaal 'substitution' phenomenon found in Modern Standard Arabic (MSA) derivational processes, particularly, in the pattern ftaʕal. Al-ibdaal 'substitution' is defined in many Arabic morphological books as removing a letter and replacing it by another. For example, the /t/ in ztaha: is replaced by /d/ resulting in zdaha: 'flourished'. The paper analyzed examples of words representing the pattern ftaʕal based on prosodic morphology; an approach used to describe non-concatenative morphological processes as those found in Arabic and other Semitic languages . The analysis of the words revealed that there is actually no substitution of sounds and the change that takes place in that pattern is phonologically motivated. The paper recommends the reconsideration of the concept of Alibdaal 'substitution' in the pattern ftaʕal in Arabic morphological books.

View PDFchevron_right
Vowels and Classification Criteria in Modern Standard Arabic
Zeynep Özkanlı

Zenodo (CERN European Organization for Nuclear Research), 2023

Plagiarism / İntihal: This article has been reviewed by at least two referees and scanned via a plagiarism software. / Bu makale, en az iki hakem tarafından incelendi ve intihal içermediği teyit edildi. Etik Beyan / Ethical Statement: Bu çalışmanın hazırlanma sürecinde bilimsel ve etik ilkelere uyulduğu ve yararlanılan tüm çalışmaların kaynakçada belirtildiği beyan olunur / It is declared that scientific and ethical principles have been followed while carrying out and writing this study and that all the sources used have been properly cited. https://dergipark.org.tr/tr/pub/ihya * This article was extracted from the third section of doctorate dissertation entitled "Arap Fonetiğinin Tarihsel Gelişimi (Historical Development of Arabic Phonetics)".

View PDFchevron_right
Arabic dialect processing
Mona Diab

MEDAR09. April, 2009

View PDFchevron_right
Arabic Morphology
Adrian Macelaru

The following abbreviations are used throughout this essay: acc. = accusative; CA = Classical Arabic; dial. = dialect; ECA = Egyptian Colloquial Arabic; fem. = feminine; gen. = genitive; imp. = imperative; impf. = imperfect; masc. = masculine; MSA = Modern Standard Arabic; nom. = nominative; obl. = oblique; pf. = perfect; pl. = plural; sg. = singular. A note on the Arabic transcription employed: I use symbols in accordance with IPA and common Semitist usage. A macron over a vowel indicates vocalic length. A dot underneath a consonant marks it as an emphatic, i.e., pharyngealized-velarized, with the exception of [. h], which is a voiceless pharyngeal fricative, IPA [ ]. The symbol [j] is IPA [ ]. *I wish to express my appreciation to the following linguists for useful comments and suggestions on a preliminary version of this essay, not all of which I have been able to follow:

View PDFchevron_right
Phonologically conditioned morphological processes in Modern Standard Arabic: An analysis of Al-ibdaal 'substitution' in ftaʕal pattern using prosodic morphology
Abdel Rahman Mitib Altakhaineh

This paper provides an analysis of Al-ibdaal 'substitution' phenomenon found in Modern Standard Arabic (MSA) derivational processes, particularly, in the pattern ftaʕal. Al-ibdaal 'substitution' is defined in many Arabic morphological books as removing a letter and replacing it by another. For example, the /t/ in ztaha: is replaced by /d/ resulting in zdaha: 'flourished'. The paper analyzed examples of words representing the pattern ftaʕal based on prosodic morphology; an approach used to describe non-concatenative morphological processes as those found in Arabic and other Semitic languages . The analysis of the words revealed that there is actually no substitution of sounds and the change that takes place in that pattern is phonologically motivated. The paper recommends the reconsideration of the concept of Alibdaal 'substitution' in the pattern ftaʕal in Arabic morphological books.

View PDFchevron_right
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved